This is several equity curves of a player

Discussion in 'Risk Management' started by harrytrader, Dec 5, 2003.

  1. <IMG SRC=>

    If you think that this guy has some special ability,
    it's an illusion since the MULTIPLE equity curves are drawn from a fair coin toss. This illusion has been discovered only in 20th century from mathematician Levy and called commonly "persistency of chance", more metaphorically "Fundamental Injustice Law of Nature" " and more technically arc sinus law. This law is NOT AT ALL INTUITIVE as one would expect that the number of curves below and above zero somehow counterbalance each other.

    Former thread here:
  2. harry, this isn't meant to be a flame or anything; I'm just really curious, do you trade?

  3. THAT'S WHY AN EQUITY CURVE ALONE DOESN'T ALLOW TO JUDGE A SYSTEM OR TRADER HAHA ! That's what snake oils sellers exploit and that's why some traders can self illusioned themselves for long.

    Yes and you ? Because if you are not interested by the problem of judging your performance adequately one can wonder ...

  4. Bolts


    Yeah, I get it. You could make a neural network or some other kind of system that would capture every single favorable coin toss in hindsight, but it would have no real predictive power. You could still get lucky, though. So how much out of sample testing or real trading experience should you expect to need to demonstrate that the neural network was bogus?
  5. Harry, nice post :D
  6. One of the more important findings from coin toss experiments -

    You can in fact identify patterns from say a sequences of 10000 trials that each producing some sort of bias for the next flip.

    Such patterns may/may not work at all on another sequence of 10000 trials.

    Translate that to the market, a system or method that supposed to perform must be fundamentally sound in terms of the design and rules must be sensible. Otherwise, you are just twisting around the paremeters to fulfill your fantasy of owning a model that perform best over your historical data :)

    One good measure is how stable the system / method performs over time.
  7. For a proper analogy to trading you need the coin flip to incur transactions costs. I think you'll find that if you do that, the chance of getting a set of distributions like those shown above is rather low.

    In any case, those distributions don't look particularly impressive. I certainly wouldn't back a trader showing me a curve that mediocre.