|
PureBytes Links
Trading Reference Links
|
Mark Johnson wrote:
>I found an interesting new betsize selection algorithm
[snip]
>Then I compared the new betsize algorithm to that old familiar
>standby, fixed fractional.
[snip]
><http://traderclub.com/discus/messages/18/1431.html>
Interesting. I don't really see how this is different from fixed
fractional at all, if one rounds to an integer number of contracts
as one should be doing. I can find a fixed fraction that will
duplicate Chuck Lebeau's scheme almost perfectly (in terms of
performance, if not actual lots traded); his difference is that
he increases the fraction after making profits and reduces it
again after a while. It seems like he developed this because
recalculating equity for each trade seems too odious -- but isn't
that why we have computers?
This brings up a peeve of mine, relating to ambiguous terminology.
The term "risk" in the context of the traditional fixed fractional
betsizing model isn't really risk at all. In the context of the
conversation Mark mentioned, the term "risk" is ambiguous, because
the thread does talk about stops. In the context of fixed fraction
using "risk" is questionable.
Saying something like this:
> ) We will risk $2,000 on each trade until our equity is below $80,000.
or this:
> ) As you can see this is very similar to a fixed fractional strategy
> ) risking 2% but I believe it is better. It is simpler and it allows
...implies that $2000 or 2% is the most that you can lose on a
trade. This is false unless you have a stoploss out there set
to that amount. Otherwise $2000 or 2% is nothing more than an
arbitrary margin allocation which may or may not have anything to do
with the actual margin requirements to trade.
If you use stops, the risk you take is your initial stop. If you
don't use stops (as in a reversal system perhaps), the risk you take
is some function of historical losses. Arbitrarily saying "I'll use
fixed fraction position sizing, risking 2% of my equity" has nothing
to do with actual risk, if all you're doing is taking that 2% and
dividing by the contract margin requirement to come with how many
contracts to trade.
A position sizing strategy truly based on 2% risk would be: take 2%
of your equity, divide that number by your initial stop size, round
to the nearest integer, and trade that many contracts, but not more
than your equity allows for initial margin requirement. My own
testing, as well as that of Van K Tharp, showed this strategy to be
vastly superior to plain old Fixed Fraction when using some form
of adaptive initial stop (like logical stops or market-noise-based
stops). If you have a constant stop all the time, then this
technique is no different from traditional fixed fraction.
Another peeve:
> ) We are trying to be conservative when losing and extremely aggressive
> ) when having winning streaks. But after each big winning streak we
> ) become conservative again. That way we won't get into trouble after
> ) big winning streaks.
This assumes that the trades have a series correlation. This may be
true for portfolios (which is the context of the thread), but not
when trading a single issue of something. You can flip a coin 100
times and get "winning streaks" of several heads in a row. So what?
One cannot take advantage of these so-called "streaks" because each
event is completely independent of the previous one. Deciding to
be "agressive" when you think you're in a streak won't give you any
advantage, because you guarantee yourself that your largest loss
will be the final trade of the streak, and your smallest wins will
occur during the streak. This is an easy thing to test using coin
flips and Excel.
--
,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
// +__> Director of Research and Development
// \
// __) Unicorn Research Corporation -- http://unicorn.us.com
|