Yashar Seyyedin / Profile

Left feedback to customer for job Update Existing EA

Why I Prefer Manual Trading

Having a busy schedule, not having enough time to trade, or psychological problems do not seem to be good excuses for preferring algorithmic trading to manual trading. The most important factor for me to prefer algorithmic trading is knowing that I can achieve better returns compared to the risk and effort I put in. ]n this post I want to explain what the minimum requirements are for using a trading robot in the financial market and in final section I will be discussing some pathologic points.

Warning: For those who have been following me for a while, you must have noticed by now that I only do my research on forex pairs and my statements do not apply to stock indices, gold, and other assets. Also note that we are talking about algorithmic trading here, not a trading algorithm... meaning that a computer decides to make a trade of a certain size, not a human. Even though in both cases there is a specific algorithm, human behavior can analyze more and more complex factors than a conventional trading robot written by an average MQL programmer.

Assessing robot performance:

1. For me to have the courage to put my personal capital at the disposal of a trading robot, I must have seen its behavior in a 1000-trade backtest for a trading system with 3 indicators (and probably the same number of optimization parameters - please don't say I might not optimize because the person who first created the indicators also optimized them) The curve should show a consistent and good behavior. The more parameters and indicators, the more trades in the backtest should be, because there is risk of overfitting with a large number of parameters.

2. Another way is to use a version of k-fold cross validation that ensures that the train and test data are in correct and timed manner. This version is known among algorithmic traders as forward testing, but to get reliable results, we must have at least 200 trades per iteration, and the number of iterations must be at least ten to get a reliable mean and standard deviation of the validation scores.

The pathology of algorithmic traders

Since algorithmic traders are much fewer in number than manual traders, and since they are much less transparent than manual traders (i.e. they don't reveal what they are doing), I have to guess at some points. By studying the algorithms commonly used in algorithmic trading and examining the performance and claims of algorithmic traders, I came to some conclusions that you may find interesting.

First, most people are not capable of getting a passing score on even one of the two tests above. However, financial markets are too complex to consistently produce consistent returns using an algorithm (even if it is optimized over time). This causes people to reduce reliable tests to simpler versions in order to pretend they are in the business. For example, in test number one, they will do 500 trades instead of 1000 trades. They may justify the bot’s failure in test number two by saying that it was due to certain external factors, such as the decisions of a certain politician or central bank, and think to themselves that it is okay. The bot is probably reliable despite its poor performance over a specific period of time. I have seen people doing k-fold validation with k=3 which is pretty optimistic.

The misleading factor of prop companies even encourages algorithmic trading to this approach because these companies benefit from risky and unreliable methods rather than from the consistent and predictable success of traders. Therefore, many of these algorithmic traders enter this business relying on simplified tests and shout about their success on social networks. Of course, there seems to be a hidden (perhaps even an open) contract between the trader and the prop company in these cases.

People who are interested in trading, despite all the motivations such as financial freedom, not having a boss, not having commitments, and getting rid of a full-time job, actually have only one motivation: to get rich. I think this is the story of all of us.

In fact, the process of learning to trade can be summarized as realizing that the return on trading is no higher than the return on most funds in bank and stocks, and that the return on trading is not even higher than long-term investments in physical assets such as gold and real estate. Trading is just a way of diversifying in managing financial resources, and that's it.

Are you one of those people who prefer to back down when you realize that getting rich in this field, like getting rich in other money-making fields, takes time and effort?

I do not really know what market structure is but I know when that low is taken you can find short entries. Clever ones of course.

Tron's black hole address receives burned TRX from transaction fees and protocol mechanisms. This used to be banks'.

Should not be a good idea to trade into new year anyway:

History has shown that the safe sectors of the economy consist of two groups: first, business owners who were owners from the beginning, not true entrepreneurs. Second, salaried workers and government contractors who are generally in their positions through family ties.
If you are a skilled worker and are not in the above two groups, you have very little chance. Don't try too much. Your effort has no effect on the outcome.

Today, both the Philly Fed and inflation data were bad enough for USD to fall, but this did not happen. Reading the news alone without knowing how much has been already priced seems difficult. Especially for those of us who are used to charts and spend less time studying macroeconomics and microeconomics. So how do we know?

In my opinion, a simple way is to monitor the price movements for a few months and analyze the reaction to the news. For example, the EUR/USD pair rose from the FOMC meeting in October into November/December. That was expected, while almost everyone knew that the rate was going to be cut, so further growth after the December meeting(last week) should not have been expected, while surprisingly the dollar fell slightly even afterwards. Now perhaps it is possible to understand why today's bad data (inflation and Philly Fed) practically did not lead to further decline in the dollar and the dollar is likely to grow in the coming days.

Left feedback to customer for job Updates on Existing Job

Hey everyone! Check out this AI generated code I have done for detecting decoupling pairs... https://github.com/yasharseyyedin-a11y/FX_Correlations_Check
The script compares recent 1-day correlations against 10-day rolling correlations (longer-term baseline). When short-term correlations fade or reverse from the 10-day norm it signals mean reversion opportunities, choppy markets ending, or trend shifts. This "fading correlation" edge spots when pairs diverge temporarily, offering entries before they snap back.

I was hoping that implementing neural networks in MetaTrader would at least give me more options for evaluation. After all, our problem is maximizing profit, not maximizing an imaginary parameter... In the current situation, Python is definitely a better option for training.

AI is the best teacher, It is even saying when you should re-optimize!

One advantage of testing with normal strategies on MetaTrader is when you get a good result you have some idea to capture why it is performing good and reject problematic backtests. This will not work with ML-driven solutions... Everything is a blackbox. All you can do is trust it is splitting data correctly during cross validation and hopefully it is not ovefit. Am I right? Does anyone has an idea?

Something really annoying about academic approach to real world problems is "The academician does not even understand the problem..."
Here you see a LSTM approach to predicting price. It is like saying "OK... today price is 100USD so probably tomorrow will be around 100USD"
Absolutely useless for a trader.

A fundamental difference between the problems solved by AI and the problem of financial markets is that those problems were solved by humans in the past and now automation has been transferred to computers. Whereas the problem of trading is an unsolved problem... The fact that someone has a positive account does not in any way mean that they have solved the financial markets.

Five years ago, programming was a safe path and trading was a little dopamine. Today, trading is the only path.

Something I realized by intuition turned out to be correct. The more optimization parameters you have the more overfit risk you are exposed to, unless you have enough train data.
Although I am still on my word... when you cannot prove that the system applies to your chart, anything is overfit.

Left feedback to customer for job Update existing EA

Considering markets have discrete structure it does not make sense to read it solely relying on continuous averages. We have to move to discrete ideas.

Here I am sharing the ups and down with my 20K challenge. Three months and a half passed and I am hardly at half point.

The success story of individuals and making money from prop firms is an exaggerated and unreliable story that has been promoted on social networks for many years. Many traders, through trial and error, endure successive failures in these challenges to access and earn money from funded accounts. This procedure is generally used by traders who do not have enough confidence in their trading systems.
In many cases, they may even make a profit, but this type of success is completely random. That is, a sequence of failures lead to a point where it is no longer possible to continue this game in the hope of winning. A successful trader may fail one out of every ten attempts he makes at challenges, but having one or two failures out of three attempts completely jeopardizes your success. A number of failures can easily bring you to a point of no return where you join the group of traders who are forced to leave the market before success.

In fact, we should accept the harsh truth that props hate profitable traders. Because they neither loose the challenge fee nor they make that much money that can bring a change for them which mostly rely on challenge sales. If you are concerned to go beyond trading personal capital you should think of another solution I guess.