Compounded Backtesting

Discussion in 'Strategy Development' started by Dhalsim, May 17, 2013.

  1. Dhalsim


    Just wanted to know if you guys use compounding profits in your back-testing results??

    I usually test all my futures strategies with max 1 contract.
    I test my stock portfolio strats with max $5000 for a trade.

    Just wondering if you guys compound your results?? What is the correct method to backtest a strat??
  2. What do you mean? Like when account level reaches > $xx,xxx, start trading greater than 1 contract?
  3. Dhalsim


    Yes, compounding returns.
  4. I only use backtesting to find statistical edges so I only use fixed position sizing.
  5. I wouldn't say there is one way to generate a strategy. If using retail software, I would be more worried about slippage and orders being filled or left unfilled. Additionally, I would pay extra close attn. to the data etc. things like this.

    You can simply write in parameters to your contracts being bought if you want to do it that way. Should only take a minute or two to figure out the correct way to input this.
  6. Dhalsim


    Thanks. Same here.

    However, if you were to hypothetically take the strat 'live' then surely you would have to compound your returns. So for somebody who is only using automated trading what do you guys do.

    When i go live i will be compounding profits, hence should i also be doing this in my backtesting to get more accurate picture of what to expect.
  7. dom993



    I import the trade-list (out of backtesting) into Excel to simulate position-sizing rules & get a compounded view.

    Note that I am trading futures, and my position size is statically defined for each strategy/instrument, so the only thing I have to simulate in Excel is the evolution of that statically defined position size over time.

    If I was trading stocks, where the position-size can truly vary to adjust to a fixed risk-size, then I would do that in the strategy itself (hence get the implication of it straight from backtesting), and only use Excel for long-term evolution of the risk-size parameter.