Machine learning in trading: theory, models, practice and algo-trading - page 3054
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
the relationship between attributes is just as likely to go out of range. Exactly the same abstraction.
I specifically took the OOS where the market changed. The training was on the falling market and the OOS on the rising market.
the relationship between attributes is just as likely to go out of range. Exactly the same abstraction.
But this is not the point, but the proposed approach, which makes some sense.
I keep waiting for normal thoughts from the forum on how to improve such a thing, because my head rarely gets fresh ideas until I read a couple more books on statistics and IO.
I specifically took the OOS where the market changed. The study was on a falling one and the OOS on a rising one.
How many models were trained before you got this one?
Collect such successful models in groups and re-train on their signals - there will be more trades.
How many models were trained before getting this one?
Collect such successful models in groups and re-train them based on their signals - more deals will be made.
10-20 for throwing out mistakes and then the final one.
10-20 to throw out the errors and then the final
Check on all available currency pairs, if the result is about the same, then the strategy is reliable.
Check on all available currency pairs, if the result is about the same, then the strategy is reliable.
You need unified signs for such a strategy.
as it was written above different dispersion and ranges usually
Forum on trading, automated trading systems and testing trading strategies
Machine Learning in Trading: Theory, Models, Practice and Algorithm Trading
Maxim Dmitrievsky, 2023.05.02 13:11
And if we put the task of searching for rules a little differently:
1. to find such trains and test, where on the test is the best. You don't have to take the whole series, you can limit these sections by year and they don't have to follow each other. Just 2 random segments of history for the traine and test. If a good test is found after overtesting and learning, then add all other examples to the model and mark them as "do not trade".
2. You can also train many models (let's say 100), get their predictions and compare them with the original labels. Collect all the errors in one place and sort them by repeatability. Mark the most recurring ones as "do not trade". Then train the final model. Or it is possible to collect only good predictions and mark the rest as "do not trade".
10-20 to throw out the errors and then the final
Do I remember correctly that your strategy is a reversal strategy, i.e. closing on the appearance of a new signal? Everyone has different approaches - I'm getting confused.
Is it at the level of intuition formulated approaches or is there some idea of a vision of on/off market patterns?
At the matstat level, I guess. If on average several models are wrong in predicting the same thing on new data (on a validation subsample), then it is unpredictable at all and is moved to "do not trade"
A specific time and corresponding values of signs/signals can be taken as "the same".Do I remember correctly that your strategy is a turnaround strategy, i.e. closing when a new signal appears? Everyone has different approaches - I'm getting confused.
3 buy/sell/don't trade classes
I just slightly modified the approach from the last article to make it less confusing to understand.
you can close on reverse signals or on stops and take-outs