|
PureBytes Links
Trading Reference Links
|
This was a great thread!
Perry recommended to me
to look at using semivariance
as well as AMR (Average
Maximum Retracement) which is the
referenced preference over SR
in his book.
I agree that these measures are
excellent choices and are producing
great autodesigned systems.
Sincerely,
Mike Barna
President,
Trading System Lab
"Your struggle to manually write Trading Models is coming to an end."
"TSL is an Algorithm that writes Algorithms."
www.TradingSystemLab.com
The divisor in Sharpe Ratio (SR) is and always has been the sd of return
(not prices). Unfortunately, SR fails two of three criteria for good
performance measure as explained in Perry J. Kaufman in "The New Commodity
Trading Systems and Methods", look at pages 390 - 391 in 1987 edition.
Figure 18-1 demonstrates:
1. Figure 18-1(a): Consecutive small losses (bad System B) and alternative
small losses & wins (better System A) are the same according to SR
2. Figure 18-1(b): Large surges of profits (System A) and large losses
(System B) are the same according to SR.
In both cases System A is better than B and SR is not good in distinguishing
them. More details in Jack D. Schwager.
DC
|