Is backtesting necessary before Walk Forward Analysis?

New comment
 

Hi everyone,

Is backtest necessary before WFA, or is it just a waste of time? Considering the fact that WFA will test the strategy's robustness and estimate its real time performance better, do you guys do backtests before WFA, or just put a strategy to test with WFA?

Thanks.

 

Use back test to estimate if it is worth forward testing at all

If it does not work in back test why wasting time? It will not become profitable in forward test

 
techmac:
Use back test to estimate if it is worth forward testing at all If it does not work in back test why wasting time? It will not become profitable in forward test

Is there a chance that a strategy might fail a backtest with whole historical data but can generate profit in a WFA? Optimizing the strategy over the whole historical data might miss some profits in my opinion. However optimizing every 3 months or any other predetermined interval might enable the strategy to accommodate to markets better.

 
kojinakata:
Is there a chance that a strategy might fail a backtest with whole historical data but can generate profit in a WFA? Optimizing the strategy over the whole historical data might miss some profits in my opinion. However optimizing every 3 months or any other predetermined interval might enable the strategy to accommodate to markets better.

Chances that it makes loses in back test and makes profit in WFA are negligible. Maybe only with very tight scalping EAs but such tight scalping EAs should never be back tested (they depend on brokers and account types too).

 
techmac:
Chances that it makes loses in back test and makes profit in WFA are negligible. Maybe only with very tight scalping EAs but such tight scalping EAs should never be back tested (they depend on brokers and account types too).

Thanks techmac. I am going to use backtesting just as a filter for bad strategies from now on.

 

Just make sure when you backtest you use tick data and also backtest for at least 3 years. I actually prefer 5 to 8 years also to check max draw down. If an EA can produce low draw down over an 8 year period I would be happy to run that on my live account for testing.

 
fxinvestfund:
Just make sure when you backtest you use tick data and also backtest for at least 3 years. I actually prefer 5 to 8 years also to check max draw down. If an EA can produce low draw down over an 8 year period I would be happy to run that on my live account for testing.

Thank you for your answer. What is the reason behind for using tick data? My systems are based on H1 timeframe, do I still have to use tick data? Can you elaborate on your reasons?

 

1. It is very useful to make a scan with an enough period backtest before running a sequence of walk forwards.

2. Yes, multiple WF will help you find a better set of parameters against a single long-term backtes.


Have you found an automated solution for multiple backtest (walk forward matrix analysis) or you want to join me to create an easy-to-use yet effective one?

please let me know asap.

New comment
Reason: