|
PureBytes Links
Trading Reference Links
|
Haven't seen any replies to your post, but if you are getting them privately
I would be interested in being copied.
Myself - and I am on the road not at my destination - I normalise results by
looking at max drawdown and adjusting position size to normalise methods.
For eg. using 40% say as an acceptable maximum d/down, if I have backtested
something and it has a past max d/d of 20%, I would look to trade double the
positions as this equates to 40% d/d. Many say risk only 1%, 2% whatever
per trade, but I feel the net result of trades is far more important than
trade by trade. A method with 40% max d/d would be treated on a one for one
basis. One with d/d greater than 40% woiuld be discarded.
This method compares results on a similar risk basis - based on the past.
Other "sanity checks" that I apply are expectancy - generally the higher the
better - and risk/reward, using average return (monthly, quarterly) v. ave
risk (based on sd of returns). If they confirm with best method re d/downs
then this is obviously a good sign.
I am also currently looking at risk of ruin measures. All things being
equal (which they aren't), a higher winning % of trades is far preferable to
a low % - but unless the %s are very low (in my book 20% or less of
winners), then expectancy and other methods are better.
Also, note that mixing systems, across markets, timeframes, methods, and
analyzing as a portfolio will produce better results than trading just one
method and market. Example, Method 1 that suffered a 30% d/d combined with
Method 2 that had a 25% d/d in combined portfolio form may have had a worst
combined d/d of 20%, ie. results are improved. However (there is always a
however), at times of extreme market moves otherwise relatively uncorrelated
assets may well correlate, which means at the precise time you don't want it
you may be carrying positions that far exceed the "theoretical" risk you
were comfortable assuming.
Interested to hear others opinions
Peter Corban
> -----Original Message-----
> From: TWA7663@xxxxxxx [SMTP:TWA7663@xxxxxxx]
> Sent: Thursday, 1 April 1999 12:29
> To: Omega-list@xxxxxxxxxx
> Subject: Spreadsheet for System Results
>
> I want to rank systems that I have backtested on stocks and futures.
> However,
> I must admit that I do not know what is the best method from all the ones
> that
> I have read.
>
> I would like to start a discussion about what the list members think is
> the
> best ranking method. In addition, any web addresses that have sample
> spreadsheets of ranking methods would be greatly appreciated.
>
> Russ
|