I have a question. By arbitrarily limiting the number of trades (presumably due to margin or risk constraint) we're randomly dropping trades from the portfolio and thus risk missing the best trades that make the day/week/month. How do we manage that? If we don't manage it we're basically introducing random filters into the systems and thus "diluting" our edge at best or worst turning a profitable system into a losing one in the case of bad luck.
What might be helpful is to see the portfolio of systems and their independent performance reports indicating what model risks exist. For example, say you've got a mean reversion system, it might not make sense to add another mean reversion system into the portfolio since the portfolio risk might double or triple as a result, i.e. the worst case day will allocate too much to this particular market dynamic. In order to mitigate this, an ideal system portfolio would have complimentary systems that act together to reduce portfolio level drawdowns, for example, a short only trend system combined with a long only mean reversion system. Showing this dynamic in action is what I suppose you're intending to do here and providing example system statistics w/analysis of the significant system risks would likely be the most beneficial way to do it. Mike
In days past I was a programmer but I'm not really clear what you are doing in the example. To me, combining strategies is pretty simple, as long as you have daily equity returns. I simply post the daily equity returns into a spreadsheet and then calculate returns based on allocation to the various strategies. This assumes that the portfolio is rebalanced daily which is not the case in my real trading, so it is not exact but it gives a good idea of the results. I am a big fan of the 80/20 rule, that 20% of the efforts gives 80% of the results .. and that is generally enough for me. I combined 3 strategies with correlations of -22%, 3%, 4% and max drawdowns of 18%, 27%, and 44%. The resulting combination had a max drawdown under 11% with annual returns close to the average of the three. Standard deviation was also greatly reduced as well.
I think that a solution could be to take the performance (you could use the daily/weekly/monthly/etc performance) for this two strategies and then use the =CORREL(A1:A2,B1:B2) function on Excel. If the number is positive, this two strategies will have a tendency to move in the same direction if negative, the opposite and if it's 0 they are uncorrelated. But be aware that there is a difference between let's say +5 and +98.
As described in the previous post, I used the Excel Correl function to find the correlation of the daily returns. It's been very useful.
No, I try to keep things as simple as possible. If I were to measure it in additional detail it wouldn't help me going forward because the correlation of my strategies is what it is and while I measured the correlation with a static number, the correlation itself is dynamic.
I'm a big fan of Spearman Rank Correlation for finding correlated pairs, but for do portfolio analytics, it really helps to be able to say that V = sigma' * C * sigma