Daytrading & Drawdown Management

Discussion in 'Strategy Development' started by halcior, Oct 15, 2012.

  1. halcior



    i have a daytrading system that i am currently forward testing. recently, it reached an equity high of +52R. this corresponded with a max drawdown of -3R from equity highs. not bad!!

    a period of drawdown then followed, taking me -14.5R from the equity high. the system is currently recovering back to its equity highs. which brings me to my question..

    would it be best to have a killswitch on the system at say, -10R from equity highs, at which point you would stop trading the system until new equity highs are made?

    my thoughts:

    the pros of this method would be that, should the system stop working, you will not trade it all the way back to zero. simply put, you would get out with a -10R loss from the equity high and if new equity highs are never made again, then you would never trade the system again.

    the cons of this method would be that you, if the system then recovered, you would have be accepting a -10R loss but not riding the +10R recovery back up to equity highs. this would affect your trade statistics.

    to me, it makes sense to have a killswitch at some point to protect my account should the system fail.

    does anyone have any comments?
  2. The answer seems simple;
    Back test your system and try to optimize the variable.
  3. If the system was working well for a long time before with a previous max drawdown of -3R then maybe consider some kind of kill switch at -4R or something tight like that. Then you could turn the system back on whenever it started making a certain amount of money again (e.g., +5R maybe).

    Alternately, you could experiment using a partial kill switch where you start trading less and less size as the system enters major drawdowns and then start gradually increasing the size again as the system starts recovering.
  4. Mr_You


    My thoughts are that you consider turning off a trading system when the drawdown is outside the drawdown results provided by your backtest or walk forward test:

    - Total drawdown amount for a single out-of-sample test period.


    - Total consecutive drawdown amount for out-of-sample test periods.

    So if your max drawdown for a single period (ie. week) is -$500 or max drawdown for consecutive losing periods (ie. 3 weeks) is -$1200 you would turn the strategy off if the drawdown excedes these amounts and periods because most likely your strategy is overly curved fitted.

    Atleast this is what I've learned from my studies.
  5. halcior


    thanks for the responses, and here i was thinking that this forum was terminally ill. i should clarify that the 14.5R drawdown, although getting up there, it was consistent with my backtesting results. the previous max drawdown is in the teens.

    the drawdown period has recovered and the returns are now into triple digits of R.

    autowealth, i am not going to optimise the variable based on the historical results. this system is not the result of data-mining and i want to keep the risk of curve-fitting to a minimum.

    thunderbolt, your solution is what i have decided to be best. i am thinking that i will cut risk in half at -5R, then cut risk in half again at -10R, and again at -15R. at -20R, i would be beyond the historical max drawdown (but that does not necessarily mean that the system has broken down). however, this point would be the time to turn it off and observe whether the system returns to new equity highs.

    the effect of the risk reduction method does reduce the returns slightly. but i think that it is a reasonable price to pay for the risk mitigation that ir provides.
  6. Handle123


    I actually do it a bit different, as I had back/forward tested for ten years my main ES daytrading method and since another for four years real time trading, when I reach 90% of the drawdown swing, I increase contract size. I as well increase size when I reach "normal" peak of losing trades in a row.

    So I think it really comes down to how far back you backtested it, number of sample size as well. I had little over 75,000 sample size over ten years as it is mainly a scalping method.

    But I do have rules in place when method just shuts down cause of volatilty as method does not do well at all when market is very excited.