StrategyTester : ExpertMAPSARSizeOptimized.ex5 / EURUSD / 01-05-2012 to 12-05-2012

New comment
 

Hello all,


Here is the result when running strategy tester with ExpertMAPSARSizeOptimized.ex5 on EURUSD market between 01-05 to 12-05 :

Initial deposit : 1000€

Net total profit : 3577.02€

Brut profit : 5215.98€

Brut loss : -1638.96€


Don't you think that is awesome ?

Thanks

 
eurosoa:

Hello all,


Here is the result when running strategy tester with ExpertMAPSARSizeOptimized.ex5 on EURUSD market between 01-05 to 12-05 :

Initial deposit : 1000€

Net total profit : 3577.02€

Brut profit : 5215.98€

Brut loss : -1638.96€


Don't you think that is awesome ?

Thanks


No. Such a short period of testing on historical data doesn't prove anything. If the strategy is expected to be robust, it has to be backtested on a long period, then prove that in the result of optmization process there are many sets of parameters, for which the strategy is profitable, then produce stable, positive results in walk-forward testing and finally - to run for some considerable time on a demo account and still produce profits.

So my guess is that above result is just because the strategy got overoptimized in the backtesting and didn't even have time to present its drawbacks due to extremely short testing period.

Moreover, there are many other parameters that might influence such a deceptive positive result, e.g. "Execution" type in strategy tester or "future leak" syndrome (I don't know if it is present in the expert you have mentioned, but this might as well be one of the reasons for such suspiciously good results) or just blind luck/coincidence. 

 

By the way, it would be much easier to say something more if you showed some of the statistics, e.g. Maximum Drawdown, Profit Factor, Win%, Loss%, Average Win, Average Loss. 

 
Enigma71fx:

No. Such a short period of testing on historical data doesn't prove anything. If the strategy is expected to be robust, it has to be backtested on a long period, then prove that in the result of optmization process there are many sets of parameters, for which the strategy is profitable, then produce stable, positive results in walk-forward testing and finally - to run for some considerable time on a demo account and still produce profits.

So my guess is that above result is just because the strategy got overoptimized in the backtesting and didn't even have time to present its drawbacks due to extremely short testing period.

Moreover, there are many other parameters that might influence such a deceptive positive result, e.g. "Execution" type in strategy tester or "future leak" syndrome (I don't know if it is present in the expert you have mentioned, but this might as well be one of the reasons for such suspiciously good results) or just blind luck/coincidence. 

 

By the way, it would be much easier to say something more if you showed some of the statistics, e.g. Maximum Drawdown, Profit Factor, Win%, Loss%, Average Win, Average Loss. 

Here is the report :

http://eurosoa.somee.com/forex-reports/ReportTester-1065311.html

But it is in french.


Another one more "awesome" (net total profit = $10 337.86 for initial deposit of $1000) :

http://eurosoa.somee.com/forex-reports/ReportTester-1065311-2.html

I just modified the following input parameter : Inp_Trailing_ParabolicSAR_Step=0.16 (initial value = 0.2)

But yes, it is relatively easy to optimize parameters when backtesting.



 

OK, so 3 obvious things that come to my mind after seeing such reports:

1) As mentioned before - testing period way too short. As by the results I could expect roughly one trade daily, test it on at least 2-3 months. (See point 2) below).

2) Only 6-7 trades made - it basically implies that the results are not statistically significant. Statistical significance begins with 30 samples, and the huge increase of stability of results begins with ~100 samples (in this case, 100 trades).

3) Maximum equity drawdown of  respectively 74% and 48% happening only after several days of testing is almost a guarantee of bankruptcy. It is just a matter of longer testing period.

 

 Just a short  digression that came to my mind due to this discussion: back in 2008 I have coded an EA in MQL4, which - during 2 months testing period in MT4 Strategy Tester - has maanged to rise from initial 10,000$ to astonishing 1,236,000$ within a month, and then plunged back to 36,000$ during the second month :) Unfortunately I didn't keep the report, but it is still funny when I think about it. However I recall that after such results, just out of curiosity I tested it on a demo account and its equity curve was a downward slope from 10,000$ to 0$ within 2 weeks or so :)

 
Enigma71fx:

OK, so 3 obvious things that come to my mind after seeing such reports:

1) As mentioned before - testing period way too short. As by the results I could expect roughly one trade daily, test it on at least 2-3 months. (See point 2) below).

2) Only 6-7 trades made - it basically implies that the results are not statistically significant. Statistical significance begins with 30 samples, and the huge increase of stability of results begins with ~100 samples (in this case, 100 trades).

3) Maximum equity drawdown of  respectively 74% and 48% happening only after several days of testing is almost a guarantee of bankruptcy. It is just a matter of longer testing period.

 

 Just a short  digression that came to my mind due to this discussion: back in 2008 I have coded an EA in MQL4, which - during 2 months testing period in MT4 Strategy Tester - has maanged to rise from initial 10,000$ to astonishing 1,236,000$ within a month, and then plunged back to 36,000$ during the second month :) Unfortunately I didn't keep the report, but it is still funny when I think about it. However I recall that after such results, just out of curiosity I tested it on a demo account and its equity curve was a downward slope from 10,000$ to 0$ within 2 weeks or so :)

Thanks for these precisions. I'm learning MQL5 and hope to be able to code my own EA soon :)


 
eurosoa:

Thanks for these precisions. I'm learning MQL5 and hope to be able to code my own EA soon :)


Maybe I should open another topic for this, but let's try here :

Is it possible, in strategy tester, to test an EA on a repetitive period of n days each month ?

Example : I would like to test an EA from January to April but on the first 5 days of each month (01/01 to 05/01, then 01/02 to 05/02, then 01/03 to 05/03 then 01/04 to 05/04) ?

Thanks

 

Hi,

As far as I am aware, it is not possible. You would have to test it separately. BTW, why do you want to test with such gaps? Do you expect to see any seasonality in the results, like e.g. "first 5 days of each month are more profitable than the rest of the month"? 

 
Enigma71fx:

Hi,

As far as I am aware, it is not possible. You would have to test it separately. BTW, why do you want to test with such gaps? Do you expect to see any seasonality in the results, like e.g. "first 5 days of each month are more profitable than the rest of the month"? 

Yes this is exactly what I expect to see :) As I do that manually, It would have helped me to save time if this were possible.


 
eurosoa:

Привет всем,


Вот результат при работе с тестером стратегий ExpertMAPSARSizeOptimized.ex5 на рынке EURUSD между 01-05 до 12-05:

Первоначальный взнос: 1000 €

Чистая Суммарная прибыль: 3577,02 €

Brut прибыль: 5215,98 €

Brut потери: -1638,96 €


Разве вы не думаете, что это удивительным?

Спасибо


На каком ТФ тестирование проводили?
 
Andrian:
На каком ТФ тестирование проводили?
Please post in English on this forum. Anyway this is an old post, not sure you will receive an answer.
New comment