Quantitative trading - page 24

[MetaQuotes]

Artificial intelligence in trading by Dr Thomas Starke | Algo Trading Week Day 6

Artificial intelligence in trading by Dr Thomas Starke | Algo Trading Week Day 6

Dr. Thomas Starke, a prominent speaker, discusses why AI is considered the next big thing in trading during his presentation. He acknowledges that AI and machine learning have existed for a long time, but due to limited compute power, their effective application was challenging. However, recent advancements in technology have drastically improved computational capabilities, enabling substantial algorithms to run efficiently on laptops and in server centers through cloud computing. Dr. Starke highlights the successes of AI in various fields, such as face recognition, image recognition, and natural language processing, which have contributed to the belief that AI can revolutionize finance as well.

Dr. Starke emphasizes that AI and machine learning are not magic bullets but scientific and mathematical tools that require a thorough understanding and application within the finance domain. While finance has scientific aspects, it is predominantly considered an art form. Hence, to harness the potential of AI in finance, one must grasp both the tools and the artistry of the field.

During his talk, Dr. Starke addresses the role of software development and programming skills alongside machine learning and statistical knowledge in applying AI to trading. He highlights the significance of strong software skills, including writing APIs and ensuring system fail-safes, as essential for effectively employing machine learning tools in the market. He argues that while machine learning tools are user-friendly, programming skills and statistical knowledge are critical for practitioners in this field. Furthermore, he addresses the question of whether a PhD is necessary for utilizing machine learning algorithms and asserts that it is not essential as long as individuals have specific goals, conduct thorough research, and are willing to put in the necessary work.

The importance of mentorship in learning AI for trading is another topic discussed by Dr. Starke. He stresses that finding a good mentor can help beginners avoid common mistakes and develop practical knowledge rather than solely relying on theoretical knowledge gained from academic institutions. Dr. Starke emphasizes that anyone can learn AI, but having a mentor who can provide proper guidance is invaluable. He also emphasizes that understanding the underlying markets and economy is more crucial than programming skills, as programming can be learned with proper mentoring.

During his presentation, Dr. Starke also emphasizes the importance of learning programming and quantitative methods in today's trading industry. He highlights that successful traders often possess a strong understanding of mathematics and programming, and those interested in trading can learn these skills relatively quickly. He points out that traders who invest time in learning quantitative methods and machine learning have a better chance of survival when the transition from screen trading to algorithmic trading occurs. However, he emphasizes that having an economic and market edge is crucial and surpasses the edge gained from programming and mathematical skills alone. He also mentions that deep learning requires businesses and individuals to explain their returns, and facing a year of negative returns can pose significant challenges.

Explaining AI algorithms and risk management practices are also discussed by Dr. Starke. He emphasizes the importance of being able to explain AI algorithms, as failure to do so can lead to problems or even withdrawal of funds. He mentions that despite the use of AI and machine learning, risk management practices remain largely unchanged, but it is necessary to explore new ways of managing risk, particularly with the end of the bull run in stocks and bonds. Dr. Starke emphasizes that machine learning is ubiquitous in trading, with various applications such as generating input signals and managing the risk of machine learning models.

Dr. Starke dives into the different models and technologies used in trading, such as principal component analysis (PCA), decision trees, xgboost, deep learning, and reinforcement learning. He discusses their applications in analyzing signal data, managing portfolio risk, and executing trades. He also highlights the importance of risk management systems in increasing geometric returns and replicating successful strategies in other markets. Dr. Starke suggests that good risk management systems can even generate alpha and be considered as long volatility strategies.

Furthermore, Dr. Starke explores how AI can be used to hedge and manage the risk of short volatility strategies in trading, potentially enhancing the alpha generated by such strategies. He emphasizes the importance of curiosity and a healthy appreciation for risk in continuously learning and developing new trading strategies. He advises against relying on out-of-the-box trading platforms and instead encourages coding strategies from scratch to gain a deep learning advantage.

Dr. Starke engages in a discussion about time-based price movements versus price-based market movements. He explains that time-based price movements can be mathematically solved by calculating indicators, while price-based market movements are determined by the underlying economics of the market. Dr. Starke emphasizes the significance of considering the underlying economic reasoning for a trading strategy rather than solely relying on mathematical techniques to outperform the markets. He recommends books by Marcus Lopez, Grinnell, and Kahn for those interested in combining AI with quantitative models in financial markets.

During the presentation, Dr. Starke emphasizes the importance of understanding factor modeling principles, which he believes are similar to machine learning principles. He suggests that understanding these principles can better equip traders to apply machine learning effectively in their systems. Dr. Starke also highlights the importance of defining what constitutes a good trading strategy, as it may not always be the most profitable one. He references books by Ralph Vince, Andreas Klenow, and Mr. Trendful, which provide valuable insights into trading strategies and the psychology behind trading.

Dr. Starke discusses how AI and machine learning can capture nonlinearities in behavioral finance, such as the Keynesian beauty contest. He explains that these nonlinear dynamics can be effectively captured by machine learning, unlike linear regression models. However, he emphasizes that having an economic reasoning behind trading strategies is still important, even if fundamental data is not explicitly used.

Furthermore, Dr. Starke explores the exploitation of certain market inefficiencies that are not necessarily fundamental. He mentions factors such as restrictions on short positions overnight and specific dates like triple reaching or quadruple witching, which can create economic effects in the market that can be capitalized upon. He also mentions market inefficiencies arising from everyday economic activity or illegal market manipulation. Dr. Starke expresses his interest in potential future collaborations but currently has no concrete plans.

In response to a viewer's question about why dreams often fail to materialize, Dr. Starke provides his personal insight. He explains that dreams initially start as concepts and that his dream life does not revolve around simply lying on the beach but rather involves exploration, running his own business, and being self-directed. He emphasizes that aligning one's true aspirations and goals with practical outcomes is crucial. The presentation concludes with the host informing viewers about the limited-time discount on Contra courses and mentioning the final session on applying machine learning in trading scheduled for the next day.

• 00:00:00 The speaker discusses why AI is considered the next big thing in trading. Although AI and machine learning have been around for a long time, there was not enough compute power to run the algorithms effectively. However, in recent years, technology has improved so much that even substantial algorithms can run on a laptop, and the cloud has enabled them to run on server centers. Additionally, there have been successes in other fields that have contributed to the idea that AI is the next big thing, and finance has not been left behind. AI has proved useful in areas such as face recognition, image recognition, and general natural language processing.
  
  
• 00:05:00 Dr. Thomas Starke discusses the potential of artificial intelligence (AI) in finance and how it could be a game-changer, as it allows for new possibilities that were not previously available. He also touches on how AI and machine learning are not a magic bullet but scientific and mathematical tools that must be understood and applied in finance, which is not intrinsically scientific. While finance has some scientific aspects, the majority of it is considered an art form. Therefore, understanding both the tool and the art of finance is essential to the successful utilization of AI.
  
  
• 00:10:00 Dr. Thomas Starke discusses the role of software development and programming skills in addition to machine learning and statistical knowledge when it comes to applying AI to trading. He emphasizes the importance of good software skills, including writing APIs and making systems fail-safe, as they are necessary for applying machine learning tools to the market. He argues that while machine learning tools are easy to use, programming skills and knowledge of stats are crucial for being a practitioner in this field. Dr. Starke also addresses the question of whether a PhD is necessary for applying machine learning algorithms and argues that it is not essential, as long as one has a specific goal and is willing to put in the necessary research and work.
  
  
• 00:15:00 Dr. Thomas Starke discusses the importance of mentorship in learning AI in trading. He emphasizes that finding a good mentor to guide you through the process can help prevent beginners' mistakes. He believes that anyone can learn AI, but it's more important to develop something that works for you practically rather than just theoretical knowledge developed in university. Dr. Starke also highlights that an understanding of the underlying markets and economy is more crucial than programming skills. He argues that one can learn programming as long as they have someone to mentor them properly.
  
  
• 00:20:00 Dr. Thomas Starke discussed the importance of learning programming and quantitative methods in today's trading industry. He stated that most successful traders possess a strong understanding of mathematics and programming, and those who are interested can learn it fairly quickly. He explained that traders who invest their time in learning quantitative methods and machine learning tend to survive in the markets when the shift from screen trading to algorithms occurs. Additionally, he emphasized that economic and market edge is crucial and surpasses an edge in programming and mathematical skills. However, he also mentioned that deep learning requires businesses and individuals to explain their returns, and one year of negative returns may bring in significant challenges.
  
  
• 00:25:00 Dr. Thomas Starke discusses the importance of being able to explain AI algorithms, especially when using machine learning tools in trading. If the algorithm cannot be explained, it can lead to problems or even the withdrawal of funds. He also states that, despite using AI and ML, the risk management practices remain more or less the same, but there is a need to reconsider new ways of managing risk, especially with the end of the bull run of stocks and bonds. Machine learning is everywhere in trading, and there are various applications such as using AI for input signals and using it for risk management of machine learning models, among others.
  
  
• 00:30:00 Dr. Thomas Starke discusses how artificial intelligence (AI) is being used in every step of trading, from analyzing signal data to managing portfolio risk and executing trades. Machine learning and deep learning are used to analyze images and sentiment signals to produce a clear signal, then principal component analysis is used to reduce the dimensionality of inputs for trade signals. Algorithms are then used to determine which input signals should be traded on. For risk management, machine learning is used to manage portfolio risk, which may be superior to classical risk management calculations. Lastly, in execution, linear models, support vectors machines, and reinforcement learning are used to help traders achieve the best execution prices.
  
  
• 00:35:00 Dr. Thomas Starke discusses different models and technologies that can be used in trading, such as PCA, decision trees, xgboost, deep learning, and reinforcement learning. Later, he addresses a question from an experienced algo trader who struggles with scaling their working system and learning new technologies. Dr. Starke suggests focusing on risk management, as it can help increase geometric returns and lead to replicating the strategy in other markets. Good risk management systems can even produce alpha and be considered as long volatility strategies.
  
  
• 00:40:00 Dr. Thomas Starke discusses how artificial intelligence could be used to buffer and hedge the risk of short volatility strategies in trading. He suggests that AI could greatly increase the alpha generated by such strategies. When it comes to motivating oneself to continuously learn and develop new strategies, Dr. Starke emphasizes the importance of curiosity and a healthy appreciation for risk. He also recommends avoiding out-of-the-box trading platforms and instead coding up strategies from scratch in order to develop a deep learning advantage. The interviewer asks Dr. Starke whether he believes in time-based price movement or price-based market movement in trading, and Dr. Starke asks for clarification on the distinction before answering.
  
  
• 00:45:00 Dr. Thomas Starke discusses the difference between time-based price movements and price-based market movements. He notes that time-based price movements can often be solved mathematically through calculating indicators, while price-based market movements are determined by the underlying economics of the market. Dr. Starke emphasizes the importance of looking at the underlying economic reasoning for a trading strategy, rather than just trying to beat the markets with mathematics. He also recommends books such as Marcus Lopez's book and Grinnell and Kahn's Active Portfolio Management for those interested in combining AI with quantitative models in financial markets.
  
  
• 00:50:00 Dr. Thomas Starke stresses the importance of understanding the underlying principles of factor modeling, which he believes are very similar to the principles of machine learning. He suggests that understanding these principles can better equip traders in applying machine learning to their systems. Dr. Starke also highlights the importance of determining what constitutes a good trading strategy since it’s not always the most profitable one, citing examples from Ralph Vince's book, Mathematics of Portfolio Management. He recommends books by Andreas Klenow and Mr. Trendful, as they not only provide valuable insight into trading strategies but also cover the psychology behind trading.
  
  
• 00:55:00 Dr. Thomas Starke discusses how AI and machine learning can capture nonlinearities that occur in behavioral finance. He explains the Keynesian beauty contest as an example of how outcomes can become extremely nonlinear and chaotic, which is a part of using behavioral methods in trading. Machine learning can capture these nonlinear dynamics, unlike linear regression, which is completely unable to do it. However, it's always good to have an economic reasoning behind what you're doing in trading, even if you are not necessarily using fundamental data in your strategies.
  
  
• 01:00:00 Dr. Thomas Starke discusses the possibility of trading a specific portfolio and exploiting certain market inefficiencies that are not necessarily fundamental. He gives examples such as knowing that people are not allowed to hold short positions overnight, which can lead to economic principles that can be exploited in the market. Additionally, he mentions the significance of certain dates such as triple reaching or quadruple witching, which can produce economic effects that arise from the market. He also talks about market inefficiencies that arise from everyday economic activity or illegal market manipulation. Dr. Starke expresses his interest in collaborating again in the future but has no plans for now.
  
  
• 01:05:00 Satwik asks Dr. Thomas Starke about why dreams often fail to materialize. Starke says it's an interesting question and gives his personal insight. He explains that his dream initially was just a concept, not his actual goal, and that his dream life is not just about lying on the beach. He loves exploring things, running his own business, and being self-directed. This, according to him, comes far closer to his true dream. Lastly, the host informs the viewers that all Contra courses are 75% off for limited time and the last session on applying machine learning in trading is tomorrow.

[MetaQuotes]

Current trends in quant finance [Panel Discussion] | Algo Trading Week Day 5

Current trends in quant finance [Panel Discussion] | Algo Trading Week Day 5

Ladies and gentlemen, welcome to today's panel discussion on current trends in quant finance. We have three distinguished domain experts joining us today to share their insights and expertise. Let's introduce our panelists:

First, we have David Jessup, the head of investment risk for EMEA at Columbia Thread Needle Investments. With extensive experience in quantitative research, risk analysis, and portfolio construction, David specializes in cross-asset factor investing and machine learning in investment management. His deep understanding of quantitative strategies and risk management will provide valuable insights into the trends shaping the industry.

Next, we have Dr. Devashes Guava, the director of machine learning and chair of the Center for Research in Technology Business at SP Gen School of Global Management. Dr. Guava's expertise lies in the application of artificial intelligence in economics and finance. His research and knowledge in this field will shed light on the intersection of AI and finance and the implications for quantitative finance.

Lastly, we have Richard Rothenberg, an executive director at Global AI Corporation. Richard brings a wealth of experience from his work at multi-billion dollar hedge funds and global investment banks. With his extensive background in quantitative portfolio management and research, he will provide valuable insights into the practical implementation of quantitative strategies in the financial industry.

Now, let's dive into the discussion on the recent trends that have shaped quant finance. Our panelists unanimously agree that the availability and quality of data have played a significant role in driving the industry forward. Furthermore, advancements in computing power have enabled the construction and analysis of complex models that were not feasible a decade ago.

