What is your advice on how long to use for back testing and forward testing?

[comfysofa]

Hello everyone,

I need some help here, I'm considering using a 1H time frame to trade, and with my EA, i want to backtest and forwardtest appropriately.

I am familiar with the procedures and all, but what do you recommend should be the time period i backtest and forwardtest with?

Should I backtest 12 months and forwardtest 3 months? Or should I backtest 6 months of data and just forwardtest 2 months?

I'd like to find the optimum settings, just that I'm not too sure how long I should backtest and forwardtest (around 25% of backtest range I heard) to find those stats.

Any recommendations from you more experienced people?

Best regards,
Comfysofa

 

[Pharaoh]

Whether it's an EA or a manual strat, backtesting is usually a worthwhile way to find out if it is worth the effort of forward testing.

The amount of time can be related to the time frame of the strategy. Something that trades off of 5 minute charts generally shouldn't need as long of a test as something that trades off of 4 hour charts.

The time for forward testing can also be influenced by how stable the returns are. If the strategy has drawdowns of less than 10% during backtests and forward tests, I'd be comfortable going live sooner than I would with a strategy that showed 50% drawdowns during testing.

 

[FringFX]

I think you should backtest until such time you are confident or (nearly) 100% knowledgeable/experienced with your own strategy.
You will dictate what is the optimum settings.

 

[TadS]

A good rule of thumb is to you use the most data available. However, you want to make sure that you are testing your EA on market conditions that are likely to persist in the future.

My recommendation would be to run a series of tests and see how much your results vary with varying back and forward testing. If the results vary dramatically it might mean your system is unstable and I would be very cautious moving forward. In my experience at least a year of back testing is a good idea and using a sliding window or walk forward analysis for 3 months is a good place to start.

Hope that helps!