What does Win% = 38 tell us?

Discussion in 'Strategy Development' started by trd, Aug 5, 2009.

  1. trd


    I recently asked a trading guru about a system having a Win% of only 38%
    and he wrote to me the following:

    "A system that generates 38% profitable trades averaging $250 per trade, against 62% losing trades averaging a loss of $75 per trade => would be a very viable system."

    He is very right, IMO Win%, much like PF, does not mean much.

    IMO, the most important characteristics of a system are:
    Number of differrent years tested
    AnnualAvgProfit% (with StdDev, Min, Max)
    AnnualAvgMDD% (with StdDev, Min, Max)

    Any other?
  2. trendy


    Yeah, one that has a little human interaction as possible, as that is clearly the weak link in the chain.
  3. His argument can best be summarized by a simple probability equation:
    Expected rtn = avgwin*%wins -avgloss*%losses
    Taken in context, he's absolutely correct.

    In theory, you could have 1 win trounce 99 losers, in which case hit rate is a bad metric. You could also have one loss trounce 99 winners (which is more likely a scenario in reality based markets), in which case, hit rate is also a bad metric. A more important metric, is how well does predicted performance match out of sample performance over large sample space. And how do you prepare for catastrophic scenario #2. I don't see these metrics much in the TA realm. But, they can be measured. Ask him what metric he uses to measure that.