The panelists highlight the expansion of quant finance beyond equities into other asset classes, including credit, currencies, and crypto trading. They also bring attention to the emerging trend of responsible investing, which is gaining traction in the finance industry. However, they note that data quality in this area still needs improvement. The panelists predict that responsible investing will continue to be a significant factor in finance over the next few years.

Moving on, the panel discusses two major trends in quantitative finance. Firstly, algorithmic trading has expanded into all asset classes, not just equities. Exotic assets are now being traded using algorithmic approaches. Secondly, there has been a substantial increase in alternative data sources, such as sentiment data from news in multiple languages and credit card transactions. The ability to process and analyze this data with advanced analytics and computational power has led to the incorporation of non-financial risk factors, such as environmental and social governance trends, in company valuations.

However, the panel also addresses the challenges of utilizing machine learning in finance. Given the low signal-to-noise ratio and the zero-sum game nature of financial markets, machine learning is not always the ideal tool to solve every problem. The panelists stress the importance of combining machine learning with other methodologies and understanding its limitations. They also clarify the distinction between machine learning and alternative data, as these two concepts are often confused.

Furthermore, the panelists discuss the unique challenges of financial machine learning within the context of market dynamics as a differential game. They highlight the importance of considering the strategic choices made by other market participants when developing trading strategies.

The discussion then shifts to the significance of high-quality data in machine learning models for algorithmic trading. The panelists acknowledge the challenge of cleaning unstructured data and emphasize the importance of starting with linear models to understand the parameters and ensure data quality. They address the issue of noise and sparsity in alternative data, making it more challenging to clean and filter. Additionally, the panelists stress the need to compare and utilize second sources of data to ensure data accuracy.

The panelists further emphasize that trading solutions should be approached as part of defining a strategy in an end-person game with opposing players who have conflicting interests. Traditional modeling methods may not always apply in this context, and the panelists stress the importance of testing different strategies to find the most effective solutions. They also discuss the unique challenges posed by alternative data sets like sustainable development data, which require different methods of analysis and may require aggregating data at lower frequencies to address sparsity. While working with sparse datasets can be challenging, the panelists believe that there are still opportunities to discover valuable signals.

Another key topic of discussion is the importance of understanding the game structure of the market when designing trading systems. The panelists highlight that while smaller players may have more leeway to take risks, larger players in commodities and crypto trading need to approach trading with caution due to the extreme volatility of these markets. They also stress the importance of diversification to mitigate drawdowns, which are significantly high in crypto assets.

The panel takes a step further and challenges the embedded assumptions in traditional finance theory. They argue that assets do not necessarily follow fixed diffusion processes with set mean and variance assumptions. Instead, they emphasize the stochastic nature of volatility and the fluctuation of mean values over time. They propose considering hidden Markov processes to tactically change the mean and standard deviation, leading to better approaches in factor investing and crypto investing. This perspective offers enticing risk-return profiles with the potential for simple diversification.

The discussion then explores various applications of machine learning in the financial industry. The panelists mention using machine learning for sex classification, carbon emission forecasting, and fixing volumes in fixed income markets. They also highlight the evolving focus on ESG factors and the expansion of sustainable development goals, which consider the impact on society as a whole and systemic risk. They consider this expanded taxonomy of risks as a significant factor in financial decision-making, with a potential to be integrated into an ESG factor model.

Another trend discussed is the utilization of committees and task forces to cluster data based on multiple factors. The panelists emphasize the growing importance of natural language processing in understanding local stakeholder sentiment to quantify non-financial risks. These risks, increasingly material to the intangible aspects of a company's balance sheet, are vital to consider in the analysis of financial markets.

Furthermore, the panelists stress the importance of having strong programming skills and statistical knowledge in the field of quantitative finance. They also caution against the pitfalls of repeatedly analyzing the same dataset, emphasizing the need to adapt and prepare for the future of quantitative trading.

Looking ahead, the panelists discuss the importance of keeping up with emerging asset classes, such as carbon and cryptocurrencies. They mention the potential game-changing impact of quantum computing, which could revolutionize encryption algorithms behind cryptocurrencies, although practical applications are yet to be realized. They also touch upon the development of large neural networks and technologies like GPT3, which are touted as pathways to general artificial intelligence. The exponential growth in hardware and software capacity shows no signs of slowing down, and the panelists anticipate a future convergence of high-performance computing, quantum computing, and AI in the field of quant finance.

In conclusion, the panelists predict a future characterized by the expansion of hardware and software capacity, leading to the development of general-purpose trading robots. These robots will possess the ability to extract and interpret data from diverse sources, including social media, utilizing image understanding, language understanding, and semantic understanding, among others. They highlight the importance of embracing new technologies and methodologies to stay ahead of the curve and adapt to the evolving landscape of quant finance.

The panel discussion concludes with the panelists expressing their gratitude to the audience and encouraging the sharing of any unanswered questions. They also announce that tomorrow's session will focus specifically on machine learning and trading, inviting attendees to join and continue exploring this fascinating field.

Thank you all for being part of today's insightful panel discussion on current trends in quant finance.

• 00:00:00 The moderator introduces the three domain experts for the day's panel discussion on current trends in quant finance. The first panelist, David Jessup, is the head of investment risk for EMEA at Columbia Thread Needle Investments and has extensive experience in quantitative research, risk analysis and portfolio construction, especially in cross-asset factor investing and machine learning in investment management. The second panelist, Dr. Devashes Guava, is the director of machine learning and chair of the Center for Research in Technology Business at SP Gen School of Global Management, specializing in the application of artificial intelligence in economics and finance. Lastly, Richard Rothenberg, an executive director at Global AI Corporation, has worked at multi-billion dollar hedge funds and global investment banks and has immense experience in quantitative portfolio management and research.
  
  
• 00:05:00 In this section, the panelists discuss the trends that have shaped quant finance recently. The availability and quality of data have been significant factors driving the industry. Additionally, the increasing power of computing has allowed for complex models to be built and analyzed in ways that were not possible even a decade ago. The panelists note that quant finance is expanding beyond equities into other asset classes, such as credit, currencies, and crypto trading. They bring up the new trend of responsible investing, which is gaining traction in the finance industry, but the data quality in this area is still lacking. The panelists predict that responsible investing will be a significant factor in finance over the next few years.
  
  
• 00:10:00 In this section, the panel discusses two of the major trends in quantitative finance. The first is the expansion of algorithmic trading into all asset classes, not just equities, including exotic assets. The second trend is the significant increase in alternative data sources, such as sentiment data from news in multiple languages and credit card transactions, and the ability to process this data with advanced analytics and computational power. This has led to an increase in non-financial risk factors, like environmental and social governance trends, that impact a company's valuation. However, the panel also highlights the challenges of using machine learning in finance, given the low signal-to-noise ratio and zero-sum game of the financial market. Bayesian statistics is another area where machine learning is being combined to come up with distributional forecasts.
  
  
• 00:15:00 In this section, the panelists discuss the benefits and limitations of machine learning in finance. One of the main points made is that machine learning is a useful tool, but it should not be the only tool in the trading box, as it is not the right tool to solve every problem. Another challenge that arises with machine learning is that it is often difficult to know when it will go wrong and it can be hard to train a model to tell when it doesn't know. The panelists also differentiate between machine learning and alternative data, stating that they are two separate things that are often confused. Finally, the panelists discuss the challenges of financial machine learning in the context of markets being a differential game that requires a different kind of machine learning, especially when dealing with strategic choices by other players in the game.
  
  
• 00:20:00 The panel discusses the importance of having good quality data for machine learning models in algorithmic trading and the challenge of cleaning unstructured data. While machine learning can be useful for forecasting distributions in short-term trading, it is important to go back to basics and start with linear models to understand the parameters and ensure the quality of the data is good. The panel acknowledges that there is a lot of noise and sparsity in alternative data, making it harder to clean and filter. Additionally, they talked about the difficulty of fixing data outliers, and the need to compare and use second sources of data to ensure the accuracy of the data.
  
  
• 00:25:00 Trading solutions are part of a game structure and must be thought of and tested as part of defining a strategy in an end person game with opposing players who have conflicting interests. It's important to keep in mind that traditional modeling methods may not apply in this context, and testing of different strategies is crucial in finding the most effective solution. Additionally, alternative data sets such as sustainable development data require different methods of analysis and may require aggregating data at lower frequencies to deal with sparsity. While sparse datasets like these can be challenging to work with, there are still opportunities for finding valuable signals.
  
  
• 00:30:00 The panelists discuss the importance of considering the game structure of the market before designing any trading system. While smaller players might be able to afford to gamble, it is not the case with larger players in commodities and crypto trading. The panelists discuss the markets that are most interesting for machine learning algorithms, mentioning crypto as a fascinating area where new challenges can be found. They advise not to focus on just one asset class or algorithm and consider the importance of alternative data sources to achieve profitable trading. Markets, in general, go through phases of being more or less predictable, and signals that were once overused can come back to gain relevance if few market players use them. Factors such as market volatility and a stable underlying data generating process may make markets more friendly for machine learning algorithms.
  
  
• 00:35:00 The discussion focuses on the obstacles faced in deploying quant strategies for crypto investing. Dr. Gughah explains that one of the principal problems is that traditional finance people have never taken an interest in crypto, as it is usually thought of as a computer geek or video game kind of field. Moreover, the extreme volatility of cryptocurrencies is also a major concern, as peak to trough drawdowns of 85 to 90 percent are unthinkable for any kind of fund manager or retail investor. For any kind of financial trading ecosystem to develop in crypto, it is necessary to recognize it as an alternative asset class and create a portfolio that is diversified enough to dampen down the drawdowns, which are a result of high correlation among crypto assets.
  
  
• 00:40:00 The speaker discusses the need to give up the idea that assets follow fixed diffusion processes with a set mean and variance assumption, which is a common embedded assumption in the field of finance. The speaker explains that volatility is stochastic and that mean changes a lot over time. Thus, it is necessary to assume that the mean and standard deviation are driven by a hidden Markov process for the state to change tactically, which is a big jump in traditional finance theory. The speaker suggests that understanding the stochastic process that drives returns can lead to better approaches to factor investing and crypto investing, resulting in very enticing risk-return profiles with simple diversification.
  
  
• 00:45:00 The panel discusses machine learning's various applications in the financial industry, such as using it for sex classification, carbon emission forecasting, and fixing volumes in fixed income markets. They also mention using it as an input in the investment process, rather than solely a trading technique. Another topic they cover is the evolution of ESG to sustainable development goals, which focuses not only on the impact on shareholders but also on society as a whole and systemic risk. This expanded taxonomy of risks includes factors beyond carbon emissions and also considers governance. They discuss this as a significant factor in financial decision-making, who say it can be thought of as an ESG factor model.
  
  
• 00:50:00 The panelists discuss two interesting trends in the field of quantitative finance. First, the use of committees and task forces to cluster data based on 17 factors and the growing importance of natural language processing to understand local stakeholder sentiment to quantify non-financial risks that are more and more material to the intangible aspect of the balance sheet for companies. Secondly, they discuss the importance of having good programming skills, statistical knowledge, and being aware of the pitfalls of looking at the same data set multiple times to prepare for the future of quantitative trading.
  
  
• 00:55:00 The panelists discuss the importance of keeping abreast of new asset classes that may become tradable, including carbon and cryptocurrencies. One area that could be a game-changer is quantum computing, which could revolutionize encryption algorithms behind cryptocurrencies. Although there are no practical applications yet, some big hedge funds are investing in the quantum area. Furthermore, they talk about the development of very large neural networks and GPT3, which is being touted as a way to general artificial intelligence. The increase in hardware and software capacity shows no signs of slowing down, and some expect deep learning to take over the world.
  
  
• 01:00:00 The panel predicts that the future of quant finance lies in the continuing expansion of hardware and software capacity that will enable the development of general-purpose trading robots. These robots would be able to extract data from various sources such as social media and make sense of it to take trading decisions. They won't be limited to numerical machine learning but rather will have image understanding, language understanding, semantic understanding, etc. Another focus area is quantum computing, which may become practical in the next five to ten years. The panelists believe that the future will be a convergence of high-performance computing, quantum computing, and AI. They think that as we start to incorporate more data and models, the future lies in the convergence of these technologies.
  
  
• 01:05:00 The panelists discuss the exponential growth of new tools and techniques in the field of quant finance, which will likely render many functions and jobs obsolete within the next five to ten years. They emphasize the importance of preparing and accelerating the incorporation of new technologies to stay ahead of the curve. The panelists conclude by thanking the audience and encouraging the sharing of any unanswered questions, as tomorrow's session will focus on machine learning and trading.

[MetaQuotes]

Using sentiment and alternative data in trading [Panel Discussion] | Algo Trading Week Day 4

Using sentiment and alternative data in trading [Panel Discussion] | Algo Trading Week Day 4

Ladies and gentlemen, thank you for joining us today for this exciting panel discussion on the use of sentiment and alternative data in trading. Before we begin, I have an important announcement to make.

I am thrilled to announce the launch of a new certification program, the Certification in Sentiment Analysis and Alternative Data in Finance (CSAF). This program has been specifically designed for financial professionals who are looking to advance their careers in trading and investment decision-making using modern methods such as news sentiment analysis and alternative data.

The CSAF program will cover various aspects of news analytics, sentiment analysis, and alternative data required in finance. It will be taught by leading experts in the fields of algorithmic trading, sentiment analysis, quantitative modeling, and high-frequency trading. These experts bring a wealth of knowledge and experience to the program, ensuring that participants receive top-notch education and training.

The program will delve into topics such as understanding sentiment analysis, leveraging alternative data sources, incorporating sentiment data into prediction models, and utilizing AI and machine learning techniques for market analysis. Participants will gain valuable insights into the role of sentiment and alternative data in trading and learn how to unlock the potential of these resources to improve financial outcomes.

In addition to the certification program, I am pleased to announce that a comprehensive handbook on alternative data will be released in the spring of 2022. This handbook will serve as a valuable resource for professionals in the field, providing in-depth information on the various types of alternative data and their applications in finance.

Now, let's turn our attention to today's panel discussion. Our esteemed panelists, including Dr. Cristiano Arbex Valle, Professor Gautam Mitra, Dr. Matteo Campolmi, and Dr. Ravi Kashyap, will be sharing their insights on the use of sentiment and alternative data in trading. They will discuss what alternative data is, why it is important, and how it can be effectively utilized to make informed trading decisions.

