Machine learning in trading: theory, models, practice and algo-trading - page 2668
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 then use the modelled dollar rate. I guess the indices will have to be taken directly from the sources themselves, unless I find something more convenient.
I have never thought before that it is more realistic to estimate drawdowns of TS through Monte Carlo, not by the maximum historical drawdown.
Some TSs showed drawdowns higher than the historical drawdown and then worked again
It would be useful to make this a standard feature of MT5 tester
https://medium.com/@ankit_quant/when-to-stop-trading-a-strategy-28d104bb20b6
I never thought before that it is most realistic to estimate drawdowns of a TS through Monte Carlo, not by maximum historical drawdown.
Some TSs showed drawdowns higher than the historical drawdown and then worked again
It would be useful to make this a standard feature of MT5 tester
https://medium.com/@ankit_quant/when-to-stop-trading-a-strategy-28d104bb20b6
I remember you and fxsaber convinced me of the uselessness of Monte Carlo)
I remember you and fxsaber convinced me of the uselessness of Monte Carlo)
God forbid I remember. I don't remember )
God forbid I remember.)
It's about that, but I'm not in contention, as you and he are right in no small part. It just came to mind)
It's about that, but I'm not in contention, as you and he are right in no small part. It just came to mind.)
Maybe I'm confusing something... the discussion was about montekarloy optimisation (as a search for TS), and here it is about risk assessment of a ready strategy. To be more precise, not even risks, but how to determine when the TS stopped working.
Yes, the link there is about validation of overfitted TS. It probably does not make sense in this way. Whether it means that there is no sense in determining the allowable drawdown is also a question.
It's about that, but I'm not in contention, as you and he are right in no small part. It just came to mind.)
I never thought before that it is most realistic to estimate drawdowns of a TS through Monte Carlo, not by maximum historical drawdown.
Some TSs showed drawdowns higher than the historical drawdown and then worked again
It would be useful to make this a standard feature of MT5 tester
https://medium.com/@ankit_quant/when-to-stop-trading-a-strategy-28d104bb20b6
The worst case scenario can be obtained not by randomly mixing 10000 times, but by simply gluing together all drawdowns. But this is if the strategy is working, as in the example. We are likely to have a retrained strategy - a forward test will only help us to evaluate it.
Worst case scenario may turn out to be unacceptable drawdown at all, something in the middle is probably more logical.
Monte Carlo forward test.Maybe I am confused about something... we discussed montecarloa optimisation (as a search for a TS), and here we are talking about risk assessment of a ready strategy. To be more precise, not even risks, but how to determine when the TS stopped working.
Yes, the link there is about validation of overfitted TS. It probably does not make sense in this way. Whether it means that there is no sense in determining the allowable drawdown is also a question.
Well, Monte Carlo gives a lot of possibilities and it can be used in different ways.
In your link I think they use random shuffling of trades (shuffle) so that only drawdown changes. In my understanding, this is not a definition of "true" drawdown, but rather whether the actual drawdown is "normal" or not. If the drawdown is too large or too small (falls into the left or right tail of the modelled histogram), it may indicate a dependence between neighbouring trades.