A specific backtesting methodology

Forums ProRealTime English forum General trading discussions A specific backtesting methodology

Viewing 2 posts - 1 through 2 (of 2 total)
  • #144744

    Hello all, I would like to discuss opinions and perspectives about a specific backtesting methodology.

    I want to set the goal of building a portfolio of strategies that run until they reach a maximum drawdown percentage (maxDD). Drawdown is defined as fall from maximum equity reached. It is coded with a little snippet that make the strategy quit if maxDD is reached. It can be thought as a portfolio of strategies, each of them running with a maxDD trailing stop on equity.

    The methodological question is: what’s the appropriate backtesting methodology to select candidate strategies?

    Let’s suppose to use a quite tight maxDD as 10%.

    Simple backtesting of a set of strategies imposing the same maxDD on a long period of time leads to highlight strategies that quit early in the backtesting period and have no evidence of actually working.

    A partial solution could be to look for more “resilient” strategies that, independently from performance (total gain) show to stay as long as possible during the entire backtesting period. Still, to my evidence, this does not guarantee to select strategies with a high probability of resilience. (A strategy discarded because of one early stop, could have worked fine after that stop.)

    All in all, “probability of resilience” (N of times a strategy quit with a profit>maxDD/ total N of time the strategy reach maxDD quit) could  be what I’m looking for, but it’s no easy to find out how to extract this from backtesting.

    WFA, as built in PRT, doesn’t help either because at each period it selects one best strategy based on performance and not “resilience”.

    Removing the maxDD quit rule and observing drawdown reached by best strategies? Not sure about what the resilience probability is when using the intended maxDD in real trading.

    Moreover, choice of the length of backtesting period is only one of the issues.

    Thanks in advance for discussing this.

    Stefano

    #144856

    You should look into the MAE (Max Adverse Execution) average. It will tell you how much orders can be in the wrong side as an average. Use it to determine you max DD / stoploss.

Viewing 2 posts - 1 through 2 (of 2 total)

Create your free account now and post your request to benefit from the help of the community
Register or Login