As we all know, news events often have a significant impact on asset prices, and sentiment data can play a crucial role in predicting future outcomes. The panelists will shed light on how sentiment data can be processed quickly and converted into numerical data for use in mathematical models, providing valuable information that is not typically captured by traditional market data.

Furthermore, our panelists will explore the challenges and opportunities associated with alternative data. They will discuss the emergence of alternative data sources, the need for rigorous data processing techniques, and the importance of avoiding overfitting while identifying signals within vast amounts of information.

During the panel discussion, we encourage you to actively participate by asking questions and engaging with our panelists. Your input and insights are highly valued, and we look forward to creating an enriching and interactive session.

Before we begin, I would like to express my gratitude to all of you for joining us today. Your presence and enthusiasm contribute to the success of events like these. I would also like to remind you to follow us on social media and wish the organizers a happy 11th anniversary.

Now, without further ado, let's commence our panel discussion on sentiment and alternative data in trading. Thank you.

As the panel discussion begins, our panelists dive into the topic of sentiment and alternative data in trading, sharing their valuable insights and experiences. They highlight the impact of incorporating news analytics and sentiment as additional input features in prediction models, emphasizing the improved results obtained, particularly in predicting asset volatility.

One key point of discussion revolves around the emergence of alternative data and its significance in informing trading decisions. The panelists stress that alternative data introduces new information, such as consumer habits, which can provide valuable insights for investment strategies. They emphasize the importance of coupling data with models, utilizing AI and machine learning techniques to predict market directions and enhance financial outcomes.

The panel takes a moment to acknowledge the moderation of Professor Gautam Mitra, founder and MD of OptiRisk Systems. With his expertise, he ensures a comprehensive exploration of the topic. They delve into the practical applications of sentiment and alternative data in trading, addressing questions regarding its definition, importance, and utilization.

Recognizing that alternative data is a constantly evolving field, the panelists highlight the dynamic nature of this domain. They discuss how what is considered alternative data today may become mainstream in the future, showcasing the continuous progress and innovation within the industry. Their focus remains on leveraging alternative data to gain an edge in finance, with the ultimate goal of maximizing returns.

Amidst the discussion, the panel acknowledges the potential bias present in sentiment data derived from news sources. They offer potential solutions to mitigate this bias, such as utilizing multiple sources and employing various techniques to analyze the data. By doing so, they emphasize the importance of comprehensive and robust data analysis to ensure accurate and reliable insights.

Moving forward, the panelists emphasize the significance of understanding the context and scenarios under which data is collected. They discuss the need for contextual information to provide a nuanced view and build effective algorithms. The panelists also touch upon the idea that biases may not always be negative and can sometimes benefit trading strategies. Their overarching message emphasizes the importance of understanding and working with the available data, even if the data source itself cannot be controlled.

The panel further explores the parameters to consider when analyzing sentiment data for trading purposes. They shed light on the classification of sentiment into positive, neutral, or negative categories by news or sentiment providers. Additionally, they discuss the relevance of considering the volume of news or tweets as a factor in sentiment analysis. The normalization of sentiment based on the average volume of news over a specific time period is also highlighted.

The conversation deepens as the panelists discuss the language-specific nature of sentiment analysis. They emphasize the use of AI and other techniques to parse and analyze text, enabling a deeper understanding of sentiment. Relevance and novelty of news events are identified as crucial factors, with companies receiving news data through subscriptions with content providers, enabling rapid processing.

Wrapping up the panel discussion, the panelists touch upon the time frames used for sentiment indicators. They clarify that sentiment indicators are not aimed at beating the speed of news reaching the market. Instead, they serve as descriptive indicators of how news flow affects stocks over time. The importance of converting text to numerical data is also highlighted, acknowledging the additional layer of processing required for text-based information.

The panelists also discuss the relevance of sentiment data and alternative data sources in trading. They address the question of how many days of sentiment data are relevant, emphasizing that the answer depends on the model's purpose and the type of trading being conducted. The discussion further extends to the performance metrics for alternative data sources, where profitability is identified as a key metric. The panelists explain the demand for historical data and its potential impact on pricing, cautioning that as alternative data sources become more popular, their value may change over time.

To conclude the panel discussion, the panelists share their insights on the challenges and importance of backtesting. They acknowledge the sparsity of historical information for certain alternative data sources, making analysis and backtesting challenging. However, they highlight the availability of statistical models and techniques that can help extrapolate data for backtesting purposes. They stress the significance of comparing the performance of a given data source to not having it, allowing traders to tailor their strategies accordingly. The panel concludes by underscoring that the value of alternative data ultimately depends on its utilization within a specific model.

We now transition to the audience Q&A session, where the panelists address two intriguing questions. The first question revolves around the use of historical data to gain a better understanding of different historical periods. The panel suggests utilizing at least seven times the time interval to obtain a comprehensive understanding of various outcomes. The second question pertains to finding reliable sources of alternative data. The panel recommends having a data scout to explore various sources and identify the best data available for quantitative teams. They highlight the challenge of finding trustworthy data and emphasize that innovative ideas often emerge from small new companies.

Expanding on the discussion, the panelists delve into the potential for small companies that identify unique data sets early on to be acquired by larger firms. They emphasize the importance of intermediaries in data aggregation and the value of derived data sets using proprietary modeling. The conversation further touches upon the impact of country-specific data sets, the identification of regional risks, and the interconnectedness of the global market. Understanding these factors becomes essential in making informed trading decisions.

As the panel draws to a close, the speakers shift their focus to the necessary skills and prerequisites for a career in finance. They emphasize the value of programming languages and a solid understanding of mathematical concepts, as these skills are increasingly crucial in the field. Networking and building connections with professionals are also highlighted, as is the importance of remaining open to diverse opportunities and continuously expanding one's knowledge.

In closing, the speaker reiterates the significance of staying informed about market trends and maintaining objectivity in financial decision-making. She emphasizes the fundamental role of managing finances and encourages attendees to actively engage in the financial industry.

With heartfelt gratitude, the speaker thanks the panelists and the audience for their valuable contributions and concludes the session.

• 00:00:00 The host announces the launch of a new certification program, the Certification in Sentiment Analysis and Alternative Data in Finance (CSAF), designed for financial professionals looking to advance their careers in trading and investment decision-making using modern methods like news sentiment analysis and alternative data. The course will cover various aspects of news analytics, sentiment analysis, and alternative data required in finance, and will be taught by leading experts in algorithmic trading, sentiment analysis, quantitative modeling, and high-frequency trading. The section also features a panel moderated by Professor Gautam Mitra, founder and MD of OptiRisk Systems, with experts Dr. Cristiano Arbex Valle, Professor Gautam Mitra, Dr. Matteo Campolmi, and Dr. Ravi Kashyap discussing the use of sentiment and alternative data in trading.
  
  
• 00:05:00 The speaker is introducing the topic of sentiment and alternative data in finance, which will be discussed in a panel discussion. The panelists will provide insight on what alternative data is, why it's needed, and how to unlock its value. The goal is to use sentiment data to predict future outcomes in finance, as news events often affect asset prices and sentiment data can be processed quickly and converted into numerical data for use in mathematical models. This data is not typically captured by traditional market data, making it a valuable source of information for decision making. A handbook on alternative data will be released in the spring of 2022, and the panel will take questions from attendees.
  
  
• 00:10:00 The panel of experts discusses the use of sentiment and alternative data in trading. They have found that incorporating news analytics and sentiment as an extra input feature in prediction models has led to improved results, particularly in predicting asset volatility. Additionally, they discuss the emergence of alternative data, which introduces new information, such as consumer habits, that can be used to inform trading decisions. They highlight the importance of coupling data with models, using AI and machine learning techniques, to predict market directions and ultimately improve financial outcomes.
  
  
• 00:15:00 The founders of Brain, a research-oriented company, explain their approach to creating proprietary algorithms and methods that extract signals related to financial markets using alternative data sets. They attribute the growing trend of alternative data sets in investing and asset management to the increased availability of data sources and the booming data science industry. While alternative data sets provide additional information for investors to work with in their models, the founders emphasize the need for a rigorous approach to processing a large amount of information and identifying signals without overfitting.
  
  
• 00:20:00 The panel discusses the use of sentiment and alternative data in trading. They cover the questions of what alternative data is, why it's important, and how to use it. They note that alternative data is constantly evolving and that what is considered alternative today may become mainstream in the future. The goal in finance is simple: to make more money. However, knowing when to buy and sell presents a challenge. That's where alternative data can provide an edge. The panel acknowledges bias can be an issue when using news as a source of sentiment data, and solutions such as using multiple sources and techniques to analyze the data are suggested.
  
  
• 00:25:00 The panel discusses the importance of understanding the scenarios under which data was collected and the potential biases present within the data. They note that back testing can be used to see how data has historically performed, but contextual information is necessary to provide a more nuanced view and build better algorithms. The panel also touches on the idea that biases may not always be negative, as they can sometimes benefit trading strategies. Overall, the key takeaway is that while the data source cannot be controlled, the focus should be on understanding and working with the available data.
  
  
• 00:30:00 The panel discusses the parameters to look for when analyzing sentiment data for trading purposes. While news or sentiment providers typically classify sentiment as positive, neutral, or negative, the team notes that the volume of news or tweets can also be a factor to consider. Depending on the provider, sentiment can be quantified as a continuous number or normalized by the average volume of news over a certain time period. The panel also highlights that social media platforms like Twitter can offer an extra element to sentiment analysis by considering who is saying something and identifying key movers with an outsized impact on the markets.
  
  
• 00:35:00 The panel discusses the use of sentiment and alternative data in trading. They note that sentiment is a language-specific factor that can be analyzed by parsing text using AI and other techniques. The panelists also talk about the importance of considering the relevance and novelty of news events, and how companies typically receive news data through subscriptions with news content providers, which can take just a few seconds for processing. Overall, the discussion emphasizes the importance of understanding sentiment and alternative data in building a prediction model for trading.
  
  
• 00:40:00 The panelists discuss using sentiment and alternative data in trading. They talk about the different time frames that can be used for sentiment indicators and how they are not aimed at beating the speed of news reaching the market. The sentiment indicators are meant to provide a descriptive indicator of how the news flow is affecting stocks over time. The panelists also talked about the importance of converting text to numbers and the extra layer of processing required for text data. They mentioned how the use case and frequency of trading can affect the quality and timing of the data used for trading.
  
  
• 00:45:00 The panel discusses the relevance of sentiment data and alternative data sources in trading. The question of how many days of sentiment data is relevant is brought up, to which the answer is that it depends on the purpose of the model and what kind of trading is being done. They move on to discuss the performance metrics for alternative data sources, with the simple answer being how much profit is being made. However, they explain that people generally want as much history as possible at a cheap price, but data sets become cheaper if more people are using them and they are getting commoditized. They also note that using alternative data sources comes with the understanding that the value of the data may change over time.
  
  
• 00:50:00 The panel discusses the challenges of using alternative data for trading and the importance of backtesting. They acknowledge that there is a sparsity of historical information, making it difficult to analyze and backtest certain alternative data sources. However, they suggest that there are statistical models and techniques that can help extrapolate data for backtesting. The panel also emphasizes the importance of comparing how well a given data source performs compared to not having it, and tailoring trading strategies accordingly. When discussing sentiment analysis, they caution that the best approach will vary depending on the specific model and its deployment. Ultimately, the panel agrees that the value of alternative data largely depends on how it is used within a given model.
  
  
• 00:55:00 The panel discusses two questions from the audience. The first question involves the use of historical data and how much of it should be used to gain a better understanding of what could happen in different historical periods. The panel suggests at least 7 times the time interval should be used to get a good idea of different outcomes. The second question refers to finding good sources of alternative data. The panel suggests having a data scout to look at various sources and find the best data available for use by the quant team. They caution that it's not an easy task to find reliable data and that the real source of alternative data are from these small new companies that find innovative ideas.
  
  
• 01:00:00 The panel discusses alternative data and how small companies that identify unique data sets early on have the potential for acquisition by larger firms. The panel also mentions the importance of intermediaries in data aggregation and the value in derived data sets using proprietary modeling. They then move on to discussing the impact of country-specific data sets, breaking down sources of risk, and how the global market is now tightly linked, making it essential to understand regional risks and their potential impact on trading decisions. The section ends with a joke before moving on to the next question.
  
  
• 01:05:00 The panel discussion, the speakers discuss the necessary skills and prerequisites for a course on using sentiment and alternative data in trading. While being comfortable with Python as a programming language is helpful, they emphasize the importance of having basic knowledge in finance and financial models. Additionally, they highlight the value of having access to data sources and being willing to participate and engage in the course. They also address a question about career advice for someone interested in becoming a quantitative research analyst, encouraging the individual to reach out to the faculty for clarification and emphasizing the importance of being open to a broad range of skills and knowledge.
  
  
• 01:10:00 The panelists discuss the skills needed for a career in finance. They suggest learning programming languages, as there are more and more datasets being created, and to develop a good understanding of mathematical concepts. Furthermore, they advise not being afraid of mathematics and programming, as these skills are becoming vital in the field. The panelists also emphasize meeting and networking with as many people as possible, becoming a valuable asset to potential employers, being prepared to take opportunities, and having a solid foundation in math.
  
  
• 01:15:00 The speaker emphasizes the importance of being aware of what is happening in the market and being open to different domains in the financial industry. She advises maintaining objectivity and avoiding being sentimental, as commerce ultimately revolves around money and managing finances. The conversation then concludes with thanks to the panelists and audience, and a reminder to wish the organizers a happy 11th anniversary on social media.
  

[MetaQuotes]

Short selling in the bull market - A Masterclass by Laurent Bernut | Algo Trading Week Day 3

Short selling in the bull market - A Masterclass by Laurent Bernut | Algo Trading Week Day 3

Laurent Bernut is introduced as the founder and CEO of Alpha Secure Capital, as well as a dedicated short seller at Fidelity Investments. The video highlights that he will be leading a masterclass on the topic of short selling, which will last for two hours. It is mentioned that there will be no Q&A session at the end of the masterclass, but viewers are encouraged to ask relevant questions during the session itself. Additionally, the speaker informs the audience about a course on short selling with Python, as well as a complementary book that explains the how and why of short selling. The book is set to be published on October 11th, 2021, and will be available on Amazon.com.

The masterclass begins with Laurent Bernut explaining the key takeaways that participants can expect to gain from the session. He asserts that top picking is bankrupt and emphasizes that short selling is the most valuable skillset for raising a successful fund. Bernut also debunks ten classic myths about short selling, shedding light on the under-researched nature of this discipline. He elaborates on the dynamics of short selling and addresses why even successful market participants struggle with the short side. Sharing personal insights, Bernut emphasizes the crucial role of money management in the course.

