|
PureBytes Links
Trading Reference Links
|
Walter,
On Sun, 23 Jan 2000 18:50:51 -0500, you wrote:
>Why? That's where all of action happens. No leptokurtosis ... no cash,
>no flame-outs.
You are right as long as you force winning and losing trades to be
represented in one probability distribution (, which imo can mean a
restrictive interpretation of trading results, especially when
using "symmetrical formulas").
Modeling winning and losing trades in separate distributions (like I
discussed some weeks before) means less prejudice and can help
(provided that appropriate probability distribution "formulas" are
used)
=> to deal more precisely with "non-symmetrical" distributions,
=> to reduce leptokurtosis of the losses (separately) to get more
consistent results, e.g. in terms of smaller draw-backs, and
=> to address these (and other) aspects directly during system
optimization.
In other words:
| Imo, the empirical probability distribution of wins & losses does not
| depend from the market behavior only, but can also be influenced
| heavily by the trading strategy.
|Keeping this in mind, the shape of the probability distribution
[of the losses] (sorry, I should have added this here)
| should be included into the system
| optimization process to find systems with less leptokurtic
| distributions, which also give more freedom in terms of money
| management.
More general: We should try to pay more attention to the messages from
our market & trading data to make the best use of the full potential
of the market, and to avoid any relevant restriction or "special
effect" that comes from the models used.
mfg rudolf stricker
PS: Can you give some practical hints on how to improve system
generalization?
| Disclaimer: The views of this user are strictly his own.
|