Best ways to backtest a strategy?

Discussion in 'Educational Resources' started by dandxg, Jul 18, 2006.

  1. I am starting this thread to get some feedback on backtesting a setup I have been using for some time now.

    I have read a couple of the threads on backtesting, but didn't help as of yet.

    I day trade futures currently and use Sierra Charts and Futuresource, mainly futuresource, for charting. I am looking for a way to accurately backtest this strategy. I am hopeing there is a way that is faster than manual going back through charts? Should I consider Trade Station and learn to write in Easy Language? I would also like input on how far to go back?

    Even if I go through charts, since I am using 5 minute candles, how do I determine if I would have been filled at that price. Switching to tick charts seems to take a long time. I am not opposed to hard work at all, I put in 12-14 hours day and I really enjoy trading. If you could point me in the right direction I would appreciate it. Best of trading to everyone.

  2. There are numerous programs that feature backtesting like TradeStation, Wealth Lab and Amibroker. Eyeballing charts is not a good approach, because you will fool yourself.

    I consider backtesting to be time well spent. While you may not come up with anything useful, it will prove to you that many popular approaches don't work. It also gives you a good feel for the effect of stop losses and various profit-taking approaches.
  3. If I can indulge AAA, how does one determine if you would be filled on entry with backtesting? It seems that you would have to have accurate tick data in order to ensure accuracy, correct?
  4. Dan,

    Tick level data is the most accurate way to backtest, but you can simulate order fills with X minute level data. It all depends on what you are doing. Testing indicator based systems can be difficult as you are tempted to do the mid bar override in realtime trading. The easiest systems to validate accuracy are those where you know your entry point ahead of time. That way if the bar contains that price you can be somewhat assured you got filled. Well that's the easy answer. As you progress you can build in safety factors into your testing. For example in a limit order system you may want to see that according to your X minute data the market extended beyond your limit price by the observed amount of the bid ask spread before recording your fill at the original limit price. Just think about "trade-thru" as a safety factor and you will avoid unrealistic situations.
  5. The trade thru is a very good point and something I was contemplating vs. using tick data to ensure a fill, although I don't how I would ensure a fill, I would simply have confirmation of how many traded at that price.