Moving forward, Bernut provides an overview of how short selling works and stresses the importance of locating the borrow. He discusses the bankrupt nature of stock picking and advocates for traders to shift their focus to other practices like short selling. Bernut points out that the industry is often fixated on stock pickers, but empirical evidence shows that the majority of active managers underperform their benchmarks consistently. This has led many to abandon stock picking in favor of passive investing and closet indexing. However, Bernut highlights the relevance of short selling during bear markets and the value it brings in terms of downside protection.

Bernut addresses misconceptions about short sellers, dispelling the notion that they destroy pensions and companies. He explains that investors seek long-short vehicles for low volatility, low correlation returns, and downside protection, something that active managers struggle to consistently deliver. Therefore, long picks from mutual fund managers are not as relevant to investors who can achieve similar results passively through exchange-traded funds. Bernut emphasizes that shorting stocks provides protection against downside risk, making the skill of short selling highly sought after, particularly in a bear market.

The speaker delves into the role of short sellers within capitalism and the responsibility of company management. He argues that short sellers, who do not participate in the management of companies, often get blamed for their failures when, in reality, it is poor management that causes the downfall. Bernut highlights the distinction between market value and intrinsic value, explaining that market value is determined by subjective judgments, akin to a beauty contest. He further clarifies that short sellers are not inherently evil speculators but often unveil paradoxes in the market. He acknowledges that regulators frown upon short sellers who engage in market manipulation, but their primary task is to expose market inefficiencies.

The video continues with Laurent Bernut discussing the corporate space-time continuum, which poses a paradox for short sellers. He brings attention to situations where companies reward employees for participating in fraud, while senior management denies knowledge of such practices. Bernut advises short sellers to adopt a non-adversarial approach toward company management, even when they are right, as there are alternative ways to short a stock. He emphasizes the risk management aspect of short selling and cautions that it should be done cautiously.

In his Algo Trading Week masterclass, Bernut emphasizes the importance of learning how to sell short and the risks associated with not having this skill, especially in anticipation of a bear market. He also touches upon how short selling can contribute to increased market volatility and the potential for share price collapses.

The video continues with Laurent Bernut thanking the viewers for their participation and engagement throughout the masterclass on short selling. He expresses his appreciation for the questions and comments received during the session, highlighting the importance of active participation and curiosity in the learning process.

Laurent Bernut then introduces an upcoming course on short selling with Python, aimed at providing practical skills for implementing short selling strategies using programming. The course will cover various topics, including data analysis, algorithmic trading, risk management, and backtesting. He emphasizes the value of combining quantitative analysis with short selling techniques, and how Python can be a powerful tool for this purpose.

In addition to the course, Laurent Bernut announces the release of a complementary book titled "Short Selling Unveiled: A Comprehensive Guide to Profiting in Bear Markets." The book will delve into both the how and why of short selling, providing insights, strategies, and real-world examples. It aims to demystify the discipline and equip readers with the knowledge and skills necessary to navigate the complexities of short selling successfully. The book is scheduled to be published on October 11th, 2021, and will be available on Amazon.com.

As the video concludes, Laurent Bernut reiterates the importance of continuous learning and improvement in the field of short selling. He encourages viewers to explore the course and book to deepen their understanding and enhance their skills. He expresses his commitment to helping individuals become proficient in short selling and emphasizes the value of staying informed and adaptable in the ever-changing financial markets.

With a final note of gratitude and encouragement, Laurent Bernut bids farewell to the viewers, leaving them with the invitation to connect, ask questions, and continue their journey in the world of short selling. The video ends, and viewers are left inspired and motivated to further explore the opportunities and challenges presented by short selling.

• 00:00:00 The video introduces Laurent Bernut as the founder and CEO of Alpha Secure Capital and a dedicated short seller at Fidelity Investments. The masterclass will focus on short selling and will last for two hours, with no Q&A session at the end. The speaker encourages viewers to ask relevant questions during the session itself. He also mentions a course on short selling with Python and a complementary book that explains both how and why to do it. The book will be published on October 11th, 2021, and will be available on Amazon.com.
  
  
• 00:05:00 Laurent Bernut explains the key takeaways of his masterclass on short selling, stating that top picking is bankrupt and short selling is the most valuable skillset for raising a successful fund. Bernut also debunks ten classic myths about short selling, while highlighting the under-researched nature of the discipline. He explains the dynamics of short selling and why even successful market participants struggle with the short side. Bernut also shares a personal insight on money management and its central part in the course. The section concludes with an overview of how short selling works and the importance of locating the borrow.
  
  
• 00:10:00 Laurent Bernut discusses the bankrupt nature of stock picking and why traders must shift their focus to other practices such as short selling. He explains that the industry is built on the cult of the stock picker, but referring to the numbers, we know that two-thirds of the active managers underperform their benchmark year after year, which has led them to change their focus from stock picking to closet indexing, and passive investing. Active managers do not underperform by large margins; it's usually plus or minus one or two percent of the index. However, during bear markets, verticals such as short selling become relevant.
  
  
• 00:15:00 Laurent Bernut dispels the myth that short sellers destroy pensions and companies. He emphasizes that investors park their money in long-short vehicles because they want low volatility, low correlation returns, and downside protection, something that active managers cannot guarantee consistently. Therefore, investors do not care about the long picks from mutual fund managers as they can perform the same function passively using exchange-traded funds. Bernut explains that shorting stocks protects against downside risk, making the short-selling skill in high demand, especially in a bear market.
  
  
• 00:20:00 Laurent Bernut discusses short selling in the context of capitalism and management responsibility. Short sellers, who do not participate in the management of companies, often get blamed for the failure of a company, when in reality, it is poor management that causes the company to fail. The history of capitalism is filled with companies that have become obsolete due to poor management, with short sellers often simply escorting them into that obsolescence. Market value is also discussed as being separate from intrinsic value, with market value being a Canadian beauty contest where the most beautiful person is picked based on who the judge thinks is the most beautiful. Lastly, the misconception that short sellers are evil speculators is discussed, with Bernut cautioning that regulators do not like short sellers who engage in market manipulation, but the job of short sellers is often to unveil paradoxes in the market.
  
  
• 00:25:00 Laurent Bernut, the speaker in the video "Short selling in the bull market - A Masterclass" talks about the corporate space-time continuum, which is a paradox faced by short sellers. The paradox arises when companies award bonuses to employees who participate in fraud, and when senior management denies awareness of such practices. Bernut also suggests that short sellers shouldn't take an adversarial view of company management, even if they're right, as there are non-adversarial ways to short a stock. Knowing that short selling is a risk management exercise, Bernut advises on the potential risks it presents, saying that short selling should be done cautiously.
  
  
• 00:30:00 Laurent Bernut, in his Algo Trading Week masterclass, discusses the importance of learning how to sell short and the risks of not knowing how to do so, emphasizing the value of training and practicing for the inevitable bear market. He also touches upon how short selling can increase market volatility and ultimately lead to the collapse of share prices. Bernut briefly delves into the concept of borrower utilization and the demand and supply of shares taken up for short selling, answering a question on minimizing losses and addressing the need to adjust spot short selling strategies in futures markets.
  
  
• 00:35:00 Laurent Bernut debunks the myth that short selling is not necessary during a bull market. Many hedge funds procrastinated learning short selling before the 2008 financial crisis, leading them to suffer enormously when the markets collapsed. Bernut believes that short selling should be learned during a bull market because it is a competitive field with a lot of alpha leakage, and learning during a bull market gives one room to make mistakes without much consequence. The myth of structural shock is also debunked. While businesses may go bankrupt, the logic doesn't apply to long positions, and those in long positions often visit and train to guard their investments.
  
  
• 00:40:00 Laurent Bernut explains the fallacy of being a "tourist" and going after structural short selling. He believes this represents a lack of humility and says that finding accounting frauds is extremely difficult. Another issue is flawed business models, which are challenging to detect because companies try to hide them. Bernut also talks about valuations, saying that when they don't make sense, they stop making sense. He explains two modes of trading, classic trend following and mean reversion, and says that people on the long side should understand the difference, as they have antagonistic profiles and payoffs.
  
  
• 00:45:00 Laurent Bernut discusses short selling strategies, emphasizing that pairs trading is not a short selling strategy despite many people associating it with the short side. He also talks about the regime definition, which triages the market into three buckets based on price: bullish, bearish, and inconclusive, and the do's: regime definition, relative series, and value traps. Bernut recommends concentrating on the underperforming, "boring" stocks with no real story or growth and high dividend yields because they tend to follow the conversation and underperform, which is what short sellers want.
  
  
• 00:50:00 Laurent Bernut discusses the stumbling block of idea generation when it comes to short selling and compares the growth of companies to the concept of aging, where their yields are higher because their growth happened years ago and there is nothing left. He also presents a graph of the S&P 500 and explains how the solid red and green lines represent the number of stocks making fresh one-year highs and lows, respectively, while the dotted line represents the same data on relative series divided by the index. Half of the constituents of any index roughly outperform, and half underperform, meaning the problematic aspect is timing the top and bottom. Bernut suggests doing sector rotations instead, and presents a heat map of all the region definition methods, leading into the discussion of regime definition.
  
  
• 00:55:00 Laurent Bernut discusses the importance of regime definition in trading, whether it be quantitative or fundamental analysis. Regime definition serves as an overview of the market, telling traders whether it is bullish or bearish. By analyzing the regime, traders may then investigate into why certain stocks are outperforming or underperforming. The answers to these inquiries come in three buckets: sector rotation, temporary mispricing, and specific stock performance. Bernut also introduces three classic trading strategies: trend following, mean reversion, and i-breed.
  
  
• 01:00:00 Laurent Bernut discusses the drawbacks of classic trend following and mean reversion strategies and emphasizes the importance of scaling out and taking short-term profits. This enables traders to capture profit quickly and raise the win rate, while lottery-like trades can run for the long term. He also explains the dynamics of short selling and the need to understand the drift of net exposure, which is fundamental to short selling. Finally, he states that money management is essential to making profits in this game.
  
  
• 01:05:00 The speaker responds to a question about their preferred time frame for short selling. They explain that their style of short selling is trend following, as per the approach taken by their friend Mike Covell who is famous for his trend following podcast. They also highlight the importance of borrowing utilization when short selling, as it indicates the involvement of institutional investors. Once borrowing utilization goes above 50, the speaker asserts that it is a sign that institutional investors have left the building, and the only ones left bringing the stock down are the stable shareholders who are unlikely to sell. Therefore, it is crucial to follow the cycle of sector rotation and get on early in relative terms because it can go on for a long time before the sector rotation ends.
  
  
• 01:10:00 Laurent Bernut discusses the concept of exposure in the stock market, specifically gross exposure, net exposure, and net beta. He explains that in a bull market, investors will typically be net long and have a positive net beta, while they will short defensive stocks that have lower volatility and be long high beta stocks. However, Bernut notes that having a negative net beta is very hard to achieve and is only done by a few investors in the world.
  
  
• 01:15:00 Laurent Bernut discusses how to be properly positioned in a bear market without having a negative net exposure. In a bear market, defensive long positions should be taken in low beta areas such as utilities, consumer staples, and food. These holdings can be supersized due to their low volatility. On the other hand, high beta stocks that rose exponentially must be taken short because they will fall the hardest. This will create a heavy long side and a lighter, more volatile short side, leading to a net exposure that is residually positive. Though stock picking failed in the past, short selling is about risk management and position sizing, making it a strategy to survive the bear market.
  
  
• 01:20:00 Laurent Bernut discusses the importance of extending winners in order to let profitable trades compound over time. He advises taking high probability profits early on and then letting the remainder of the trade run, allowing for the potential of big gains from trend following. However, when short selling, it's important to take money off the table quickly due to the risk of short squeezes. Bernut also emphasizes that stop-loss orders should not be part of normal trading decisions and should only be placed at a point where the investment decision is either invalidated or reversed. Placing stop-loss orders within the volatility band can lead to noise and negatively affect gain expectancy.
  
  
• 01:25:00 Laurent Bernut discusses stop-losses and how they are the number one variable in gain expectancy, as they influence three out of four of the variables - average loss, win rate, and loss rate. He advises making stop-losses a logical and budget issue and experimenting with them to find the best solution. In response to a question about specializing in shorting commodities, Bernut admits to his lack of knowledge but suggests that using his method of taking the relative stage and looking at regimes makes trading stocks easier than it seems. He concludes by emphasizing that money is made in the money management module, and stock picking is not the primary determinant of returns.
  
  
• 01:30:00 Laurent Bernut talks about the importance of money management returns and how it's not just stock picking that makes the returns. He gives an example of the Fidelity company where different portfolios had the same stocks but had varying returns due to the size of their investments. Bernut then goes on to explain different money management algorithms using a graph as an example, including equal weight, equal risk, convex, and concave. He compares the different algorithms to driving a car at different gears and emphasizes the importance of modulating risk according to market conditions.
  
  
• 01:35:00 Laurent Bernut emphasizes the importance of money management in making gains in the finance industry. He suggests that the biggest advances will be made in position sizing, risk management, and portfolio management. He believes that the modulating of concave and convex bottom lines can be used to modulate risk and create a different risk profile, which is an efficient way to manage the portfolio's size being put into a trade. According to Bernut, this is where machine learning and AI will contribute over the next generation in making big advancements in finance.
  
  
• 01:40:00 Laurent Bernut addresses questions from viewers about managing risk, the Black Litterman optimization in a long-short portfolio, and the relationship between long-short portfolios and leverage. Bernut explains that managing risk involves knowing when to re-accelerate or decelerate slowly, while leveraging a long-short portfolio can magnify returns, but can also be a double-edged sword if not used carefully. He also notes that while he hasn't personally used the Black Litterman optimization, it is an iteration of capping. Finally, Bernut shares his personal approach of using open risk and relative risk adjusted return to manage risk in his own trading.
  
  
• 01:45:00 Laurent Bernut discusses his preference for managing stocks independently rather than using a composite approach. When it comes to time frames, he trades daily since the market offers numerous opportunities at this level. He also believes that it's easier to have a positive expectancy at longer time frames, but this might result in lower returns and slower turnover. Finally, Bernut addresses the challenge of trading against computers, reminding us that we're competing with them rather than being punished by them.
  
