Best Measure of Performance

Discussion in 'Automated Trading' started by 4EXJOE, Nov 5, 2009.

  1. 4EXJOE


    What is the best measure of performance for an unleveraged (cash only) system?

    Sharpe? Sortino?

    Some have told me that because the system doesn't use leverage, that the sharpe may not give an accurate representation of performance.


    Also. What are considered "good" sharpe and sortino ratios?

    Any other measures to use that would be more appropriate?

  2. Money
  3. P/L. All else is just statistical(?) masturbation.
  4. autocore


    Sharpe is fine.

    Because performance is just measured relative to the volatilty of your p/L , leverage does not matter at all. That is the reason why it is a good measurement

    Assuming a world of no fat tails, one can just say eg that a sharpe of 1 and more will occure at 15 of 100 players that trade random driven. So if you achieve 1 that is fine . 2 will be done by 2,7 out of 100.
    So finally if you have 1 that is good . More than 2 will only occure if you really have an edge .
  5. 4EXJOE


    Exactly. But understanding this is the difference between attracting capital and trading from your mom's basement in your tighty-whiteys.
  6. 4EXJOE


    Thanks for the reply, AC. What would qualify as a "good" sortino?

    Also, how did you come up with your figures (x out of 100) for each sharpe ratio? Is there a formula for that.

    Thanks again.
  7. autocore


    I am not too familar with the sortino concept but i have the impression that excluding the upside deviation does not make to much sense for me .

    Concerning the x out of 100 thing you should have a look at the so called "table of distribution" you can google it. .just replace the z-score by the sharpe ratio you have in mind and you can see how big the chance is to achieve it. For 1 you will see 0,85 for example . so 85 out of 100 will have a sharpe below 1 if they trade random driven.
  8. Syprik


    If your intention is to attract outside capital (insinuated), industry standard still revolves around Sharpe & IR. If I had to choose between Sharpe/Sortino, prefer latter given its omission of upside vol. These two, however, only speak to two moments of your return distribution, so you're only getting a partial picture. Plenty of critics will point out that the information drawn from higher moment's of return distributions can contradict results drawn from traditional mean-variance analysis = significant info is being ignored through over-reliance on Sharpe/Sortino.

    Any consideration for omega? I've been slightly more satisfied using a modified version of omega (ratio on single threshold as opposed to the ratio across thresholds) when testing strategies. Omega is the probability weighted ratio of gains over losses @ given level of expected return and has the benefit of completely describing the distribution of returns.

    My position is that the best measure depends quite heavily on the nature of your trading system(s) (ex, the nature of your return distribution) and your objectives. Can you provide any more insight regarding your systems/objectives?
  9. edbar


  10. dchazin


    I have been using a formula that takes overall net profit (after commissions) and maximum drawdown (i.e., the size of the biggest losing streak, historically) into consideration. Something like: ( NetProfit / ( 2 * MaxDrawDown + MinimumAllowableAccountSize) ). I think it is important to take into consideration, not only profit, but also the size of account necessary to generate that profit.

    #10     Nov 18, 2009