BackTesting vs WalkForward

Discussion in 'Automated Trading' started by ButterMilk, Aug 14, 2010.

  1. I hear allot about backtesting systems, including TV commercial from retail brokers. My experience, however, is that walkforward testing (with varying test periods) is significantly more effective than backtesting, when it come to consistence and profitability.

    All comments and experiences would be appreciated.
  2. rickty


    What software are you using to perform walk-forward testing?

  3. Pippi436


    If you apply rigorous backtesting methods, the results should be the same, no ? For example, if you forward-test your system (either papertrade or live with samll size) for the next month vs. doing a back test in a month for the same time should yield the same outcome? Past perfmormace won't guarantee future profits and all that, in any case.
  4. There's no reason why the period(s) you forward test on should be more any more representative of market conditions AFTER that than the period(s) you backtest on. Even the recent past is only representative until it's not.
  5. Using a modified NT6., with a GO. The "period" being the in sample data in trading days, then walking forward.
  6. Retail brokers do not understand back/forward testing. They are marketing departments and all they know well is how to trick you into giving them your money using any buzzword they can think of.

    You do not understand backtesting if you are comparing it to forward testing.
  7. Please educate me: :cool:
  8. Buy Pardo's book for more detail.
  9. januson


    Very simple, no need to buy a book for such.

    Backtesting is based at curvefitting over a period, for instance curvefitting/ optimizing the last year looking for 30min trades. Then your optimixer will calculate the gain/ loss and you will clap your hands and look very happy sitting there imagining yourself as a millionaire.

    Forward testing, you don't have anything to start with... as the data arrives you will build a strategy, refining it over time and then let the system run forward.

    Backtesting really sucks bigtime, forwardtesting is much more realistic, though you can easily "cheat" performing forwardtesting in the same manner as backtesting optimizing the whole period.

    Backtesting can of course easily be treated as forwardtesting only curvefitting a small period. But people need some distinction between the concepts, hence the naming :)
  10. Craig66


    I think everybody is kind of missing each other in this thread...
    I'm interested in the reasons the OP started this thread given that he seems to have had some success with walk-forward testing (if one views previous posts).
    #10     Aug 16, 2010