  
• 01:50:00 Laurent Bernut discusses the use of AI and machine learning in retail trading. He believes that using AI and machine learning for market predictions is bound to fail because randomness is here to stay. As for retail traders building institutions, it is easier than one would think. One can hop on the tail of big institutions by observing where volumes are moving and stocks are trending. While retail traders cannot beat big institutions in high-frequency trading, they can always trade trends and niches that do not require extensive research. All in all, Bernut advises against trying to compete with big institutions in pick-by-pick trading and encourages finding areas that allow for easier entry and success.
  
  
• 01:55:00 Laurent Bernut discusses how short sellers have an advantage in processing information faster than analysts and how it's important to find your niche and concentrate on negative expectancy. When asked about keeping a shorting mindset, Bernut advises staying humble and accepting when you're wrong, moving on and setting stop losses. He even compares stock hunting to fishing, explaining how to be ruthless in cutting stocks and casting your net wide. Finally, he answers whether low and medium frequency traders have a chance against high-frequency traders, stating that it’s not a matter of HFT being more profitable, but more so about each trader finding their own niche and being successful in their own right.
  
  
• 02:00:00 The speaker discusses the role of HFTs in trading and how they essentially act as a tax that takes a cut no matter what. He also mentions that winning in HFT is an arms race, and it's either you're ahead of the pack or you pay for somebody else's lunch. Rather than looking at trends in terms of time, he believes that time is the wrong container and it's better to look at the regime instead. In terms of long versus short exposure, he talks about the standard long extension 130/30 or 140/40 models used by quants or gunslingers in mutual funds. These are popular because they are still classified as long only in the asset allocation game, but he suggests considering a net beta benchmark other than just gross exposure levels.
  
  
• 02:05:00 The speaker explains how the 130/30 strategy works in terms of asset allocation and how the residual is still classified as long. The 130/30 strategy means having 130% long and 30% short positions, giving a net exposure of 100% and a cash balance close to zero. The funds using this strategy can still be classified as long-only and benchmarked to their respective indexes. The speaker provides an interesting academic point regarding whether a net beta mandate would be possible but shares that it is a technical question that he has not thought of before. The session ends here with a suggestion to give feedback and ask questions, and a reminder to attend the next session on using sentiment and alternative data in trading.

[MetaQuotes]

How to choose the best stocks and live trade by Dr. Hui Liu | Algo Trading Week Day 2

How to choose the best stocks and live trade by Dr. Hui Liu | Algo Trading Week Day 2

During the introduction to Algo Trading Week Day 2, the speaker acknowledges the previous sessions featuring experts in quant and algo trading. They briefly mention the valuable insights shared by these experts, setting the stage for the day's presentation. The focus of Day 2 is on selecting the best stocks and engaging in live trading, with Dr. Hui Liu taking the lead as the presenter.

The speaker also highlights the ongoing Algo Trading Competition, which encompasses three distinct tests covering the foundations of quantitative and algorithmic trading. The winners of the competition will be announced in September, adding an element of anticipation and excitement to the event. Additionally, the speaker reveals that the following day's session will be a two-hour masterclass on short selling, led by Aloha Bendu. The timing of this class will be adjusted to accommodate participants from different time zones.

Dr. Hui Liu begins his presentation by discussing the process of generating a trading idea, validating it, and constructing a machine learning model to test its historical performance. He suggests that traders can derive ideas by reading financial reports or monitoring social media platforms to gauge a company's performance. Dr. Liu also introduces the SPY ETF, which tracks the S&P 500 index and serves as a valuable historical data source. He emphasizes the importance of employing statistical models and conducting backtesting to validate trading ideas before proceeding to create a trading robot using iBridgePi.

The basics of trend trading and the significance of buying low and selling high are then explained by Dr. Liu. He elaborates on the collection of historical data and the utilization of Python on Jupyter Notebook to develop a machine learning model. Dr. Liu demonstrates how the model can be employed to create a stock screener, aiding in the identification of the most promising stocks for trading purposes. He underscores the significance of verifying trading ideas through backtesting and live trading.

In his next segment, Dr. Liu provides a hands-on demonstration of utilizing Python to retrieve historical data from the Yahoo Finance API and manipulate it for building a machine learning model. Specifically, he retrieves daily bar data for the SPY and employs the "request historical data" function. Dr. Liu adds additional columns to the data that calculate the percentage change in the close price from the previous day to the current day, as well as from the current day to the following day. He explains that a negative close price change from yesterday to today, combined with a positive change from today to tomorrow, signifies an opportunity to buy stocks when the price decreases, as his prediction suggests an impending price increase.

The process of constructing a machine learning model to predict stock prices is then detailed by Dr. Liu. He acquires data on the close price, yesterday's price change, and the price change from today to tomorrow. By utilizing a linear regression model, he fits the data and analyzes the results. Dr. Liu displays a plot where the black line represents the predictions of the machine learning model, while scattered data points depict daily stock prices from Yahoo Finance for the S&P 500. He explains that a negative coefficient signifies a negative correlation, indicating that when the price declines, it is likely to rise, and vice versa. Dr. Liu contemplates the viability of using this model for automated trading to potentially generate profits.

Dr. Liu proceeds to discuss the process of selecting the best stocks and engaging in live trading. He recommends traders examine the price at the end of the trading day to determine its upward or downward movement before placing orders near the market close. He demonstrates the construction of a stock screener to gain insights into how the model performs with various stocks and identifies favorable stocks to follow. Dr. Liu acknowledges that his model is relatively simplistic, relying on yesterday's price to predict tomorrow's, and thus considers the incorporation of advanced indicators such as the Moving Average Convergence Divergence (MACD) to enhance prediction accuracy and filter trades.

The utilization of MACD to predict and filter stocks is explored by Dr. Liu, along with a comparison to the buy low sell high model. He presents the results obtained when employing MACD 10 and 30 on the SPY, revealing a relatively weak trend. Consequently, Dr. Liu concludes that using MACD for future predictions may not yield as favorable results as before. He proceeds to discuss the construction of a statistical machine learning model and considers the buy low sell high model as a potential means of generating profits. Dr. Liu highlights Average Pi, a Python platform facilitating backtesting and live trading, underscoring its 100% privacy feature, compatibility with multiple accounts, and flexibility in terms of data providers. He illustrates the simplicity and efficiency of building a buy low sell high model in Average Pi using only a few lines of code.

Dr. Liu explains the process of setting up a configuration for trading using Algo Trading Week Day 2. He emphasizes the execution of the initialize function at the start to define variables and establish the configuration. As an example, he schedules the "buy low, sell high" function to execute every trading day, one minute before the market closes, instructing it to invest 100% of the portfolio into the SPY if yesterday's price was lower than today's. Dr. Liu delves into the topic of backtesting, illustrating how historical data from brokers or third-party providers can be utilized on various time frames, including minute by minute, hourly, or daily.

Next, Dr. Liu demonstrates the process of backtesting a chosen strategy using different data providers and packages. He advises selecting a start time and an end time for the backtesting period, along with confirming the chosen data provider for execution. Transitioning to demo mode, Dr. Liu showcases the process, indicating that data providers like Interactive Brokers (IB) or local historical data can be used for backtesting strategies. He provides guidance on configuring the backtesting setup, utilizing available historical data stored in local files.

Dr. Liu proceeds to demonstrate the use of backtesting for testing the effectiveness of a trading strategy using historical data. He acknowledges the challenge of obtaining meaningful daily bar data for extensive backtesting timeframes. To overcome this obstacle, he introduces the concept of simulated minute bar data, where the close price of the daily bar can be utilized to simulate the data. This simplifies the process for traders struggling to access the precise data required for backtesting purposes.

Dr. Liu presents the results of backtesting a "buy low sell high" model in comparison to a buy-and-hold strategy for the S&P 500 from 2000 to 2020. The model outperforms the buy-and-hold strategy, resulting in a portfolio value of $800,000 compared to $200,000. He acknowledges that despite the small correlation observed through simple linear regression, the model still delivers positive outcomes. Dr. Liu then transitions to the topic of live trading, indicating that it can be as straightforward as modifying two lines of code to select the desired strategy and input the account code for Interactive Brokers before executing the program. He concludes the presentation by inviting attendees to contact him via email for coding assistance or to arrange an in-person meeting in San Jose, California.

During the Q&A session, a question is posed regarding the certainty of a backtested strategy providing identical results in live trades. Dr. Liu explains that while historical data represents the past and the model may exhibit statistical stability, the price itself is volatile, particularly near the market close. Therefore, variations in predicting the future are inevitable. However, over an extended period, the overall model should hold true. He notes that he utilizes the linear regression model due to its simplicity and ease of understanding, but he acknowledges that more sophisticated machine learning models could potentially yield better results. Dr. Liu also addresses the question of transaction costs and slippage, stating that they should be considered when implementing live trading strategies and can have an impact on the overall performance of the strategy.

Another question is raised regarding the use of other technical indicators in conjunction with the buy low sell high model. Dr. Liu responds by highlighting the flexibility of the Average Pi platform, which allows traders to incorporate additional indicators into their strategies. He mentions that the Moving Average Convergence Divergence (MACD) indicator could be a valuable addition to filter trades and enhance prediction accuracy.

A participant asks about the significance of the time interval between the trading signal and the market close. Dr. Liu explains that the time interval chosen depends on individual preferences and trading strategies. It could be a few minutes or even hours before the market close, depending on the desired trade execution time. He advises traders to experiment with different time intervals to find what works best for their specific strategies.

In response to a question about the impact of market volatility on the buy low sell high model, Dr. Liu acknowledges that increased volatility can potentially generate more trading opportunities. However, he warns that higher volatility also carries higher risk, and traders should carefully consider their risk tolerance and adjust their strategies accordingly.

A participant asks about the potential limitations of the buy low sell high model. Dr. Liu acknowledges that the model's simplicity is both a strength and a limitation. While it can generate positive results, it may not capture more complex market dynamics and could potentially miss out on certain trading opportunities. He suggests that traders who want to explore more advanced strategies and models should consider diving deeper into quantitative finance and exploring other machine learning algorithms.

The Q&A session concludes with Dr. Liu expressing his willingness to assist attendees with any further questions or coding assistance, encouraging them to reach out to him via email.

• 00:00:00 The speaker introduces Algo Trading Week Day 2 and briefly mentions the previous sessions with experts on quant and algo trading. The focus of the day's presentation is on how to choose the best stocks and live trade, presented by Dr. Hui Liu. The speaker also briefly talks about the Algo Trading Competition and its three different tests on the pillars of quantitative and algorithmic trading, with the winners being announced at the end of September. Tomorrow's session will be a two-hour masterclass on short selling by Aloha Bendu, which will be conducted earlier than usual depending on the participant's location.
  
  
• 00:05:00 Dr. Hui Liu discusses how to come up with a trading idea, validate it, and then build a machine learning model to test the performance in the past. He suggests that reading financial reports or using social media to get a feeling about a company's performance can be one way to come up with a trading idea. He then talks about the SPY ETF that tracks the S&P 500 index and how it can be used as a historical data source. Dr. Liu also mentions using statistical models and backtesting to validate the trading idea before creating a trading robot using iBridgePi.
  
  
• 00:10:00 Dr. Hui Liu discusses the basics of trend trading and the importance of buying low and selling high. Building on this concept, he explains how to collect historical data and build a machine learning model using Python on Jupyter Notebook. He also demonstrates how to use the model to create a stock screener that can help identify the best stocks for trading. Finally, he emphasizes the importance of verifying your trading ideas through backtesting and live trading.
  
  
• 00:15:00 Dr. Hui Liu demonstrates how to use Python to retrieve historical data from Yahoo Finance API and manipulate the data to build a machine learning model. The data is retrieved for SPY with a daily bar and the function used to retrieve historical data is "request historical data." To build a machine learning model, Dr. Liu adds a few columns that calculate the close price change from yesterday to today, and from today to tomorrow, in percentage. He explains that if the close price change from yesterday to today is negative and the close price change from today to tomorrow is positive, it means there is an opportunity to buy stock when the price goes down, as his prediction is that the price will go up.
  
  
• 00:20:00 Dr. Hui Liu explains his process for building a machine learning model to predict stock prices. He begins by collecting data on the close price, yesterday's price change, and the price change from today to tomorrow. He then uses a linear regression model to fit the data and analyzes the results. The black line on the plot represents the machine learning model's predictions, and the scattered data points represent daily stock prices from Yahoo Finance for the S&P 500. Liu explains that a negative coefficient means a negative correlation, which indicates that when the price drops, it is likely to go up, and when the price increases, it is likely to go down. Ultimately, Liu considers whether this model can be used for automated trading to potentially make a profit.
  
  
• 00:25:00 Dr. Hui Liu discusses how to choose the best stocks and live trade. He suggests that traders look at the price at the end of a trading day to see if the price goes up or down and then place their orders at the end of the trading market. He demonstrates how to build a stock screener to understand how the model works for other stocks and which stock is a good one to follow. Dr. Liu explains that his model is too simple because he uses yesterday's price to predict tomorrow, so he considers using an advanced indicator like Moving Average Convergence Divergence (MACD) to predict and filter trades.
  
  
• 00:30:00 Dr. Hui Liu discusses using MACD to predict and filter stocks, and how it compares to the buy low sell high model. He shows the results when using MACD 10 and 30 on Spy, revealing a relatively weak trend, and concludes that using MACD for future predictions will not be as successful as before. Dr. Liu discusses building a statistical machine learning model and considering the buy low sell high model to potentially make a profit. He then transitions to discuss using Average Pi, a Python platform for backtesting and live trading, emphasizing its 100% privacy feature, which allows for managing multiple accounts and using any data provider for backtesting. Finally, Dr. Liu walks through how to build a buy low sell high model in Average Pi with only a few lines of code, highlighting its simplicity and efficiency.
  
  
• 00:35:00 Dr. Hui Liu explains the process of setting up a configuration for trading using Algo Trading Week Day 2. Dr. Liu runs the initialize function at the start of the execution to define variables and set up the configuration. In one example, Dr. Liu schedules the "buy low, sell high" function to run every trading day, one minute before the market close, and buy 100% of their portfolio into SPY if the price yesterday was lower than the price today. Dr. Liu then goes on to explain how to backtest using historical data from brokers or third-party data providers on minute by minute, hourly, or day by day basis.
  
  
• 00:40:00 Dr. Hui Liu demonstrates how to backtest your chosen strategy using different data providers and packages. He explains that the easiest way to backtest is to choose a start time, pick an end time for backtesting, and confirm the data provider to run it. Dr. Liu switches to demo mode to show the process and mentions that one can use a data provider like IB or local historical data to backtest their strategy. He further guides on how to set up for the backtesting process while using historical data available in the local file.
  
