|
PureBytes Links
Trading Reference Links
|
Mike asked:
>When backtesting a system, should you always eliminate the largest
>winning trade... especially if it is very large?
I always do, when I calculate expectancy (see
http://uncorn.us.com/trading/expectancy.html ).
My reasoning is:
(1) If you have an "outlier" win that isn't statistically
representative of all your other trades, and occurs only once, then
you cannot reasonably count on that kind of thing happening again.
The trade should therefore be eliminated from your analysis to get a
more reasonable picture of the performance of the strategy. It's a
way to be conservative in evaluating your strategy.
(2) On the other hand, if your largest win is representative of your
other large wins, then eliminating it will have a negligible effect
on the win/loss ratio and other performance results (especially if
you have 100 or so other trades), so it doesn't hurt to get rid of
it.
>How about if one or two trades are responsible
>for much of a system's profitability?
If it's 1 or 2 trades out of 200 trades that make or break the
strategy, I'd look for another strategy.
>I'm testing a system across 200 securities and there
>are often 1 or 2 highly profitable trades.
I encountered a similar thing testing a strategy over 400 stocks in
the 1980s markets. My trading strategy would make money like crazy.
Then I discovered that this was due to ONE stock in my database
that happened to behave in EXACTLY the right way for my strategy
to lock onto (it was GAP if anyone cares). That one stock made
all the money; the other issues had about equal wins and losses.
Eliminating that stock totally destroyed the strategy. I was never
able to make it work well enough to be worthwhile to trade.
--
,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
// +__> Director of Research and Development
// \
// __) Unicorn Research Corporation -- http://unicorn.us.com
|