What type of drawdowns do you go through with your system

Discussion in 'Automated Trading' started by trackstar, Jan 28, 2009.

  1. tommaso

    tommaso

    Thank you RoughTrader. I see your point. It's not that I wish to reinvent any wheel...

    I use this approach, because I feel it more suitable to the class of strategies I use, which are stricly *intraday*.
    I regard each trading day as an entity and a story completely independent from the other ones.

    Clearly, by "summation" it is still possible to create and use an equity curve, but all my interest is mostly focused on the minimization of the following ratio:

    Ratio
    = Avg daily Final Balance / abs(Min "temp" daily Balance Seen)

    because that is the criterion I feel (so far) more suitable to define an intraday strategy better than another one. So far I am satisfied with it. Probably my requirements are more demanding, but I prefer to sleep at night! :) or at least stay up and work without too many worries :))

    [Probably an implication holds so that the above criterion also ensure good strategy, defined using drawdown computed on an equity curve, but that should be investigated ... ]

    Besides, I personally define "Robustness" of a class of strategies the mean (sampling) Ratio over the space of all possible determinations of the class (intuitively obtained by varying the class-strategy parameters in all possible ways).

    A good strategy class should, in my opinion, have a mean ratio around or possibly above 15-25% with clearly 100% of Ratios >= 0 (and hence mean avg of the daily final balance >= 0). Clearly, the Min avg daily final balance should be as high as possible (for a given daily "seed" capital).

    [This, in my experience, is attainable only with a very few securities.]

    Tommaso
     
    #31     Feb 1, 2009
  2. Oh I see, I misunderstood your post. Yes, looking at individual session PNLs, in terms of avg session return vs. worst individual session return also makes sense. I never really thought about it, but I think this would also be a pretty good metric.

    rt
     
    #32     Feb 1, 2009
  3. Apologies if I missed it but do your equity curves include commission/slippage adjustments? This could have the effect of transforming flat periods into negative periods.
     
    #33     Feb 1, 2009
  4. 5 dollars per contrect and 1 tick on slip per side. running it live has never slipped more than 1 tick.
     
    #34     Feb 1, 2009