Can't Believe Mean Reversion Backtest Results

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

  1. omega_350

    omega_350

    I FINALLY STARTED TRADING STRICTLY WITH ALGORITHMS FOR MY ENTRIES AND I DO THE REST AND I HAVE 14 DAYS WITH OUT A SINGLE LOSING DAY AND VIRTUALLY NO STRESS ANYWAY I HAVE DESIGNED A MEAN REVERSION STRATEGY FOR THE ES AND ZB AND I CANT BELIEVE MY RESULTS ON EITHER ONE ESPECIALLY THE ES PLEASE SOMEONE REVIEW FOR COMMON ERRORS AND SUGGEST ANYTHING I MIGHT BE MISSING
     
  2. It's perfect.

    You've found It.

    Time to go live, with all you've got.
     
  3. mean reversion works. Some say as much as 80% of the time. But when the market trends, it wipes you out.

    Imagine one of your longs gets filled at the high and never sees that price again. Then what? Stops won't save you because if you are using stops you will get chopped to death the 80% of the time mean reversion (the chop) works.

    Wouldn't it be nice if we could come up with a sytem that always works?

    It's gambling man, and you have to make a bet.

    But now you have a system to use when you want to make a bet on mean reversion.

    Then your choice is to stay out until you think the market is favorable to the chop, or come up with another system you can use when you want to bet the market will trend.
     
  4. dom993

    dom993

    How much slippage is assumed for stop-loss orders?

    In any case, I would say the avg/trade doesn't leave room for the performance degradation that always takes place once you go live.

    If you are really going to trade that frequency, do yourself a favor and look into the CME ECMP program ... that will save you quite a bit on the comms
     
  5. omega_350

    omega_350

    THANKS A LOT I HAVE BEEN RIGOROUSLY TRYING TO DISPROVE THESE FINDINGS EVEN THOUGH I CHARTED THE CONCEPT ADJUSTED PARAMETERS UNTIL I SAW MY THEORY IN ACTION BUT THIS IS ONE OF THE FIRST STRATEGIES I DESIGNED THAT HAS ATTAINED THESE KIND OF ROBUST RESULTS WITHOUT MULTIPLE DEGREES OF FREEDOM AND OPTIMIZING JUST A FEW PARAMETERS ANYWAY I WILL KEEP YOU POSTED AND PROBABLY POST MY LIVE RESULTS CHECK OUT THE ZB IF YOU DON'T MIND LET ME KNOW
     
  6. bone

    bone ET Sponsor

    I think it's a mistake to read too much into very high frequency trading strategy backtesting without accounting for the trade execution slippage characteristics associated with a live market order book - especially given how modest the average winner is. Not a critique necessarily, just a dose of reality. When the ES or ZB bid or offer trades, it usually goes all at once. In other words - if you miss the bid and it trades out, do you then join the offer at that price and hope for a fill, or do you hit the next best bid ?
     
  7. itsame

    itsame

    I can't remember where its from but I've heard that 3 or 4 out of the top 4 or 5 hedge funds are strictly algo driven mean reversion strategies.
     
  8. vikana

    vikana Moderator

    Just remember that your avg trade is less than half a tick for ES. That's doable, but extremely tight. You would do well by working to reduce the largest losses.

    One last thing: slip is something you need to worry about given how tight your game is. Assuming you generally use limit orders, try to set stop/mkt orders to two slips in tick.

    If it's still acceptable, give it a spin.
     
  9. backtest for 10 years, blowup in 3 days
     
  10. ocean5

    ocean5

    i touched her thigh and death smiled
     
    #10     Apr 4, 2012