Machine learning in trading: theory, models, practice and algo-trading - page 1829
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
And what will it do? Nothing.
it's all working with BP again, it won't work
It is not clear how you want to present the data. If it is a series. Looking for some repetitions of combinations - patterns in this series? Describe it clearly without emotion.
Well, a row is a row, a humpbacked grave...
We should not go through the row by a sliding window, but to go in a cycle till the fulfillment of some trigger rule, and then go further till the next rule, and so on...
One and the same pattern, but the pattern is not in the price but in the sequence of events (rules), the event can be anything
Start search of all combinations and see what consequences (patterns) work...
It is as if we create a trading system in automatic mode
Well the row should be left next to each other, the hump is the hump...
But do not go through the row by a sliding window, but go in a cycle until some trigger rule is executed, then we go on to the next rule, and so on...
One and the same pattern, but the pattern is not in the price but in the sequence of events (rules), the event can be anything
Start search of all combinations and see what consequences (patterns) work...
You kind of create a trading system in automatic mode
So you kinda go back to indicators and optimizer
I don't know...
You can imagine anything as a time series, there is always time.
But if no BP prediction algorithm can predict the price in any way, isn't the answer obvious?
Guys, I wanted to know what data do you use to look for patterns? I read here about the zigzag and the floating window. What are they?
According to the cause-and-effect model of price change. Cause - Effect on price. What is the cause for price change????
Model: Smile -> OI, Delta, Volume -> Price -> Indicators. This is not a temporal model, but a causal one. Everything else is bogus. Good luck!!!! Because I've already said too much :-)
Well, who's on the knives to punish, sciccuns ?????
i.e. you kind of go back to indicators and optimizer
Why? Explain your logical chain, how did you come to this conclusion?
I propose the concept, which allows you to create TS of any complexity, of any kind in automatic mode. Who has such a concept?
I propose a concept by which you can create TCs of any complexity, of any kind in automatic mode. Who has such a concept?
What are the advantages of this approach?
Minus: The trader must know the essence of the TS, and what are the pluses?
It's time to move on from the "can you use MO to predict price movement or not" argument, you're stomping in place.
I've had a neural network working since early March. If we assume that this is not a show I made up, and everything is true, then here is some important information:
1. YES, there are regularities and you can train a neural network on them.
2. In practice, using my method, you can get with acceptable accuracy the answer only in about 1% of the questions asked. Simply speaking, if you ask the net "where the price will be in 5 min - higher or lower?" at every minute candlestick, the net will give the answer only 1 time out of 100. If we collect all answers later it will appear to be 65-70% accurate. These results were obtained by testing each model, the test includes about 100.000 questions, from this we get 1000 answers + all this confirms the real market for the fifth month.
But you can not wait for 1 answer per 100 questions. There is a way to get an answer at each candlestick. These answers, of course, with very low "confidence" of the network. In practice, if the network is forced to answer on every candlestick, the guessing will be only 51-52%, it is not enough for trading, but it allows you to see how it thinks. It turns out that it thinks quite adequately, at the level of a beginning trader. If her answers are listed on a scale from 0 to 100, we can clearly see where she expects a reversal, and where she expects the trend to continue.
4. 4. The above is a prediction for a definite period, for example 5 minutes or 5 hours. That is why each model sees a certain "harmonic" but not the whole picture. The price movement consists of the sum of such harmonics. If you create models with the step of one minute, for example, from 5 minutes to 60 minutes, you can get 55 responses at each minute candle. These answers show which harmonic is stronger at the moment. Having this amount of data, you can mathematically process them and get a larger picture.
Yes, 1 answer per 100 questions with 65% quality is not enough, but it is a real result, it works! Maybe this prediction quality is close to the limit of real chaos and you can not go further, I hope it is not so.
Stop already on the following knowledge obtained from experience:
1. The quotes are predictable.
2. The regularities are very few, but they are there!
3. The market does not change quickly. This 1% is quite stable for at least a year.
And following advice - don't get hung up on solving a concrete problem like "how to set a stop or a take with a neural network", try to understand the nature of the object in question first. After that, try to get at least some stable result, any!
It's time to move on from the "can you use MO to predict price movement or not" argument, you're stomping in place.
I've had a neural network working since early March. If we assume that this is not a show I made up, and everything is true, then here is some important information:
1. YES, there are regularities and you can train a neural network on them.
2. In practice, using my method, you can get with acceptable accuracy the answer only in about 1% of the questions asked. Simply speaking, if you ask the net "where the price will be in 5 min - higher or lower?" at every minute candlestick, the net will give the answer only 1 time out of 100. If we collect all answers later it will appear to be 65-70% accurate. These results were obtained by testing each model, the test includes about 100.000 questions, from this we get 1000 answers + all this confirms the real market for the fifth month.
But you can not wait for 1 answer per 100 questions. There is a way to get an answer at each candlestick. These answers, of course, with very low "confidence" of the network. In practice, if the network is forced to answer on every candlestick, the guessing will be only 51-52%, it is not enough for trading, but it allows you to see how it thinks. It turns out that it thinks quite adequately, at the level of a beginning trader. If her answers are listed on a scale from 0 to 100, we can clearly see where she expects a reversal, and where she expects the trend to continue.
4. 4. The above is a prediction for a definite period, for example 5 minutes or 5 hours. That is why each model sees a certain "harmonic" but not the whole picture. The price movement consists of the sum of such harmonics. If you create models with the step of one minute, for example, from 5 minutes to 60 minutes, you can get 55 responses at each minute candle. These answers show which harmonic is stronger at the moment. Having this amount of data, you can mathematically process them and get a larger picture.
Yes, 1 answer per 100 questions of 65% quality is not enough, but it is a real result, it works! Maybe this prediction quality is close to the limit of real chaos and you can not go further, I hope it is not so.
Stop already on the following knowledge obtained from experience:
1. The quotes are predictable.
2. The regularities are very few, but they are there!
3. The market does not change quickly. This 1% is quite stable for at least a year.
And following advice - don't get hung up on solving a concrete problem like "how to set a stop or a take with a neural network", try to understand the nature of the object in question first. After that, try to get at least some stable result, any!
What can I say. What amazes only your confidence in what you say, and so we all listen to a fierce rap :-)
https://www.youtube.com/watch?v=4FJ5njoJ6tM
And so for fun :-)
https://www.youtube.com/watch?v=0UPz0L8cTcI