Strategy Quant Checks

Optuma Forums Feature Requests Strategy Quant Checks

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #55997
    AvatarMandeep
    • Topics: 37
    • Replies: 18
    • Posts: 55

    Hi,

    I recently relistened to a Andreas Clenow presentation.   One of the things he discusses to see whether a strategy has merit is to compare the strategy results to  randomly selected stocks in the universe using a random number generator.  Funny thing happens is that most of the random samples actually outperform the S&P500 over a full cycle and some of the strategies tested (question of robustness then arises).

    • Can we add a random number generator to the backtests to see this happen?
    • In the rebal of the backtest can we add a stop loss or an exit based on a condition e.g. close < ma(200)
    • Only rebal a stock if its no longer e.g. in the top 50 rank (even though we are selecting the top 20 stocks).  This apparently improves return / reduce risk.

    Thanks

    Mandeep

     

     

    #55999
    MathewMathew
    • Topics: 28
    • Replies: 1,386
    • Posts: 1,414

    Hi Mandeep,

    is to compare the strategy results to  randomly selected stocks in the universe using a random number generator

    This is what the MonteCarlo Simulation is doing. It randomly selects stocks in the same universe and also randomly selects the date range. What it maintains is the hold time so that the exposure to the market is unchanged.

    In the rebal of the backtest can we add a stop loss or an exit based on a condition e.g. close < ma(200)

    You need to think of the reverse of this in the filter. When we rebal, we first find securities which match the “Filter” script then we order them by the “Ranking” script. So in this case if you want a stop of close < ma(200) we just make the filter “close > ma(200)”

    Note: Make sure you use a Filter Script (greater than, less than etc) and not a Signal Script (CrossesAbove etc). A Signal script is a discrete event while a Filter script is a condition which remains true for a long time.

    Only rebal a stock if its no longer e.g. in the top 50 rank (even though we are selecting the top 20 stocks).  This apparently improves return / reduce risk.

    This one is a bit harder. Until we have Survivorship Bias for the top 50 I can not see how this would work.

    Hope that helps

    Mathew

    #56011
    AvatarMandeep
    • Topics: 37
    • Replies: 18
    • Posts: 55

    Thanks for your reply Mathew.   I see your point re the monte carlo – that’s a great point.

    In regards to the second point what about adding a simple % stop loss?  As an example if we rank on a 12m ROC momentum and define a entry condition as close()>ma(200) – quite often momentum can crash especially if its rebalanced on a quarterly/monthly basis.  During this period it may well close significantly below its 200d ma.  Would be great to see if we improve risk adjusted return with a simple 10%-15% stop loss.

    thanks

    Mandeep

    #56013
    MathewMathew
    • Topics: 28
    • Replies: 1,386
    • Posts: 1,414

    Do you mean to drop the equity between rebalances? eg if we Rebalance Dec 1 and then on Dec 14 the stock plummets 15% we kick it out then rather than waiting until Jan 1 to rebalance?

    Answer is “no” to that, but it is something we can look at add.

    #56023
    AvatarMandeep
    • Topics: 37
    • Replies: 18
    • Posts: 55

    Hi Mathew – yes exactly what I mean.

    If the stock drops x% or closes below its 200d ma on the 14th Dec we kick it out and goto cash.  We dont wait until the next rebal period.

    cheers

Viewing 5 posts - 1 through 5 (of 5 total)
  • You must be logged in to reply to this topic.

Pin It on Pinterest

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.

Close