  
• 00:45:00 Dr. Hui Liu demonstrates how to use backtesting to test the effectiveness of a trading strategy using historical data. He explains that the data needs to be meaningful, but traders may struggle to find daily bar data for long backtesting time frames. However, to solve this problem, he introduces the concept of simulated minute bar data, where if the data isn't available, the close price of the daily bar can be used to simulate the data. This can help simplify the process for traders who struggle to find the correct data for backtesting.
  
  
• 00:50:00 Dr. Hui Liu demonstrates the results of backtesting a "buy low sell high" model compared to a buy-and-hold strategy for the S&P 500 from 2000 to 2020. The model outperformed the buy-and-hold strategy, yielding a portfolio value of $800,000 compared to $200,000. He notes that even though the correlation using simple linear regression was a small number, it still produced good results. Dr. Liu then transitions to live trading, which he says is as simple as changing just two lines of code to choose the strategy and input the account code for IB (Interactive Brokers) before running the program. He concludes the presentation by inviting attendees to contact him via email for coding help or to meet up in person if they happen to be in San Jose, California.
  
  
• 00:55:00 Dr. Hui Liu addresses a question about the certainty of a backtested strategy providing identical results on live trades. He explains that historical data is just the past, and while a model may be statistically stable, the price is volatile, especially close to market close. Therefore, there will always be variations to predict the future, but for a long period, the overall model will still be true. He notes that the reason he is using the linear regression model is that it is the easiest one to understand, but other models can be used, such as the random forest model. However, it would be challenging to explain the model briefly, and simpler models should be used to avoid overfitting.
  
  
• 01:00:00 Dr. Liu discusses how frequently to retrain a model and how to manage risk through stop loss or take profit. He suggests that retraining depends on the kind of model and the amount of data you have; having more data will lead to better results and predictability. When it comes to managing risk through stop loss, he advises that while it's hard to put it into the model directly, it can be incorporated in a backtesting framework to set up a stop loss point and compare results with a strategy without stop loss. Lastly, he cautions regular traders against getting involved in high frequency trading as there is no way to beat institutions in that area.
  
  
• 01:05:00 Dr. Hui Liu discusses the lowest level needed to backtest and the amount of data required for reliable results. He says that the lowest level for reliable testing is the highest number of data points you can obtain, and you should use your judgment based on the daily or hourly bars. He suggests that when comparing the number of factors to your model, for each factor, you should have a hundred data points to fit your model, or your model will not be as great. Finally, the hosts express their gratitude to Dr. Liu and announce the next class.

[MetaQuotes]

How to become a successful quant | Dr Ernest Chan | Algo Trading Week Day 1

How to become a successful quant | Dr Ernest Chan | Algo Trading Week Day 1

The Q&A session with Dr. Ernest Chan begins with the speaker introducing an algorithmic trading competition designed to provide beginners with an opportunity to learn the basics of algorithmic trading while allowing experts to refresh their knowledge. The competition offers prizes such as scholarships and certificates of achievement for the top three winners. Dr. Chan, the founder and CEO of PredictNow.ai and QTS Capital Management, as well as the author of three books on quantitative trading, shares his expertise with the audience.

Dr. Chan starts by highlighting the dominance of quantitative trading over the past decade, with estimates suggesting that up to 90% of trading volume on U.S. exchanges is attributed to algorithmic trading. While he doesn't claim that quantitative trading is superior to discretionary trading, he emphasizes the importance of not overlooking the opportunity to automate or systematize trading strategies. In terms of individual traders competing against institutions, Dr. Chan suggests that niche strategies with limited capacity offer the best opportunity. These strategies are often unattractive to large institutions and involve infrequent trading, making them viable options for independent traders.

The discussion continues with Dr. Chan addressing the importance of finding a niche in algorithmic trading where big institutions are not competing. He advises against direct competition with large players and recommends seeking out areas where there is little to no competition. Dr. Chan responds to questions about the significance of having a Ph.D. in quantitative and algorithmic trading. He emphasizes that having "skin in the game," meaning putting one's own money on the line, is crucial to becoming a successful quant. He suggests that traders focus on developing an intuitive understanding of the market through backtesting trading strategies themselves and reading blogs and books on trading, rather than solely relying on theoretical knowledge.

Dr. Chan advises that a successful quantitative trader should prioritize practical experience and market understanding over a Ph.D. He notes that it takes time to become a successful quant and suggests distinguishing oneself when seeking to join a top quant fund by writing original research in the form of a white paper, focusing on a trading strategy or specific market phenomenon. He cautions that a short track record, such as a single successful trade, is not sufficient to prove consistency and knowledge. In response to a question about incorporating order flow data into trading strategies, Dr. Chan acknowledges its value as an indicator but emphasizes that it should be used in conjunction with other indicators, as no single indicator is comprehensive on its own.

The limitations of using individual indicators to build a trading strategy are discussed by Dr. Chan. He points out that many people use these indicators, reducing their effectiveness. He suggests incorporating them as one of many features in a machine learning program. When asked about ageism in the quant industry, Dr. Chan highlights that if someone operates as a sole proprietor, ageism is not a problem. He also shares his view on the use of machine learning in generating alpha, cautioning about the risk of overfitting and recommending it as a tool for risk management instead. Regarding low-latency trading, Dr. Chan argues that quantitative trading is a necessity in this domain. Finally, he advises that beyond a successful track record, management skills are essential for anyone looking to start a quant-based hedge fund.

Dr. Chan emphasizes that successful fund management involves not only trading skills but also management and business development skills. Having leadership qualities and a background in business management is crucial. When asked about understanding the Indian market quantitatively, he admits to lacking knowledge primarily due to regulations. On the question of how much time one should spend on paper trading before going live with a strategy, Dr. Chan explains that it depends on the efficiency of trading. For high-frequency trading strategies that execute trades every second, two weeks of paper trading may be sufficient to go live. Conversely, for holding strategies, paper trading for three months may be necessary to earn statistical significance based on the number of trades conducted.

Dr. Chan further discusses whether the time series approach should still be the core of one's alpha portfolio, despite recent studies showing that profitable alphas are mostly non-price based. He suggests attending industry conferences, networking with professionals through platforms like LinkedIn, and building a strong track record in trading to attract the attention of experienced quants. He encourages individuals to seek out mentors and take proactive steps in reaching out to potential collaborators.

Moving on, Dr. Chan shares insights on how to hire and train a successful quantitative trading team. He advises that individuals hired should possess demonstrated expertise in the specific function the team is focused on, whether it be risk management, derivatives pricing, or data science. If the team's goal is to develop profitable trading strategies, it is best to hire someone who already has a track record in that area. Additionally, Dr. Chan highlights that there is no universally ideal market for trading, and teams should focus on what they know best. He also explains how high-frequency traders have an advantage in predicting short-term market direction compared to medium and low-frequency traders.

The discussion continues with Dr. Chan delving into the challenge of accurately predicting market movements beyond short timeframes and the complexities involved in utilizing high-frequency trading predictions. He shares his personal approach to trading, which involves hiring skilled traders instead of trading himself. Dr. Chan emphasizes the importance of hiring traders with strong track records, regardless of whether they employ discretionary or quantitative strategies. When asked about his cumulative annual growth rate, he clarifies that he cannot disclose this information due to SEC regulations. Lastly, he notes that quant traders typically do not use the same strategy across all asset classes, making it challenging to compare programming languages like Python and MATLAB for algorithmic trading purposes.

Dr. Chan discusses the use of MATLAB and Python in trading, acknowledging that while he personally prefers MATLAB, different traders have their own preferences, and the choice of language is not the most critical factor. He believes that optimizing transaction costs is difficult even for experts in the field, so it should not be a primary priority for traders. Regarding revising or retraining machine learning strategies, he suggests doing so only when the market regime undergoes significant changes. He also recommends expanding opportunities by learning new languages such as Python or MATLAB to enhance trading skills.

Dr. Chan concludes the session by offering career advice for individuals interested in becoming quant traders. He suggests exploring different areas, such as options trading, to gain a better understanding of personal strengths and weaknesses. He mentions that his current focus revolves around making his machine learning-based risk management system more widely available and clarifies that he does not have plans to release a second edition of his machine trading book in the near future. When hiring traders, he looks for a long and consistent track record and recommends using time series techniques and econometric models for trading at short timeframes. Exit strategies should align with the specific trading strategy, with stop or profit target exits implemented accordingly.

As the video concludes, the host expresses gratitude to Dr. Ernest Chan for his valuable insights and time spent answering a variety of questions related to becoming a successful quant. Viewers are encouraged to email any unanswered questions to ensure they are addressed. The host announces additional sessions in the coming week with other esteemed guests in the field of algorithmic trading, expressing appreciation for the audience's support and encouraging them to continue tuning in.

• 00:00:00 This Q&A session with Dr. Ernest Chan, questions from the audience will be addressed alongside preselected questions. Before diving into the Q&A, the speaker introduced the algorithmic trading competition which provides an opportunity for beginners to learn the basics of algorithmic trading while allowing experts to brush up their knowledge. The top three winners of the competition will receive prizes such as scholarships and certificates of achievement. Dr. Chan is the founder and CEO of PredictNow.ai and QTS Capital Management, and author of three books on quantitative trading.
  
  
• 00:05:00 Dr. Ernest Chan explains that quantitative trading has already been a dominant form of trading for the past 10 years, with some estimates suggesting that up to 90% of trading volume on U.S. exchanges is due to algorithmic trading. While he does not claim that quantitative trading is better than discretionary trading, he emphasizes that ignoring the opportunity to automate or systematize one's strategy would not be wise. When it comes to individual traders competing against institutions, Dr. Chan suggests that niche strategies with limited capacity present the best opportunity. These strategies are often unattractive to large institutions and involve infrequent trading, making them viable options for independent traders.
  
  
• 00:10:00 Dr. Ernest Chan discusses the importance of finding a niche in algo trading where big institutions are not competing and avoiding competition at all costs. He advises against competing with big players and recommends finding a niche where there is no competition. He also answers questions about the importance of having a Ph.D. in quantitative and algo trading, where he advises that having skin in the game is crucial to become a successful quant. Without putting your own money on the line, one will never learn to trade and focus on secondary matters such as mathematics or data science. It is essential to develop an intuitive understanding of the market by backtesting trading strategies yourself and reading blogs and books on trading.
  
  
• 00:15:00 Dr. Ernest Chan advises that the focus of a successful quantitative trader should be on the markets themselves rather than on theoretical knowledge gained from a PhD. He suggests that a singular focus on practical experience is required and that it takes time to become a successful quant. To distinguish oneself when looking to join a top quant fund, he recommends writing original research in the form of a white paper, focusing on a trading strategy or particular market phenomenon. He also warns that a short track record, such as a successful trade, is not sufficient by itself to prove consistency and knowledge. In response to a follow-up question about incorporating order flow data, he advises that it is a good indicator, but not sufficient as a standalone indicator, and there are many other indicators that should be utilized in conjunction.
  
  
• 00:20:00 Dr. Ernest Chan discusses the limitations of using individual indicators to build a trading strategy due to the large number of people that use them. He suggests incorporating them as one of many features in a machine learning program. When asked about ageism in the quant industry, Dr. Chan points out that if someone is a sole proprietor, ageism is not a problem. He also shares his view on the use of machine learning in generating alpha, highlighting the risk of overfitting and recommending it as a risk management tool instead. With regards to low latency trading, Dr. Chan argues that quantitative trading is a necessity for this domain. Finally, he advises that beyond a successful track record, management skills are essential for someone who wants to start a quant-based hedge fund.
  
  
• 00:25:00 Dr. Ernest Chan emphasizes that successful fund management involves not only trading skills but also management and business development skills. Thus, having leadership qualities and a background in business management is crucial. When asked about understanding the Indian market quantitatively, he admits to having no knowledge of it primarily due to regulations. On the question of how much time one should spend on paper trading before going live with a strategy, Dr. Chan explains that it depends on the efficiency of trading. For high-frequency trading strategies that trade every second, two weeks of paper trading are enough to go live. Meanwhile, for holding strategies, paper trading for three months to earn statistical significance based on the number of traits may be necessary. Lastly, he discusses whether the time series approach should still be the heart of one's alpha portfolio despite recent studies showing that profitable alphas are mostly non-price based.
  
  
• 00:30:00 Dr. Ernest Chan suggests attending industry conferences, networking with professionals through LinkedIn or other platforms, and building a strong track record in trading. He also advises seeking out mentors and being proactive in reaching out to potential collaborators. Building a solid reputation and showing a willingness to learn and improve can help attract the attention of experienced quants.
  
  
• 00:35:00 Dr. Ernest Chan discusses how to hire and train a successful quantitative trading team. He advises that the person hired needs to have demonstrated expertise in the specific function that the team is focused on, whether it be risk management, derivatives pricing, or data science. If the team's goal is to develop profitable trading strategies, it's best to hire someone who already has a track record. Additionally, Chan states that there is no universally good market for trading, and teams should focus on what they know best. Finally, he discusses how high-frequency traders have an advantage in predicting market direction in the short-term compared to medium and low-frequency traders.
  
  
• 00:40:00 Dr. Ernest Chan discusses the difficulty of accurately predicting market movements past short timeframes, and the challenge of utilizing high-frequency trading predictions. He also touches on his personal approach to trading, which involves hiring skilled traders and not trading himself. Chan emphasizes the importance of hiring traders with strong track records, regardless of whether they utilize discretionary or quantitative strategies. When asked about his cumulative annual growth rate, Chan states that he cannot disclose this information due to SEC regulations. Finally, he notes that quant traders typically do not use the same strategy across all asset classes, and that comparing Python and MATLAB for algo trading is difficult.
  
  
• 00:45:00 Dr. Ernest Chan discusses the use of Matlab and Python in trading. While he personally prefers Matlab, he acknowledges that different traders have their own preferences and that language is not the most important aspect. He also believes that optimizing transaction costs is difficult, even for experts in the field, so it should not be a priority for traders. When it comes to revising or retraining machine learning strategies, he suggests doing so only when the market regime has drastically changed, and upskilling by learning new languages such as Python or Matlab can help traders expand their opportunities.
  
  
• 00:50:00 Dr. Ernest Chan discusses career advice for individuals interested in becoming a quant trader. He suggests trying out different areas, such as options trading, to gain a better understanding of personal strengths and weaknesses. He also mentions that his current focus is on making his machine learning-based risk management system more widely available and that he does not have plans to release a second edition of his machine trading book in the near future. When hiring traders, he looks for a long and consistent track record, and he recommends using time series techniques and econometric models for trading at short timeframes. He notes that exit strategies depend on the specific trading strategy and suggests implementing stop or profit target exits accordingly.
  
