Backtest or forward test value to use in Re-optimization?

[ezraluandre]

Hi,

I'd like to ask about backtest and forward test optimization. I understand the general idea about backtest and forward test but kind of confuse with which value should be used.
Let's say from backtest optimization I fount that OptimizeA as the winner but at the forward optimization the best winner is OptimizeB. From those result if I want to set my setting, should I set it as OptimizeA or should I set it as OptimizeB?

 

[Enivid]

You need to check the OptimizeB settings via a backtest to make sure that they can produce consistent results (that the forward test period wasn't just a stroke of luck). Then the OptimizeB settings should take priority over OptimizeA even if they weren't as profitable as OptimizeA during the backtets.

 

[ezraluandre]

So to sum it up:
Scenario 1:
- if OptimizeB profitable on forward test but failed in backtest, the forward test is a lucky coincidence

Scenario 2:
- if OptimizeB profitable both on forward test and also in backtest then priorities the forward test setting over the backtest setting

Scenario 3:
- if OptimizeB profitable on backtest but failed on forward test then the settings isn't robust an could resulting in loss

Does this correct?

 

[Enivid]

So to sum it up:
Scenario 1:
- if OptimizeB profitable on forward test but failed in backtest, the forward test is a lucky coincidence

Scenario 2:
- if OptimizeB profitable both on forward test and also in backtest then priorities the forward test setting over the backtest setting

Scenario 3:
- if OptimizeB profitable on backtest but failed on forward test then the settings isn't robust an could resulting in loss

Does this correct?

More or less. In any case, any backtesting period or a forward testing period may just show profit due to a lucky coincidence. But the more testing time (and especially forward testing time) you have, the unlikelier it is.

 

[ezraluandre]

Ok, thanks for your help