|
PureBytes Links
Trading Reference Links
|
It is always fun to spar with The Omega Man!
At 6:51 AM -0400 9/21/00, The Omega Man <editorial@xxxxxxxxxxxxx>
wrote:
>Mr. Schwager covers this topic in his "Complete Guide to The Futures
>Markets" also.
Perhaps I am shortsighted but I tend to prefer the teachings of the
Nobel Prize winners to those of Mr. Schwager when it comes to
evaluating investment risks.
>To summarize, the problems with the Sharpe ratio are:
>
>1.) It does not distinguish between intermittent and consecutive
>losses.
So losses occur when they occur. You can be assured that they will
not reoccur in real trading as they did in backtesting so I fail to
see why this objection is important. Measures such as drawdown depend
heavily upon the order of losses, which, in a decent system will be
totally random. (If they are not random, you should take advantage of
the systematic loss patterns in the system design.)
>2.) The calculation depends (heavily!) on the time interval length
>chosen for measuring returns (weekly, biweekly, monthly,...)
Not quite correct. If the distribution of returns of returns is
"normal" (Gaussian), you get the same Sharpe Ratio no matter what
interval you use to sample the equity curve. And the equity
distributions of most trading systems I have seen is pretty close to
"normal". The trades occur when they occur, almost randomly, and the
Sharpe Ratio samples the equity curve at exact time intervals so you
would only get systematic errors if the trades were somehow
synchronized with the sampling period, which is unlikely. The
distributions of buy/hold equity curves is not quite normal (has
"fatter tails" than does a normal distribution) but the error caused
by this is still pretty small. And keep in mind that we are not
talking about high accuracy here. We are talking about the difference
between a Sharpe Ratio of 1 or 2 or 3 - not 1.03 vs. 1.07.
>3.) It does not distinguish between positive and negative
>fluctuations in returns (!) (All volatility is considered to be bad.)
This is a common complaint. Sure, everyone dislikes negative
fluctuations more than they dislike positive fluctuations. I
personally like an estimate not biased toward positive or negative so
that I can have an unbiased assessment of what is most likely to
occur. In real life we have a choice between various investment
options and if the estimates are biased one way or the other, we will
definitely make a suboptimum allocation of our resources.
>4.) It does not distinguish between retracements in unrealized
>profits versus retracements from trade entry equity.
So the investor who bought GE stock back in 1950 does not think of
his investment as worth it's present market value? I don't know about
you but to me a $5,000 retracement for an unrealized profit hurts
just as much as a $5,000 loss from a brand new investment.
>The Sharpe ratio really is a very, very poor performance measure for
>traders.
I guess it is a matter of opinion. Perhaps the promoters trading
systems, trading software, and trading books prefer to use their own
figures of merit because this makes their wares look better.
>"It is wonderous that with so little wisdom, the world is ruled in
>high places."
"In the land of the blind the one-eyed man is king"
Bob Fulks
|