What's the optimal time period for backtesting?

To add comments, please log in or register
thomas2004
337
thomas2004  

By backtesting I will ask myself quite often: How long should I trace back to do the backtesting? If I backtest the 15M timeframe, a three years historical data should be long enough? Surely many people will say the longer the better. I doubt this.

How is your opinion?

Marco vd Heijden
Moderator
8166
Marco vd Heijden  
You can just skip it.
thomas2004
337
thomas2004  
Marco vd Heijden:
You can just skip it.

Hi,

I notice you've posted the same answer everywhere to my questions. Why?

Marco vd Heijden
Moderator
8166
Marco vd Heijden  
Because what you are doing might not make any difference on the long run.
thomas2004
337
thomas2004  
Marco vd Heijden:
Because what you are doing might not make any difference on the long run.

on the long run? 1 month is long? 3 months are long?

Marco vd Heijden
Moderator
8166
Marco vd Heijden  

That depends on how long you continue with this, before trying something else.

I have talked to some guy in the past, who claimed to have backtested every robot in market...

That's a lot of time wasted, right there.

Anthony Garot
1829
Anthony Garot  
thomas2004:

By backtesting I will ask myself quite often: How long should I trace back to do the backtesting? If I backtest the 15M timeframe, a three years historical data should be long enough? Surely many people will say the longer the better. I doubt this.

How is your opinion?

Statistical analysis is not done on time, it is done on trades. The more trades, the better the statistics can suggest if your strategy has an edge.

To add comments, please log in or register