  
• 00:55:00 The video is ending with the host thanking Dr. Ernest Chan for his time and insightful answers to a variety of questions related to becoming a successful quant. Viewers are encouraged to email any questions that weren't answered during the session to ensure that they get addressed. The host announces that there will be additional sessions over the course of the next week with other esteemed guests within the field of algo trading. The audience is thanked for their support and encouraged to continue tuning in.

[MetaQuotes]

Before you get into quant and algorithmic trading... [Panel Discussion] | Algo Trading Week Day 0

Before you get into quant and algorithmic trading... [Panel Discussion] | Algo Trading Week Day 0

The Algo Trading Week kicks off with an engaging panel discussion led by the host and featuring industry experts. The host begins by inviting the head of marketing and outreach initiatives to provide some background on the event and its purpose. The head of marketing explains that the primary goal of Algo Trading Week is to make algorithmic trading more accessible and bring it into the mainstream. The event aims to achieve this through various educational initiatives such as webinars, workshops, and free resources. Additionally, Algo Trading Week is a celebration of the company's 11th anniversary and will span over the course of 7-8 days, offering a wide range of sessions and competitions.

The speaker then introduces their Quantra courses, highlighting that a significant portion, around 20-25 percent or more, of the courses are available for free. This is made possible through the support and contributions of the community. The speaker expresses their desire to do more and explains how this led them to organize a week-long learning festival. The festival brings together some of the industry's top experts who will share their knowledge and insights. The speaker expresses gratitude for the positive responses received.

Moving on, the speaker introduces the panel members who will be part of the discussion. The panel includes Ishaan, who leads the Contra content team, Nitish, the co-founder and CEO of QuantInsti, Pradipta, the VP of Blue Shift, and Rajiv, the co-founder and CEO of iRage. These esteemed panelists bring diverse perspectives and expertise to the table.

The discussion then transitions to the topic of necessary skills and educational backgrounds required for a career in quant and algorithmic trading. The panel emphasizes the importance of aligning one's interests and passions before delving into this field. They advise individuals to be prepared to commit a significant amount of time and effort and stress the need for a clear understanding of financial markets, programming methods, and statistics and econometrics. The panel emphasizes that expertise in one or two of these areas is necessary, but a minimum level of qualification criteria must be met in all three. The panel also discusses how short duration courses can help individuals develop the necessary skills to become competitive players in the field.

The panelists then delve into the benefits of taking courses in quantitative and algorithmic trading. They highlight the importance of following a proper trading process and utilizing mathematics and statistics to explore anomalies in the market. The courses teach the skill of Python, which is essential for backtesting and verifying hypotheses. Moreover, participants gain the ability to paper or live trade their strategies on platforms like BlueShift. The panelists also discuss the different sources of alpha in the markets and how retail users can benefit from using research and live trading platforms rather than relying solely on ready-made strategies. They emphasize that assessing the risk of a trading strategy should not only consider the strategy in isolation but also its impact on one's position and overall portfolio.

The importance of testing strategies and having access to alpha is further discussed by the panel. They stress the significance of utilizing platforms like BlueShift for systematic research rather than building one's own platform, which requires a different set of skills and processes. The panelists note that trading can be categorized into different styles, and the impact of market developments varies accordingly. They use the analogy of machine learning chess programs to illustrate how the quant trading industry can benefit from advances in technology and data analysis. They also highlight the substantial volume of information available for mid and high-frequency trading strategies due to increased market volume and data availability.

The panelists shift their focus to the impact of technology on quantitative and algorithmic trading. They emphasize the growing importance of big data and automation and acknowledge that high-frequency traders face increasing competition. The panelists address the concerns of retail investors considering entering the field, cautioning against implementing strategies too quickly.

The panelists emphasize the importance of thoroughly testing and understanding a strategy before investing in it. They highlight the need to avoid the dangers of rushing into implementation without proper evaluation. They stress that it is crucial to comprehend why a particular strategy is expected to be successful before using it.

The panelists emphasize the significance of focusing on inputs such as alpha ideas, testing, and risk management to increase the probability of success in trading. They acknowledge that this process may seem slow and tedious, but it is necessary to stick with it and avoid hasty decision-making. For those looking to transition from discretionary to systematic trading, the panelists recommend acquiring a basic understanding of market trading, elementary math and strategy skills, and programming, particularly Python. They also advise individuals to read about successful traders and learn from their experiences to avoid losses through trial and error.

The potential pitfalls of algorithmic trading and how to avoid them are discussed by the panelists. They stress the importance of identifying biases in strategies and ensuring that they work across various market conditions through thorough backtesting and analysis. The panelists caution against underestimating the modeling of exchange activity, as a lack of understanding can lead to missed opportunities or significant delays in trade execution for high-frequency trading strategies. They recommend adopting a systematic approach to strategy development and extensively testing it with both simple and complex factors. The panelists suggest acquiring the necessary skills through courses, webinars, and practice to become proficient and successful quant traders.

The panelists provide valuable advice to individuals interested in algorithmic trading. They caution against look-ahead bias, over-reliance on backtests, and excessive confidence in high returns without considering the associated risks. The panelists also stress the importance of avoiding over-leveraging and remind traders to consider the risks involved when evaluating returns. They highlight the presence of biases that can skew backtest results and emphasize the need to understand and address these biases appropriately.

The speakers emphasize the significance of using the right tools and methods when backtesting to improve the chances of success in trading. They highlight the opportunities available with the rise of open-source systems and data science libraries that are freely accessible to traders who possess the ability to interpret data correctly. Additionally, they mention the possibility of using cloud infrastructure to rent servers on a flexible basis, which can help reduce costs. The speakers acknowledge the challenges of achieving success in trading and stress the importance of being objective and systematic in one's approach to avoid emotional influences such as fear and greed in trading decisions. They recommend taking courses like those offered by Quantra to enhance skills in quantitative and algorithmic trading.

The speaker then discusses the importance of learning all the building blocks of trading objectively and being aware of the various strategies that exist. They highlight the value of investing in one's education, whether in quantitative and algorithmic trading or any other field. The speaker announces a competition for individuals interested in learning the basics of trading, open to traders, programmers, and anyone looking to enhance their knowledge. The competition will consist of three quizzes covering financial markets, math and statistics, and programming and machine learning. The speaker provides resources for test preparation.

The speaker provides detailed information about the upcoming quiz for Algo Trading Week, specifying the dates and topics to be covered. Participants are encouraged to prepare using the indicated resources or any other means they prefer, as the scores will determine the final leaderboard. The speaker suggests taking all three quizzes to increase the chances of ranking among the top three or top ten participants. Additionally, the speaker discusses the hardware requirements needed for a quant setup, explaining that execution hardware can be as simple as a laptop or a minimum configuration on the cloud. However, more advanced research capabilities may require a better computer with at least 4GB of RAM.

The panel then delves into the hardware requirements for high-frequency trading (HFT) and computationally heavy funds. They emphasize that HFT necessitates frequent hardware upgrades and enhancements to achieve faster exchange connectivity, which is a crucial factor in their alpha generation. Trading strategies that require speed and extensive research and data analysis require server-grade infrastructure. The panel also cautions against treating algorithmic trading as a "fire and forget" mechanism, emphasizing the need to regularly monitor strategy performance and take corrective actions if necessary, even when utilizing a cloud-based trading system.

As the panel discussion comes to a close, the panelists express their gratitude to the audience for tuning in and actively participating in the session. They appreciate the patience demonstrated throughout the hour-long discussion and bid farewell until the next session, which will take place on the following day. The panel concludes with a final round of thanks and well-wishes to everyone attending the event.

• 00:00:00 The host of the Algo Trading Week kicks off the event with a panel discussion featuring industry experts. The host invites the head of marketing and outreach initiatives to give some background on the event and why it was created. The head of marketing explains that the goal is to make algorithmic trading more accessible and to take it mainstream through educational initiatives like webinars, workshops, and free resources. The Algo Trading Week is a celebration of the company's 11th anniversary and will feature a variety of sessions and competitions over the next 7-8 days.
  
  
• 00:05:00 The speaker talks about their Quantra courses, stating that 20-25 percent or more of the courses are free due to the support and contributions of the community. They explain that they wanted to do something more, leading them to organize a week-long learning festival with some of the best people in the industry to share their knowledge, which received positive responses. The speaker then moves on to introduce the panel members, including Ishaan, who heads the Contra content team, Nitish, the co-founder and CEO of QuantInsti, Pradipta, the VP of Blue Shift, and Rajiv, the co-founder and CEO of iRage. The discussion then transitions to the topic of necessary skills and educational backgrounds needed for quant and algorithmic trading.
  
  
• 00:10:00 The panel discusses the importance of aligning one's interest and passion before diving into the world of quant and algorithmic trading. They advise that individuals need to be willing to commit significant time and effort and possess a clear understanding of the financial markets, programming methods, and statistics and econometrics. They stress that all three pillars are equally important and that expertise in one or two is necessary, but a bare minimum level of qualification criteria must be met in all three. The panel also discusses how short duration courses can help users build the necessary skills to become competitive players in the field.
  
  
• 00:15:00 The panelists discuss the benefits of taking a course in quantitative and algorithmic trading. The course emphasizes the importance of following a proper process in trading and using math and statistics to explore anomalies. The skill of python is taught in the course to backtest and verify your hypothesis, while also giving you the ability to paper or live trade your strategy on platforms like BlueShift. The panelists also discuss the different sources of alpha in the markets and how retail users can benefit from using research and live trading platforms instead of relying on ready-made strategies. The risk of a trading strategy is not just standalone, but also in relation to your position and overall portfolio.
  
  
• 00:20:00 The panel discusses the importance of testing strategies and having access to alpha, even for retail traders. They also discuss the benefits of using a platform like Blueshift for systematic research instead of building your own platform, which requires a different set of skills and processes. The panelist notes that trading can be bifurcated into different styles, and the impact of developments in the market varies accordingly. They use an analogy of machine learning chess programs to explain how the quant trading industry can benefit from advances in technology and data analysis. They also mention the increased volume on exchanges and the huge amount of information available for mid and high-frequency trading strategies.
  
  
• 00:25:00 The speakers discuss the impact of technology on the field of quantitative and algorithmic trading. Big data and automation are becoming increasingly important, and high-frequency traders are facing more competition. The speakers address the concerns of retail investors who are considering entering the field, warning them about the dangers of implementing strategies too quickly and emphasizing the importance of thoroughly testing and understanding a strategy before investing. It is crucial to understand why a particular strategy will be successful before using it.
  
  
• 00:30:00 Focusing on inputs such as alpha ideas, testing, and risk management is crucial to ensuring a higher probability of success. While it may seem like a slow and boring process, it is necessary to stick with it and avoid rushing into implementation too quickly. For those looking to transition from discretionary to systematic trading, acquiring a basic understanding of market trading, elementary math and strategy skills, and programming (particularly Python) is important. It is also recommended to read about successful traders and avoid losing money through trial and error.
  
  
• 00:35:00 The panelists discuss the potential pitfalls of algorithmic trading and how one can avoid them. They emphasize the importance of identifying any biases in a strategy and ensuring that it works across various market regimes through backtesting and analysis. Additionally, they caution against underestimating the modeling of exchange activity, where a lack of understanding can lead to missed opportunities or significant delay in trade execution for high-frequency trading strategies. The speakers recommend taking a systematic approach to strategy development and extensively testing it with simple as well as more complex factors. Finally, they suggest acquiring the necessary skills through courses, webinars, and practice to become a proficient and successful quant trader.
  
  
• 00:40:00 The panelists provide advice to those interested in algorithmic trading. They warn against look-ahead bias, over-reliance on backtests, and being too confident in high returns without considering the associated risks. The panelists also caution against over leveraging and advise traders to keep in mind that returns in themselves do not have much meaning without considering the risks involved. Furthermore, they suggest that there are many biases that can skew backtest results, and it is essential to understand and address these biases.
  
  
• 00:45:00 The speakers emphasize the importance of using the right tools and methods when backtesting in order to improve the chances of success in trading. They also highlight the opportunities available with the rise of open source systems and data science libraries that are freely available to traders who understand the language of data and can interpret it correctly. Additionally, they note the possibilities of using cloud infrastructure to rent servers on a per-minute, per-hour, per-second, or per-day basis in order to save costs. The speakers also highlight the difficulty of being successful in trading and emphasize the need to be objective and systematic in one's approach to avoid emotions like fear and greed affecting trading decisions. They recommend taking courses like those offered by Quantra to improve skills in quantitative and algorithmic trading.
  
  
• 00:50:00 The speaker discusses the importance of learning all the building blocks of trading objectively and being aware of the various strategies that exist. They emphasize investing in one's education, whether it be in quantitative or algorithmic trading or anywhere else. The speaker then introduces a competition for individuals interested in learning the basics of trading. The competition is open to traders, programmers, and anyone looking to brush up on their knowledge and will consist of three quizzes covering financial markets, math and stats, and programming and machine learning. The quizzes will take place on specific dates, and winners will be announced at the end of September. The speaker also provides resources for test preparation.
  
  
• 00:55:00 The speaker provides information about the upcoming quiz for Algo Trading Week, specifying the dates and topics to be covered. Participants can prepare for the quiz using the indicated resources or any other means they prefer, but the scores will determine the final leaderboard. The speaker suggests taking all three quizzes to increase the odds of featuring in the top three or ten. The speaker then discusses the hardware requirements needed for a quant setup and explains that the execution hardware can be as simple as a laptop or a minimum configuration on the cloud while more advanced research capabilities require a better computer with at least 4GB RAM.
  
  
• 01:00:00 The panel discusses hardware requirements for high-frequency trading (HFT) and computational heavy funds. They note that HFT requires frequent hardware upgrades and enhancements in order to reach the exchange faster, which is their key alpha. Trading strategies that require speed and massive amounts of research and data need server-grade infrastructure. They also caution against treating algorithmic trading as a "fire and forget" mechanism, noting that it is important to monitor strategy performance regularly and take corrective actions if necessary, even if the trading system is cloud-based.
  
  
• 01:05:00 The panelists thank the audience for tuning in and listening to their discussion. They express their gratitude for everyone's patience during the hour-long session and bid farewell until they meet again on the following day for the next session. The panel signs off with a final round of thanks and well-wishes.

