What results to look for in backtesting systems?

Discussion in 'Strategy Development' started by elit, Oct 20, 2006.

  1. elit

    elit

    Hi, I'm backtesting some systems.

    What are the key results i need to look for?
    What are some usual pitfalls?

    Z-ratio? standard deviation (per trade/week/year?)
    Win/loss-ratio? Number of trades? procent per trade? Days held?


    And obviously a nice _positive_ return would be nice! :D
     
  2. You choose. It is like selecting a house to buy. What is best for me might not be best for you. Some people like maximum profit. Some people like to win most of the time. I like to see trading methods and security price histories that show a wide range of profitability regardless of the parameters used. Some people might prefer to minimize volatility. If I was attempting to minimize volatility then I might examine the standard deviation of profits and losses of all simulated trades. You might want to study Sharpe Ratio, Calmar Ratio, Sterling Ratio, and Traynor Ratio. There are probably other measurements of system performance also.
     
  3. If you are trading more than one security then you might want to model the performance of the portfolio. I find sometimes a security might not show a greatly profitable backtest individually but it improves portfolio performance. The risk level is important. I am studying the effect of risk on portfolio performance now.
     
  4. kut2k2

    kut2k2

    Yes, on number of trades; a certain minimum is required.

    Also, how you did versus the buy-and-hold. No matter how profitable a system is, if it doesn't beat the buy-and-hold, it's a waste of time and money.

    Other stats of worth: profit factor, and Sortino ratio.