Can't Believe Mean Reversion Backtest Results

Discussion in 'Automated Trading' started by omega_350, Apr 3, 2012.

  1. sma202

    sma202

    whats the right way to test fills/slippage?
     
    #21     Apr 5, 2012
  2. ================

    I dont mind the CAPS or the higher % hit rate on your system;
    but both tend to get smaller as time goes on.LOL:D Not saying it has to be that way, simply a trend observation.

    And its not just the live slippage;
    but some of the best & brightest are in ES. 14 days without a loser, with your average hold time... exspect some changes .

    Congrats on the stress test, thats better than Citicorp perhaps...:cool: Hope this helps.
     
    #22     Apr 5, 2012
  3. Your largest losing trade is 10X the largest winning trade. This is a NO NO. You have done what almost every newbie does when they start backtesting. You have set a very small profit objective and a huge stop and you exit at the close of the entry bar (maybe). You were lucky for now but one day you won't see your price again for several days, not minutes, and you will blow your account.

    With a system like this I once many years ago made 1000% in a month trading bond futures and I thought I would be extremely rich in a year only to lose everything in a few days. Now someone will say: "why didn't you apply a stop". But if you apply a stop this system turns into a loser right away.
     
    #23     Apr 5, 2012


  4. Mean reversion is largely misunderstood.

    It means a security...
    Returns to a historical "Fair Value"...
    Relatively to other RELATED securities.

    That means is can work for niches...
    Of strongly correlated securities...
    Say, for example, the universe of all REITS...
    But is completely worthless for broad market indices...
    Or trading a single commodity...
    So applying it to ES is Noob Error 101.

    Also, the majority of your profits come from spread capture...
    And a smaller part (if anything) from the actual reversion...
    So if you have negative slippage...
    You are totally screwed in the long run.

    But then doing it properly is hard work...
    It's much easier to just give your money away.
     
    #24     Apr 6, 2012
  5. +
     
    #25     Apr 7, 2012
  6. ES is related to NQ which is related to YM which is related to GC which is related to CL which is related to ZN which is related to DX and as long as they all return to their historical mean values we are are all cool. But when one of them starts violating the norm on a consistent basis we are either in a trend or we are all screwed and if you're trading mean reversion that means all of the above.
     
    #26     Apr 7, 2012
  7. PLEASE STOP POSTING IN ALLCAPS... thanks.
     
    #27     Apr 7, 2012
  8. jakeM

    jakeM

    Backtesting in Ninja is very dangerous. It usually gives a very good result that will make you think you found the holy grail. Make sure to run your strategy using the Market Replay before going live and prepare to be disappointed. In my experience, using a multi-timeframe is the best approach in Ninja to make its backtesting more accurate.
     
    #28     Apr 9, 2012