|
PureBytes Links
Trading Reference Links
|
Stefan wrote:
>Expectancy as put forward by Tharp or Matulich makes sense. It expresses
>the edge a system has. The expectancy score ignores the ratio of
>transaction costs to average wins and therefore is somewhat biased in
>favour of marginal systems.
I wouldn't say that. At least the way I calculate
expectancy score in my _SystemQuality EL function on my page
http://unicorn.us.com/trading/el.html, the transaction costs *are*
included in the profits and losses that are used to calculate the
averages. At least I *think* so -- doesn't PositionProfit() include
commissions and slippage?
I'm not sure how it would be biased in favor of marginal systems
either. Both Tharp and I include the small wins in the calculations
but ignore the small losses as scratch trades -- which bias a
system's score toward the conservative side, because the average win
is reduced by the small wins and the average loss is increased by
ignoring the small losses. I even go one step further and eliminate the
largest win as a statistical outlier, but keep the largest loss in the
calculation.
A marginal system, then, having a slow equity growth, would be
penalized more than a "good" system.
>I personally use a simpler rule: No more than 60% losers, and at least 2:1
>avg win/avg loss. If a system fails either or both criteria I look elsewhere.
That's a good rule. You should realize that your rule is exactly
equivalent to demanding an expectancy of at least 20 cents per
dollar risked; i.e. (0.4 * $2 - 0.6 * $1)/$1 or better. Tharp
recommends a minimum of 50 cents per dollar for EOD trading (which I
can't seem to get), less for daytrading.
What's missing in your rule is the opportunity factor. Given two
systems with the same expectancy (say both have 40% winners and
a 2:1 avg win/avg loss), you would want the one that trades more
often, giving you greater opportunity to realize your expectancy.
That's where the expectancy score comes in, expressed as
expectancy*opportunities_per_year in my _SystemQuality function.
--
,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
// +__> Director of Research and Development
// \
// __) Unicorn Research Corporation -- http://unicorn.us.com
|