System Ranking Metric, 11/18/06



I've been working on a system ranking metric which looks like this:
h System Rank = (z score * ratio of variability / f$ * c1) + (n^1/2 * c2)
where f$ is the largest losing trade / optimal f,

where the ratio of variability is the average return / standard deviation of returns,

where n is the number of trades and c1 and c2 are constants just to make the result easy to read
This is nice for looking at a group of systems because it tends to make collinearity and correlation stand out. This can be a sign of stagnant system design or show a lack of new insights into the market.

System ranking also exposes outlier systems - systems whose methodologies may deserve more attention and exploration.


Henry Carstens
Vertical Solutions
carstens@verticalsolutions.com