[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Sharpe & K-Ratio - requisite time-frame - I



PureBytes Links

Trading Reference Links

This message was posted to the list yesterday but probably didn't make it because it exceeded the size limit. I will try breaking it into two parts.

Bob

-----

At 8:39 AM -0700 6/1/00, Dennis Holverstott wrote:

> > Why can't monthly or even weekly returns be used as a proxy for this
>> important measure ?
>
>They can. Or even daily. Bob Fulks posted some nice Sharpe code. You
>have to annualize the numbers. For the standard deviation calc, multiply
>the stddev by squareroot(periods per year) to annualize it.

At 11:22 AM -0500 6/1/00, M. Simms wrote:

>Thanks for the reply. I noticed that one of the net's largest hedgefund
>sites www.hedgefund.net is using Sharpe to rank the managers. They are using
>MONTHLY data for this.
>
>Why can't monthly or even weekly returns be used as a proxy for this
>important measure ?
>
>Why are futures an exception ?


The Sharpe Ratio is related to the annualized rate of return and the
annualized standard deviation of returns. That said, the calculations
can be based upon sampling the equity curve at any fixed interval -
days, weeks, months, calendar quarters, years, etc. (Sampling monthly
is not strictly fixed intervals because the number of trading days in
a month varies a bit.)

If you used weekly samples (neglecting compounding) you would
multiply the average weekly return my 52 and multiply the standard
deviation of weekly returns by the square root of 52 to get the
annualized values for each quantity. If you used monthly returns, the
corresponding numbers would be 12 and the square root of 12.

The annualized values should be about the same no matter how
frequently you take the samples. In theory, they will be independent
of how frequently you take samples if the returns have a "normal"
(Gaussian) distribution. In practice, returns tend to not quite be
"normal" but are a little narrower in the middle and have fatter than
normal "tails". The attached GIF picture shows this clearly. This is
a plot of the weekly returns of a futures trading system. The red
curve is the best-fit "normal" distribution.

I try to sample frequently enough to get at least 30 samples. So
weekly samples would work well for 30 to 100 weeks of trades; daily
samples would work well for 30 to 100 days of trades, etc.

I find that the Sharpe Ratio is the best single measure of the worth
of a trading system or the performance of a money manager or of the
performance of a mutual fund. It is also the only valid way of
comparing the results of using a trading system vs. buy/hold. The
Morningstar web site lists the Sharpe Ratio of mutual funds but you
have to search for it.

Continued in Part II to follow:
Attachment Converted: "f:\eudora\attach\Sharpe.gif"