|
PureBytes Links
Trading Reference Links
|
Pompatis and others:
What you are suggesting is that there is inter-trade correlation in
a trading system that can be exploited.
I believe that Ralph Vince has some good statistical assessment tools
for this in his Portfolio Management Formula's book. He conclusion is
that it is extremely rare for substantial correlations to exist in
trading systems. If you are concluding there are correllations based
on simple observation of trade history, you are almost certainly getting
misled by our human tendency to see patterns where none exist. Particularly
if your trade history is "small" (<200); you just don't have enough data
demonstrating sequence variation that will really validate correllation.
And then finally there's the problem of the lack of a stable distribution
of results in any system, due to changing market conditions....
I personally completely shy away from using any correlation methods for
bet sizing. I strictly use anti-martingale.
Also, I believe the "don't risk more than X% of equity" is too conservative
particularly when X is small, as in <=2%. With a good system (>75%
winners, decent average win size to loss size, reasonable max loss),
this figure is really underutilizing capital, and keeping risk of ruin
down in the <<1% range. I prefer to keep risk of ruin at the 2-3% range.
That's not risk of ruin in a single trade, that's risk of ruin if I
just kept on making the exact same bet size over and over...which I
don't, if I lose, I recompute for a new risk of ruin based on bet size
and new equity amount, and adjust positions. Anti-martingale...
At any rate, I typically find that with a good system, using this approach,
I'm risking more in the range of 10-20% of equity per trade. I know that
sounds high, but IF you have a sound system with a good real-time track
record, I think for myself I want to bet it hard, with reasonable
risk control. 2% max loss of equity is betting it way way too soft.
Just my thoughts and ways of doing things.
-Kevin
|