A System Ranking Metric, 01/08/09
A System Ranking Metric for comparatively ranking and forecasting the relative performance of trading systems and traders across trading vehicles:
ranking = ((z * sharpe) / f$) * 10000 where,z = the z score of the system which represents the probability the results occurred by chance
sharpe = the monthly sharpe ratio for the system which represents the stability of the results over time
f$ = optimal f$, the kelly criterion for the system which represents the ability to leverage the system
10000 = scaling constant
Example of four systems which trade the SP e-mini futures:
System A has a z score of 6.8, f$ of 6768 and monthly sharpe ratio of .55,r = (6.8 * .55) / 6786 * 10000 = 5.5System B has a z score of 8.2, an f$ of 4348 and a monthly Sharpe of .16r = (8.2 * .16) / 4348 * 10000 = 3.0System C has a z score of 4.8, an f$ of 2664 and a monthly Sharpe of .20r = (4.8 * .2)/2664 * 10000 = 3.6System D has a z score of 7.6, an f$ of 5404 and a monthly Sharpe of .08r = (7.6 * .08)/5404 * 10000 = 1.1Which gives a ranking table of:System A     5.5
System C     3.6
System B     3.0
System D     1.1
And here's an example of a bond system which drops into the table above at position four:
System E has a z of 5.4, an f$ of 3877 and a monthly Sharpe of .17NOTES:r = (5.4 * .17)/3877 * 10000 = 2.4
System A is the best system I have ever seen, the ranking metric supports that.
My guess is that systems with a rank of less than .5 will have problems going forward.
When developing systems for new markets, the ranking system provides a baseline for evaluating the relative strength of the new systems.
The longer the look-back period (the above systems go back to 1997) the more reliable the metric, a general rule.
Future performance of discretionary traders can also be forecast by this metric.