Is backtesting necessary before Walk Forward Analysis?

[kojinakata]

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.

 

[Money Trust]

In my personal opinion, I believe that forward testing is more important...particularly paper trading. Back testing is okay if you want to save time.

 

[dbphoenix]

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.

How will you determine what to forwardtest without backtesting?

 

[kojinakata]

How will you determine what to forwardtest without backtesting?

I don't reject any system that I find before doing WFA. I do WFA on them, and based on the results, reject or accept then move on to paper trading.

 

[kojinakata]

In my personal opinion, I believe that forward testing is more important...particularly paper trading. Back testing is okay if you want to save time.

I do forward testing too, but before testing the system live, you should do WFA testing to see how your system fares with changing market conditions. To do this live would take ages, however WFA can estimate how your strategy might do on forward testing and live. How would backtesting save time?

 

[globalarbtrader]

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.

If you backtest first the performance of your WFA will be massively overstated, since you're only testing stuff that you know already works. This could give you overconfidence, leading to overbetting and overtrading.

 

[kojinakata]

If you backtest first the performance of your WFA will be massively overstated, since you're only testing stuff that you know already works. This could give you overconfidence, leading to overbetting and overtrading.

This is true if you backtest and OPTIMIZE. If you just backtest, WFA will optimize the system in its in-sample data and test it on out-of-sample (OOS) data. So my opinion is, WFA will optimize the system to test with OOS data, so why bother with backtesting and optimizing?

 

[dbphoenix]

If you backtest first the performance of your WFA will be massively overstated, since you're only testing stuff that you know already works. This could give you overconfidence, leading to overbetting and overtrading.

Not if you do it right. The backtesting yields only preliminary results. These are verified by forwardtesting to help avoid a variety of biases.

 

[kojinakata]

How will you determine what to forwardtest without backtesting?

Are you referring to parameters within a system or a system as a whole?

 

[volpunter]

What? Have you studied the basics of time series analysis? Pretty much everything you said in this post is factually incorrect.

If you backtest first the performance of your WFA will be massively overstated, since you're only testing stuff that you know already works. This could give you overconfidence, leading to overbetting and overtrading.