The Motley Fool Discussion Boards

Previous Page

Investing/Strategies / Mechanical Investing


Subject:  Re: The Best Measure for the Best Blend Date:  11/19/2007  12:29 AM
Author:  Zeelotes Number:  204048 of 263489

Eric wrote:
I would guess it doesn't make too much difference.

For alpha I would definitely think it would not, but for other measures, it can make a huge difference.

The problem I've run into with Alpha, is that backtested performance is so far above the S&P 500 that when I try to optimize for alpha I get blends that have unacceptably (to me) high GSD's (I'm talking in the 30's). But that's using only monthly data.

The median GSD is 31.87 for ranks 1-4 and 28.75 for ranks 1-10. I personally would think there are other measures that work better so there is no use in using this one.

Another approach to your optimization is to use the Sharpe ratio, but modify the risk-free rate as a constant rather than using the treasury yield. For example, something I might call Sharpe(8) using a figure of 8% as the risk-free rate. You could vary that number and see how it affects future returns.

Here is a backtest of the Sharpe using various MAR values from 2-15%. I also have the ability to easily test the granularity of the data used so I included that in the backtest. This is using daily, weekly, monthly and yearly data. As you can see with the Sharpe-based sort below, the highest Sharpe Ratio is attained using yearly granularity, but the highest CAGR is obtained by using a larger MAR and daily data. The default in the list that follows is using the Sharpe with the 1 year t-bill.
MAR Fixed  Look-back Period