System Metrics Evaluation

Discussion in 'Strategy Building' started by Maverick1, Jul 7, 2016.

  1. Maverick1

    Maverick1

    Hi All,

    What would an impressive but yet realistic trading system have in terms of Win Rate, Profit Factor, Trade Frequency, Distribution of Wins/Losses?

    It seems like reliable systems with Profit Factors greater than 2, a fairly even trade distribution (unlike TF systems where bulk of profits may come from a handful of trades) that also trade regularly (especially daily systems)are very hard to come by?
     
  2. This should be an interesting topic -- I must imagine the responses vary greatly; o_O:)
    I'm part of the Pro-Discretionary trading crowd.
     
  3. Maverick1

    Maverick1

    Agreed LL. Have you seen real discretionary trading with the above profile and a profit factor north of 2?
     
  4. It goes without saying, someone that has that sort of system won't share the trades without receiving some sort of compensation. The next question would be how much would you be willing to pay for a system with desirable and realistic metrics that you seek?
     
  5. Maverick1

    Maverick1

    Hi CP,

    Yeah it's not so much about sharing trades, but rather whether ET members have seen trading systems with a profit factor over 2 over a large trade sample ( 300+ to thousands), in addition to the other metrics being desirable as well, such a good distribution of wins, win rate, sharpe etc.
     
  6. Sergio77

    Sergio77

    K-Pia likes this.
  7. K-Pia

    K-Pia

  8. Maverick1

    Maverick1

    Hi Sergio,

    Agreed. I should add that my question would assume that the purported trading system would have avoided the naive pitfall of optimization, i.e., would be based on robust parameters which are all chosen for good reason and not because they are the result of an optimization.

    So, have you seen a system like that which sports a high win rate, PF higher than 2, good sharpe etc?
     
  9. A system that i`ve seen(performance over 12 years).A few metrcis:

    1.Winning months - 70%

    2.Max DD - 17%

    3.Sharpe - 1.9

    4.Correlation vs S&P 500 - -0.11

    Goes for $20K per month or 20/2 in fees.
     
  10. Bootstrap is not useful in ML. It can be distorted by bias.
     
    #10     Jul 11, 2016