EA's that pass out-of-sample testing methodologies

To add comments, please log in or register
Marius Lombaard
71
Marius Lombaard  

Hi all,

I was wondering how many of the community members have EA's that actually deliver results on out-of-sample data, and the follow up is whether they produce consistent results on live data?

I'm presupposing that such EA's would not implement grid or martingale and other such bad principles, while imlementing good principles like proper stop loss implementations etc.

Are there EA's on the market that claim to meet the above criteria?

Marius Lombaard
71
Marius Lombaard  

I hope the silence does not mean what I think it means...

Sergey Golubev
Moderator
106119
Sergey Golubev  
Marius Lombaard:

Hi all,

I was wondering how many of the community members have EA's that actually deliver results on out-of-sample data, and the follow up is whether they produce consistent results on live data?

I'm presupposing that such EA's would not implement grid or martingale and other such bad principles, while imlementing good principles like proper stop loss implementations etc.

Are there EA's on the market that claim to meet the above criteria?

If you are talking about the Market so any discussion about the Market products is prohibited on the forum ... so - that is why silence ...

Marius Lombaard
71
Marius Lombaard  
Sergey Golubev:

If you are talking about the Market so any discussion about the Market products is prohibited on the forum ... so - that is why silence ...

okay then, people can mention that they have EA's that deliver results on out of sample data without referring to the market. my bad. sorry.

Sergey Golubev
Moderator
106119
Sergey Golubev  
Marius Lombaard:

okay then, people can mention that they have EA's that deliver results on out of sample data without referring to the market. my bad. sorry.

What is 'out of sample data'?
Incorrect backtesting?

Marius Lombaard
71
Marius Lombaard  
Sergey Golubev:

What is 'out of sample data'?
Incorrect backtesting?

sorry insufficient terms used.


what I'm talking about is forward testing. its a methodology where you use the strategy tester to optimize parameters for an EA and then test those parameters on "out of sample data". this is continually done to cater for market changes.


as an example, you'd optimize for 3 months - lets say begin January to end of March - then you'd use those optimized parameters on April and see how it fares, then you optimize for feb-april and test on May etc etc..


You optimize for a period, then back test without optimizing on the subsequent period. It is a methodology that allows you to test a strategy on historic data in a more realistic way. optimization results always look good because, well, it was optimized for the period you tested. forward testing eliminates the bad assumption that optimized parameters will produce similar results on a period of data that it was not optimized for.


I would post a link to external resources that explain better, but there may be marketing on external links so I'd rather PM anyone who wants more info.

To add comments, please log in or register