Can I really win this way?

Discussion in 'Automated Trading' started by hopeful, Aug 7, 2010.

  1. hopeful


    I have programmed a strategy into NT6.5 , it is quite simple really , I'm very impressed with the results, it "seems too good to be true" , it's not all that good anyway, but here are the details:


    profit factor 1.92
    38% winning trades
    5 trades per day
    avg time in mkt 60 mins
    max conseq winners 9 , losers 10

    Any comments wood be kool. Thx.
  2. januson


    Have you curvefitted it?

    If you are optimizing your strategy, your strategy is wrong :)
  3. Are these simulated results? if they are

    (1) what was the timeframe
    (2) How many trades you got in the backtest
    (3) Avg. win to avg. loss
    (4) How many variables in the system?
  4. Retief


    Looks like a good system based on your stats. Put some trades on with a simulation account and if works there, trade it with real money.
  5. If PF is 1.92 and %winners is 38%, then avg win to avg loss must be 1.92 X (1 - 38%) / 38% = approx 3.1

    - how many trades do the stats cover?
    (there is something not right about seeing 9 consec winners for a system that only wins 38% of the time; either you have a nice big trade sample, big enough to see such an "outlier" event, or your winners are all concentraded in a relatively short timeframe part of your sample ... can you post the system equity curve? ... do the wins all come at once, and the rest are losers? ... was it a particular market event that triggered the wins?)
    - whether simulated or forward tested, do the results account for commissions/slippage?
  6. Of course smarty, but I wanted to check his numbers, not your ability.

    This is a really good comment.
  7. hopeful


    Thanks for your valuable input.

    I have optimized only one parameter and it didn't make much difference anyway.

    I ran the strategy again to include a bit more data, here's some more details:

    #of trades 1860
    % profitable 34
    avg trade +18
    avg win +132
    avg lose -42
    win/lose 3.15
    max conseq wins 6 , losers 17
    largest winner 0.3516, loser 0.00302 (not sure why it reports like this)
    I set the quantity to $50,000 trades, all in all out.

    I can play with the variables and get different results, but in each case it still gives a pretty good upward sloping equity line. However, it is not very robust in that it performs best on EUR/USD but doesn't impress on anything else. It doesn't like time-based candles either, it prefers ticks or ranges.

    I will leave it running for a week and see how it goes, then maybe I'll go live (it will be my first attempt at bot trading).
  8. Be careful using range bars in NT. Last I saw on their message boards there was a problem using them to backtest (not calculated correctly, giving bogus results)...

  9. maxpi


  10. %K = w - (1-w)/R = .34 - (1 - .34)/(142/32) = .34 - .21 = .13

    or 13%

    18% is wrong
    #10     Aug 13, 2010