How to set the Monte carlo trades manipulation filtering criteria?
8 replies
eastpeace
7 years ago #242435
Hello, everyone. How to set the monte carlo trades manipulation? What is the reasonable setting criteria?
1, Make the testing individually for each condition, or could be combined?
2, What is the reasonable number of simulations? 100, 200, or 500?
3, Waht is the reasonable filtering value for Net profit(MC trades) , and max DD%?
I guess level 95% is acceptable. But how about right value? >= 50% net profit of main data, 150,or 200% of max DD% of main data?
Thank you in advance for sharing your experience.

0
coensio
7 years ago #242436
I do not know if there is one proper way to do it, but here is an example how I do it:
1. In the first place I would not mix different MC tests together, otherwise you will not be able to learn from your statistics (what is failing where etc…).
2. I use minimal 200 sims using Exact method
3. It depends what you are aiming for, but in any case I would always check using some higher level of confidence e.g. 98% or even 100% for example Ret/DD@100% > 1.
This is a false statement.
0
Joseph
7 years ago #242437
Like so many other settings, this another subjective one depending on what you’re trying to achieve like coensio said and when you are new (like I am) and not sure what u’re trying to achieve at times, I would start with a minimum of what was suggested in instructions : 95% confidence level values greater that 50% of actual.
The important thing then is to try to understand these values. My understanding is that the above setting means 95 % probability that the value generated by the MC simulation is greater than half of the value on the reference(original) test. ANYONE CAN PLS CORRECT ME IF I’M WRONG.
This then means that if you wanted to be more strict with your evaluation you can use higher confidence levels and instead of using 50% of the actual, you can reference 10%.
Hope it all makes sense.
0
hankeys
7 years ago #242439
with Resampling method achieve 50% of original RD is very rare
use the Exact method
https://strategyquant.com/article/what-is-monte-carlo-analysis-and-why-you-should-use-it/
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
0
eastpeace
7 years ago #245047
Thank you, coensio.
100% level is an interesting and very rigorous test. That is also a good reference.
But don’t you compare the MC results with the original values in Cross_checks?
0
eastpeace
7 years ago #245048
@Joseph, @hankeys
Thank you, Joseph and hankeys.
According to your suggestion, and I read the article again.
1, Use each method separately.
2, Randomize trades order with exact method.
3, With Conf. level 95%, Net profit>50% of the Original value, Max DD% <= 200% of the original value.
Can these settings be acceptable?
0
hankeys
7 years ago #245051
i am using only 50% of original with return /drawdown ratio
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
0
coensio
7 years ago #245060
100% level is an interesting and very rigorous test. That is also a good reference. But don’t you compare the MC results with the original values in Cross_checks?
I will answer with an open question 😉
If I have a system with Ret/DD of 25 and in MC test it drops to Ret/DD=11 @100% confidence (more than 50% drop but still a very strong value), then is it pass or fail according to you?
This is a false statement.
0
Tomas Matejka
7 years ago #245585
Good point Coensio. Percentage expression can be misleading in many cases. Better to consider a specific value like for example you do not want Ret/DD to be < 2 in all RT tests performed. If all tests provide higher Ret/DD than the value specified for the RT test we can keep such strategy
0
Viewing 8 replies - 1 through 8 (of 8 total)