[MetaQuotes]

How To Automate A Trading Strategy | Algo Trading Course

How To Automate A Trading Strategy | Algo Trading Course

Rishabh Mittal is a Quantitative Analyst working in the content team at Quantra. His expertise lies in applying unsupervised learning techniques, particularly K-Means, to generate tradeable signals. He is actively involved in developing innovative algorithms for position sizing in the financial markets, utilizing methodologies such as Constant Proportion Portfolio Insurance (CPPI), among others. Before joining Quantra, Rishabh gained experience in creating systematic trading strategies using TradingView for various clients.

In this webinar titled "How To Automate A Trading Strategy," Rishabh will delve into the process of automating trading strategies and guide participants on how to take their systematic trading strategies live. The webinar will commence by addressing the prerequisites necessary for automating a strategy.

Rishabh will then focus on the event-driven approach essential for automated trading. He will explore topics such as connecting with a broker, fetching real-time data, generating signals based on the acquired data, and ultimately placing an order with the broker.

To conclude the session, Rishabh will provide a step-by-step demonstration of setting up a demo strategy for paper-trading in the markets using Blueshift. Participants will gain practical insights into implementing and testing their strategies in a simulated trading environment.

Join Rishabh Mittal in this informative webinar as he shares his expertise on automating trading strategies and offers valuable guidance on taking your systematic trading approach from theory to practice.

[MetaQuotes]

How To Create A Trading Algorithm From Scratch [Algo Trading Webinar] - 22 July 2021

How To Create A Trading Algorithm From Scratch [Algo Trading Webinar] - 22 July 2021

During the webinar, Ashutosh shared his extensive experience in the field of financial derivatives trading, spanning over a decade. He highlighted his expertise in applying advanced data science and machine learning techniques to analyze financial data. Ashutosh holds a prestigious master's degree and is a certified financial analyst (FF). Currently, he is a valuable member of the Quantum City team, responsible for the development and instruction of the EPAT course, the world's first verified algorithmic trading certification.

The webinar primarily focused on guiding participants through the process of creating a trading algorithm from scratch. Ashutosh emphasized the significance of understanding trading algorithms, their various applications in the market, and the conversion of ideas into strategies and eventually into trading algorithms. Essentially, an algorithm serves as a computer program that assists traders in making profitable decisions by analyzing data and generating buy and sell orders based on predetermined rules. It also facilitates interactions with the external environment to send and receive orders effectively.

Before diving into the practical aspects of trading, Ashutosh highlighted the importance of defining one's trading universe and determining the desired alpha. Alpha represents the driving force behind profits, which can originate from diverse sources such as unique market perspectives, gaining an edge over the competition, or implementing specific strategies tailored to individual goals.

The video content covered the three fundamental phases of trading: research, trading, and post-trading. Ashutosh elucidated these phases and provided examples of different trading strategies, focusing on the process of transforming ideas into concrete trading algorithms. He demonstrated how even simple rules, such as buying a stock when its rate of change (roc) surpasses 2 within the last 63 days, can form the foundation of a trading algorithm.

Throughout the webinar, various traders showcased their approaches to building trading algorithms from scratch. One trader utilized visual coding, leveraging data from the Indian market, and incorporated order limits and commission per share. Another trader demonstrated the step-by-step process, beginning with defining their trading universe, followed by creating an alpha function to calculate the roc, establishing trading rules, and finally implementing the strategy using logic blocks.

The video provided comprehensive insights into the essential components of a trading algorithm, namely the conditions, order sending, and order receiving. Additionally, it showcased how to schedule algorithms for automatic execution. Strategies based on beta and momentum were presented as a means to exploit market trends, alongside the inclusion of a mean-diverting strategy.

Ashutosh explained the process of creating a trading algorithm from scratch, covering key aspects such as defining a universe of stocks, calculating relevant hedges, and executing trading rules. He also emphasized the significance of running backtests on the algorithm and optimizing it for enhanced performance.

Quantitative methods and their role in improving trading skills were discussed, with an emphasis on utilizing beta and correlation with the market to make informed decisions. Ashutosh also offered participants the opportunity for a free counseling call to further support their trading journey.

Furthermore, the webinar explored the different types of data that can be utilized within an algorithm and addressed the process of cost assessment for the EPAT course. Attendees were also provided with a list of course counselors for guidance and support.

Ashutosh's webinar delivered a comprehensive guide to creating trading algorithms from scratch. Attendees were encouraged to submit any unanswered questions they may have had during the presentation, ensuring a thorough understanding of the topic.

• 00:00:00 Ashutosh discusses how he has been involved in the field of financial derivatives trading for more than a decade, and has experience in applying advanced data science and machine learning techniques to financial data. He also has a masters degree from a prestigious university, and is a certified financial analyst (FF). He is currently a part of thequantumcity team, and is responsible for developing and teaching the epact course, which is the world's first verified algorithmic trading certification.
  
  
• 00:05:00 This webinar is focused on how to create a trading algorithm from scratch, and it covers topics such as why we need to learn about trading algorithms, the different ways in which algorithms are used in the market, and how to convert ideas into strategies and strategies into trading algorithms.
  
  
• 00:10:00 An algorithm is a computer program that helps traders make profitable trading decisions. It analyzes data and generates buy and sell orders based on preset rules. It also interacts with the outside world to send and receive orders.
  
  
• 00:15:00 Before starting to trade, it is important to define your trading universe and determine what alpha you hope to achieve. Alpha is the reason behind profits, and it can come from a variety of sources, such as looking at market variables in a special way, having an edge over the market, or following a specific strategy.
  
  
• 00:20:00 This video explains how to create a trading algorithm from scratch, focusing on the three phases of trading: research, trading, and post-trading. The video provides examples of different trading strategies and explains how to convert ideas into trading algorithms.
  
  
• 00:25:00 In this video, the presenter explains how to create a trading algorithm from scratch. Rules can be as simple as buying a stock when its rate of change (roc) is greater than 2 in the last 63 days.
  
  
• 00:30:00 In this video, a trader explains how to create a trading algorithm from scratch using visual coding. The trader uses data from the Indian market, and sets limits on orders and commission per share.
  
  
• 00:35:00 In this video, a trader shows how to create a trading algorithm from scratch. First, they define their universe, which includes all of the stocks they will use in the trading strategy. Next, they create an alpha function to calculate the roc, and then define trading rules. Finally, they show how to use the logic block to execute the trading strategy.
  
  
• 00:40:00 This video explains how to create a trading algorithm from scratch. The video covers the three parts of an algorithm: condition, order sending, and order receiving. The video also shows how to schedule the algorithm to run automatically.
  
  
• 00:45:00 The video explains how to create a trading algorithm from scratch. The strategy is based on beta and momentum, and is designed to exploit market trends. The video also covers a mean diverting strategy.
  
  
• 00:50:00 The presenter explains how to create a trading algorithm from scratch, including defining a universe of stocks, calculating relevant hedges, and executing trading rules. The presenter also explains how to run back tests on the algorithm and how to optimize it.
  
  
• 00:55:00 This video discusses how to create a trading algorithm from scratch, including the importance of beta and correlation with the market. The presenter also discusses how to improve your trading skills with the help of quantitative methods. Finally, the presenter shares information about a free counseling call.
  
  
• 01:00:00 This webinar discusses how to create an algorithm from scratch, and discusses the different types of data that can be used in an algorithm. The webinar also discusses how to cost a course, and provides a list of course counselors.
  
  
• 01:05:00 In this video, Ashitosh Sharma from Blueshift shares how to create a trading algorithm from scratch, using Blueshift's visual editor and strategies. Blueshift allows users to backtest their strategies and even paper trade them.
  
  
• 01:10:00 The presenter discusses how to create a trading algorithm from scratch. Afterwards, attendees can submit questions that were not answered in the presentation.

[MetaQuotes]

Machine Learning and Sentiment Analysis [Algo Trading Project Webinar]

Machine Learning and Sentiment Analysis [Algo Trading Project Webinar]

Ladies and gentlemen,

I hope all of you can hear me clearly.

Welcome to Quantum City's YouTube channel. For those of you who regularly attend our webinars, you may remember one of our recent Algo Trading Project webinars, which focused on machine learning in sentiment analysis and portfolio allocation. We had the pleasure of inviting two esteemed EPAT alumni, Carlos Peral and Vivian Thomas, to present their project work. Unfortunately, the post-presentation was interrupted by a hardware failure, and we couldn't cover it in much detail at the time. However, we were fortunate that Carlos took some extra hours to record his presentation separately and share it with us.

So, without further delay, let's proceed and watch Carlos' presentation. Thank you.

"Hello, everybody. For today's presentation, I'm going to show my final project for the EPAT (Executive Programme in Algorithmic Trading) program, which was completed last March. First, let me introduce myself. My name is Carlos Martin, and I have a bachelor's degree in computer engineering. I have been working for over 10 years for several clients, mainly located in Spain and Belgium. My main skill lies in software development, and I have been working for European institutions for the past five years.

The motivation behind this project stems from my interest in machine learning, particularly in sentiment analysis. I believe that these techniques have seen impressive advancements in recent years, with machine learning models being applied in various domains such as text analysis, speech recognition, language translation, and sentiment analysis, which is the focus of this project. The main objective is to find a correlation between news sentiment and price sensitivity and leverage sentiment scores to generate trading signals.

Unlike traditional approaches that rely on technical or quantitative analysis, this project utilizes qualitative data as a new source of information. The goal is to translate this qualitative data into trading signals. The project is divided into two main parts: text analysis and trading strategy implementation.

The text analysis part involves downloading news, performing pre-processing, and implementing a machine learning model to generate sentiment scores. For this project, I chose a long short-term memory (LSTM) model to generate sentiment scores. The trading part involves implementing the trading strategy, analyzing stock prices, and evaluating the strategy's performance.

Let's delve into the project's structure in detail. The text analysis part consists of the news manager, which handles the initial retrieval and pre-processing of news data. I used a class to connect to an external web service and retrieve the news in JSON format. These news data are then stored in a CSV file. The sentiment analysis part includes the pre-processing of text and the NLP (Natural Language Processing) handler, which generates polarity scores using a library called Analytic Evaluator. This library assigns binary scores to the news, labeling them as either negative (-1) or positive (1). This step is crucial for training the model.

The model takes the pre-processed news and is trained using a sigmoid function for binary classification. The output sentiment scores are classified as either positive or negative. The trading strategy is then implemented, and the generated sentiment scores are translated into trading signals. A value of -1 represents a sell signal, while a value of 1 represents a buy signal.

The project was tested using four stocks: Apple, Amazon, Twitter, and Facebook. The sentiment score strategy was compared to a buy and hold strategy. The performance was evaluated using returns, the Sharpe ratio, and strategy returns. The results varied across stocks, with some stocks showing improved performance using the sentiment score strategy compared to the buy and hold strategy. However, there were cases where the sentiment score strategy did not perform well, especially during certain periods.

In conclusion, this project highlights a correlation between negative trends, bad news, and potential trading opportunities. By incorporating sentiment analysis into the trading strategy, it becomes possible to leverage qualitative data and capture market sentiment in a systematic manner. This approach can provide an additional layer of information that complements traditional technical and quantitative analysis.

However, it is important to note that sentiment analysis is not a foolproof method, and its effectiveness can vary depending on various factors. Market conditions, the quality and reliability of the news sources, and the accuracy of the sentiment analysis model all play a role in determining the success of the strategy.

Furthermore, it is crucial to continuously evaluate and refine the sentiment analysis model to adapt to changing market dynamics and evolving news patterns. Regular monitoring of the strategy's performance and making necessary adjustments is necessary to ensure its effectiveness over time.

Overall, this project demonstrates the potential of sentiment analysis in algorithmic trading. It opens up new avenues for incorporating qualitative data into trading strategies and provides a framework for further research and development in this area.

I would like to extend my gratitude to the EPAT program and the Quantum City team for providing the platform and resources for me to undertake this project. It has been an enriching experience, and I believe that sentiment analysis can offer valuable insights in the field of algorithmic trading.

Thank you for watching, and I hope you found this presentation informative. If you have any questions or would like to discuss further, please feel free to reach out to me. Have a great day!

• 00:00:00 Carlos Peral presents his final project for the QuantInsti program, which focuses on machine learning in sentiment analysis and portfolio allocation. Carlos has a background in computer engineering and has been working for more than 10 years in software development, with his main interest being in machine learning topics. He discusses the impressive growth that has occurred in recent years in machine learning models and the wide range of domains where they are used, including sentiment analysis. The aim of Carlos's project is to find a correlation between sensitive prices and news sentiment scores and to take advantage of this information to create trading signals. He approaches the problem using qualitative data as a new source of information, rather than a quantitative one, and translates this data into trading signals.
  
  
• 00:05:00 The speaker discusses the two main parts of the algorithm for their sentiment analysis and machine learning project. The first part focuses on text analysis, where the sentiment scores are generated through a long short term memory model and the pre-processing of news retrieved from a web service. The second part involves implementing the trading strategy and analyzing stock prices. The five steps of the project include news downloads, pre-processing, creation of the LCT model, news prediction for sentiment scores, and stock price analysis. The sentiment analysis component includes pre-processing text to obtain better performance. Classes related to retrieving news and generating sentiment scores are available in the project.
  
  
• 00:10:00 The speaker explains the different steps involved in their approach to a sentiment analysis algo trading project. First, the news is downloaded and pre-processed before being classified as positive or negative using a library called Analytic Evader. The next step involves training a model to classify the previous news and classified data. The backtesting phase follows to validate the trading indicators. The project was tested between March 2018 and December 2020.
  
  
• 00:15:00 The speaker discusses the importance of setting values for sentiment scores, especially because it has a significant impact on backtesting. They explain that they do not use neutral news and set negative scores for news that are under 0.08 and positive for higher scores. The speaker then goes on to explain how they generated scores using NLTK library and trained and compiled the LCT model using a sigma function for binary classification. They compare the returns of the sentiment score strategy with a buy and hold strategy for four stocks: Apple, Amazon, Twitter, and Facebook. The speaker shows graphs for each stock and explains that sentiment-based trading seems to outperform buy and hold for some periods, including an improvement during the pandemic.
  
  
• 00:20:00 The presenter concludes that there is a certain correlation between negative trends and bad news, and good performance when compared to a buy and call strategy. However, he suggests that it is essential to work with high-quality data sources and improve the refinement in sentiment polarity to classify news accurately and assign positive or negative labels. He believes that more work needs to be done to fully trust a strategy based on sentiment and suggests that a good model could be incorporated into certain strategies. In conclusion, the presentation sheds light on the potential of incorporating sentiment analysis into investment strategies but warns that it requires further research and improvements in refining sentiment polarity.