M15 backtesting

To add comments, please log in or register
FH111
167
FH111  

Hi,

I have a few questions that I could do with some help with answering if anyone has time...

If I am back testing an EA on M15 is 4 months of data sufficient? (Every tick)

The last 4 months have covered numerous conditions in the pair I am looking at and at a highish-frequency like M15 I am coming to the conclusion that 4 months is ok.

Regarding accuracy, at this timeframe, I have compared backtesting to actual realtime and it is exactly the same for the last few days. Is there a reason to think backtesting is inaccurate as I have heard this. Or is it more of a case of not over-optimizing for available data and ensuring you have sufficient and varied data to work with?

Would be good to get your thoughts.

Thanks

FH111

Greg Pawlak
226
Greg Pawlak  
Hi, 4 month is too short period for backtest,try to use 10 last years. Regards Greg
ArashB
73
ArashB  
FH111:

Hi,

I have a few questions that I could do with some help with answering if anyone has time...

If I am back testing an EA on M15 is 4 months of data sufficient? (Every tick)

The last 4 months have covered numerous conditions in the pair I am looking at and at a highish-frequency like M15 I am coming to the conclusion that 4 months is ok.

Regarding accuracy, at this timeframe, I have compared backtesting to actual realtime and it is exactly the same for the last few days. Is there a reason to think backtesting is inaccurate as I have heard this. Or is it more of a case of not over-optimizing for available data and ensuring you have sufficient and varied data to work with?

Would be good to get your thoughts.

Thanks

FH111

Hello man,


your answer is very depend on your strategy and timefame. if you are using the strategy over M1 and during each hour there is a position (totally more than 200 position) , you are good to go over 4 month back test 

but if you have worked on D1, 4 month will not be enough as during that ime you may have just 10-15 position which is not enough at all

Juvenille Emperor Limited
Moderator
9435
Eleni Anna Branou  
FH111:

Hi,

I have a few questions that I could do with some help with answering if anyone has time...

If I am back testing an EA on M15 is 4 months of data sufficient? (Every tick)

The last 4 months have covered numerous conditions in the pair I am looking at and at a highish-frequency like M15 I am coming to the conclusion that 4 months is ok.

Regarding accuracy, at this timeframe, I have compared backtesting to actual realtime and it is exactly the same for the last few days. Is there a reason to think backtesting is inaccurate as I have heard this. Or is it more of a case of not over-optimizing for available data and ensuring you have sufficient and varied data to work with?

Would be good to get your thoughts.

Thanks

FH111

As the previous members have stated, generally the lower the timeframe the shorter period it need to get trustworthy results.

I would suggest some months for M1 scalping strategies, about 1 year for M5 scalping strategies, 1-2 years at least for M15-M30 scalping/trend strategies and 2-3 years for H4-D1 longer term strategies.

Multi year back testing is not really proving something, because the market and currency pairs behaviour is changing rapidly over the last years, so something that works for the past 1-2 years may never worked 5 or more years ago.

There is no strategy that works in all pairs, all timeframes and all years, this is a holy grail dream.

Greg Pawlak
226
Greg Pawlak  
Hello, there is something called "overfitting" to history, that is why we should test strategies on larger history. Regards Greg
To add comments, please log in or register