|
PureBytes Links
Trading Reference Links
|
Bob,
Sadly I dont support Ryan's approach either. I read the
book 5 months ago after getting it out of the local library.
I was very pleased that I hadnt paid for it.
Basically Gary's arguments are valid although in his latest
email I think that he hastily said "That means you require
increasingly large profits per contract to increase your
position size by 1" when I think that he meant increasingly
large profits overall (not per contract). There is also a
good argument in the book reviews at Amazon.com that
demolishes the book.
If you were lucky enough to use the method when you started
small then you just didnt hit a bad drawdown early on. I
suspect that a montecarlo analysis would show the flaws
pretty well ... and look forward to testing it on Alex's
spreadsheet once my paypal account is set up.
I'd like to discuss a different method which was originally
posted at Chuck LeBeau's site. This is a variation on
optimal F in which you don't decrease the number of
contracts traded as the drawdown occurs until you hit a
critical point (for example at the historical maxDD). This
has some intuitive attraction because reducing contracts
traded results in a reduction of the systems expectancy
(when you trade your way back from a drawdown with a lower
number of contracts).
I tested this out of curiosity optimising it and the
traditional Optimal F to give the same percentage maxDD.
What I found was that OptimalF had a higher percentage bet
size (because as a drawdown goes on it bets lower so it can
continue longer from a given starting percentage) which
would tend to overcome the reduction in expectancy. I found
that if you had long strings of loosers (as you'd expect
from time to time especially with a low win ratio medium
term system) then optimal F was much better. If, however,
you had a system with a high win ratio (and thus lacking
many long strings of loosers) then you were better off not
reducing the bet size. Non-reducing optimal F (or NROptF)
gave much better results than Optimal F.
Caveat: I invented the strings of trades to look at the
characteristics of the two approaches as I was looking to
see where one might be superior to the other. When I have
got Alex's software I'll be able to look at the
distributions with greater validity.
Does anyone have any thoughts on this?
John
|