Max Drawdowns in systematic trading.

Discussion in 'Automated Trading' started by blox87, Sep 4, 2010.

  1. blox87

    blox87 Guest

    At what point would it be smart to pull the plug on a system or method you use?

    Personally I have an automated system that I am pretty set on at the moment that is working great now but I know the time will come when I must re evaluate.

    Could some of the experienced in this area please advise on your opinions related to .

    Trading your equity curve.

    Having a max DrawDown point where you will stop taking signals for the time being .

    Would that Max Drawdown point be decided upon through monte carlo simulation or by other means?

    Whether someone is trading discretion or systematic there comes a time when you must re evaluate your system/method so I'm just looking for some thoughts on this.


  2. Jack Schwager: "So far you've mentioned a 1.5 percent maximum loss limit on a single position and 4 percent on the entire portfolio for any given day. Are there any other risk management rules you use?"

    Monroe Trout: "We have a maximum loss point of 10 percent per month. If we ever lost that amount, we'd exit all our positions and wait until the start of the next month to begin trading again."

    Page 166, The New Market Wizards by Jack Schwager
  3. A 20% drop from initial equity is what a good number of professionals use.

    I have a different approach. I start tr\ding a system only after it has generated a high number of consecutive losers in paper trading. Then I stop the system if the equity drops to my initial bakroll.

    Conservative , eh? But it saved me in 2008. I ended up with +$10 in my account when most I knew suffered 50% up to 100% loss.

    I mean if a system loses some of you capital, even 1%, you should stop right there. 0% tolerance. There is no reason giving your money away. Your objective is to make money, not to lose.

    Thus, I recommend you set your cummulative stop at something like 0.1%. I know it sounds unrealistic for most who are willing to lose money, I am not. I am NOT giving any money to the market. Not even $1. This should be your objective.
  4. blox87

    blox87 Guest

    Largest DD so far for me peak to trough has been 5% . Equity has never dipped below my initial deposit amount and I don't plan on letting it get that far. If my max Draw down on monte carlo sim was 15% then I think 20% would be a good time to see if their was a major shift that has affected my system.

    Next question,

    When do you resume using the system?

    The 10% a month guy says just stop trading until next month.

    Any other thoughts or suggestions on resuming the trading after said max DD has been reached?
  5. nLepwa


    If the underlying distribution that you use for placing bets isn't stationary your system is doomed to fail.

    Therefore unless you can guarantee stationarity max DD is total ruin because the fat tail will always get you in the long run. (And the long run might be in 25 years as well as tomorrow..)

    If you want to be rigurous in your analysis, without stationarity, you need to consider the worst case (instead of 3 or 4 std's). And the worst case is -100% max DD.

    Now, of course, most people won't consider the worst case since it prevents them to use the adequate leverage they need to get rich fast. So they settle for an arbitrary DD, usually between -30% and -15%.
    However, ignoring risk hasn't proved to be a good mitigation strategy throughout the years. :)

    No wonder we see so many blowups..

  6. blox87

    blox87 Guest

    "If the underlying distribution that you use for placing bets isn't stationary your system is doomed to fail."

    What would be stationary and what wouldn't be stationary? Could you provide some examples. I don't understand what your saying.
  7. gov


    Maybe I can help; don't know but this is what I do.

    Take the last 100 trades and evaluate the current application of your edge. For instance, my trading yields about 16 %, roughly 58% wins, 42% loss. Now, since I use statistics tables, I round down to a 0.10 edge (more conservative). Now, I look at the capital units deployed and check the probability of having x amount of losses in a row. How are the current results compared to theoretical? At 100 trades they should be extremely close. If they differ greatly, you must answer why.

    Ultimately, you are simply running a number of time series back to back, and you need to analyze the effect of leverage on your risk of ruin.
  8. nLepwa


    There you go:

  9. blox87

    blox87 Guest

    A stationary distribution is a specific entity which is unchanged by the effect of some matrix or operator: it need not be unique.

    So basically anyone that has any kind of variable in their entries are doomed . Is that what you are saying?

    What about slippage?
  10. And if you're a Major League Baseball player and you strike out, you should retire right there. There is no reason to strike out. Your objective is to get on base, not to make outs.
    #10     Sep 5, 2010