The Motley Fool Discussion Boards
Investing/Strategies / Mechanical Investing
|Subject: Re: The Best Measure for the Best Blend||Date: 11/16/2007 6:13 PM|
|Author: StevnFool||Number: 203972 of 264210|
I want to lay out an idea for further research if it interests you.
You have done a great job in demonstrating through testing the possible use of Sharpe/GSD as a useful prediction metric.
I have commented many times before that (CAGR-rfr)/GSD is a reasonable substitute for Sharpe when looking at our typical screens and where rfr = the average risk free rate over the backtest period. RRR4 as suggested by me a long time ago and as used by some is the version of this that uses 4% as the risk free rate.
My main point here though is that Sharpe essentially has a CAGR component in its numerator and a GSD component in its demoninator. When I originally suggested the idea of looking for low GSD screens which ultimately led to using Sharpe/GSD, the logic was that the GSD component of the Sharpe ratio is a better predictor than the CAGR component and thus the GSD component needs to get extra weight. We started doing this by selecting low GSD screens first before selecting the high Sharpe, but using Sharpe/GSD is just another way of overweighting the GSD as we essentially have GSD^2 in the denominator now.
This raises the question. "What is the optimimum amount by which we should overweight GSD?"
For example, which of the following is a better predictor of future Sharpe?
Sharpe/(GSD^X) where X could be any value > 0
A suggestion might be to try X = 0.5 and X = 2 and see how they compare with what you have already done (X=1).
If you would be interested in testing this idea, I would love if you posted the results. If not, let me know and I might try it if I get time.
|Copyright 1996-2016 trademark and the "Fool" logo is a trademark of The Motley Fool, Inc. Contact Us|