|
PureBytes Links
Trading Reference Links
|
This seems interesting, but it may be dangerous to use flat periods as a
statistical measure. For example, suppose you have a nice equity curve
representing a test sample of hundreds of trades going back several
years. It has a good Sharpe ratio, but it also has 5 flat periods, one of
which lasted 8 months, causing too much capital erosion because you are
living off of it. So, in a sense, there is only 1 flat period you need to
worry about.
Suppose you have an even better equity curve, but with two of these
"dangerous" flat periods. While the other measurements have statistical
significance, you are penalizing the curve that has a better Sharpe ratio,
standard error, etc., because it has just one more dangerous flat
period. In the end, you end up choosing the curve that has no serious flat
periods, only to end up having one shortly after you begin trading the
strategy.
Of course, when you compare one curve with a few flat periods against
another curve that has several, the difference should be detected by the
statistical measures that characterize a smooth equity curve.
But if you're worried about your best test strategy because it has one or
two flat periods, perhaps you should take a qualitative approach and find
out what the market was doing during those times, which would help you make
decisions about your strategy in the future. One 8 month flat period in the
past may reveal a problem with your strategy during certain market
conditions, but it doesn't say anything about how long the next flat period
will be, unless you have statistics elsewhere that shows predictable time
durations of various market hiccups and behaviors.
At 11:54 AM 1/6/2002 -0800, Alex Matulich wrote:
> >So my question, while the Sharpe ratio, gives us a very good
> >indication of the performance, is there more to it than just this
> >ratio?
>
>It seems that the problem you encountered (a high Sharpe ratio for
>result that have a long flat period) might be fixed by the "Sharper
>ratio" described in the Excel spreadsheet at
>http://www.medianline.com/sharperatioexample.xls
>
>The difference between the Sharpe ratio and the Sharper ratio is
>the way they penalize drawdowns versus profits. The Sharpe ratio
>penalizes upside as well as downside volatility. The Sharper ratio
>doesn't penalize profitable periods but does penalize drawdowns (and
>possibly flats; I haven't looked into it).
>
>I'm not well versed enough in TS yet to know how to implement it,
>or how to incorporate my OWN calculation as something for TS to
>optimize. (is that even possible?)
>
>--
> ,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
>// +__> Director of Research and Development
>// \
>//___) Unicorn Research Corporation -- http://unicorn.us.com
|