difference in results between testing and live?

Discussion in 'Strategy Development' started by nr1-poker.com, Jun 5, 2007.

  1. At this moment i am testing a strategy, and with 1 future, i can make a net result of around 1.000$ a day. Before buying this program, and opening an IB account i would like to know if this could be more, less or the same in 'live' trading. Does anybody know?
  2. man


    bidask spread. overfitting. nothing tells more than trading
  3. TraDaToR


    It's highly unprobable. Don't create your account based on such expectations.

    Can you say the instrument traded? If you get such results, major pitfalls are lack of slippage involved in tests, tests based on limit orders where you have no slippage but in fact you will never get filled, lack of liquidity in traded instruments for pair trading, or simply a bug related to software or datas.

    I've never seen a system giving 1K per day per future.
  4. man


    especially high frequency trading can be easily flawed.
    if you test for example assuming you can buy or sell
    the close of the next minute bar, let's say, you miss
    that you actually pay about half of bid ask spread and
    that makes a huge, huge difference. my fastest trading
    system pays about 25% OF CAPITAL into bid ask spread.
    and that is trading far from "high frequent" ...
  5. I know that testing is not the same as live trading, but it seems that my strategy is working, and that in live trading it will also make money for me, perhaps less, but 500$ per future is also nice
  6. man


    1000 quickly moves below nill if the assumptions are flawed ...

    what happens is that you fit a system to exactly these erroneous
    assumptions. the only reason reason why the system exists could
    be the false assumptions.

    but don't worry, everybody did that at least once ... part of the game.
  7. TraDaToR


    In fact, 500 $ per MONTH per future is not bad.

    You can get higher on some instrument like ER2, CL, DAX but it's pretty rare in real trading, at least to me...
  8. maxpi


    What kind of testing are you doing?

    If it is backtesting then you can expect different results in realtime. If you are testing it on a simulator or live market data your results in real trading might be more in line with your tested results.
  9. I am testing it on simulated data, and i make around 250 trades a day
  10. JackR


    Is your system an automated system that sends in orders or do you see a signal (on a chart or a pop-up or an audible alert)?

    If you must see\hear the signal and manually input the order I think you will have difficultly executing 250 trades a day. I tried manually executing a system I designed, back-tested and found to be profitable. I ran against a real-time paper trading account and could not keep up. There is also the occasional break for eating\eliminating.

    If you trade that often it I'd guess it is a scalping system. How much slippage did you assume in your testing if you used market orders? What commission rate? Will the future that you are trading actually trade as much as you'll need it to? Does your backtest work with market or limit orders? Can you set up the test to not execute until the limit price has been hit "x" times? In the real world your limit order goes into a queue.

    There is lots more to consider, that's why they sell books on this stuff, but I hope these and the other comments above will make you aware of the pitfalls. Review your test results and test assumptions before you throw any real money into the fray.

    #10     Jun 6, 2007