[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

the correct model for trading systems



PureBytes Links

Trading Reference Links

<x-flowed>
	When you adopt a trading system that gives you entries and exits, your 
profit on any individual trade is either positive or negative or zero.  The 
collection of these positive and negative and null results forms a 
distribution.  If the mean of this distribution is positive, the system 
will on average gain you a profit; if it is negative, the system will on 
average lose.

	A distribution with a positive mean represents what is called a superfair 
bet; a distribution with a negative mean represents a subfair bet.   If you 
imagine the histogram of the distribution to be made of rigid wooden blocks 
with a fulcrum at the zero point on the horizontal axis, a superfair 
distribution will topple to the right and a subfair distribution will 
topple to the left.

	The only way to win consistently in any gambling game is to make superfair 
bets consistently.  A system's theoretical distribution and its statistics 
are not available but can only be estimated using back data.  Any 
conclusions derived from back data are subject to sampling error and this 
error must be sufficiently small as not to invalidate the conclusions.

	The standard way to examine the profitability of a system is to back test 
it.  This is what TradeStation does.  However, the right way to look at any 
trading situation (and any investment situation) is to consider it as a 
gamble, with positive or negative payoffs derived from the distribution of 
payoffs, and to examine the distribution of those payoffs.

	Price movement is a stochastic process, and the task of the trader is to 
find conditions that cause him or her to jump on and off that process so 
that s/he on average buys low and sells high, in whatever order, yielding a 
superfair profit distribution.  A setup is a market condition that causes 
the trader to jump onto that process and an exit strategy, well-defined or 
improvised, is what causes the trader to jump off.

	Call the combination setup/exit strategy a "system".  An intuitive or 
discretionary trader can also be considered to have a system, or a number 
of systems, each with a profit distribution, although the entries and exits 
are not hard coded or as easily testable.

	A system is a gambling strategy and has a distribution of profit.   Each 
execution of a trade triggered by the system is a gamble.  It is a good 
gamble or a bad gamble according to whether the mean profit is positive or 
negative.  If the system is superfair, it will yield regular profits as 
long as proper money management is used.

	The standard deviation of the distribution is also important;  among other 
things, it tells you how consistent the system is.  Additionally, the tails 
of the distribution contain information that is valuable to the trader of 
the system.   The right hand tail describes the structure of the profits 
and the left hand tail describes the structure of the losses.

	The trader improves the performance of the system by reducing the leverage 
(weight times distance from the center) of the left hand tail and 
increasing the leverage of the right hand tail.  "Let your profits run" is 
advice that is supposed to increase the leverage of the right hand tail, 
and stop losses are meant to reduce the leverage of the left hand 
tail.  More analytically, the tails are rebalanced by understanding the 
conditions that accompany them.  Such conditions are sometimes complex and 
can only be unearthed with great difficulty.

	A successful trader is one who has identified, either analytically or 
intuitively, market conditions that generate superfair distributions and 
who regularly makes superfair bets.   A trader improves only by adjusting 
the tails as described.

	Anyone familiar with stochastic processes and gambling theory will 
recognize immediately that this is the right way to look at trading 
systems.  I have developed a software package for my own use that derives 
these distributions, displays them and their important pieces, uses pattern 
recognition techniques and genetic algorithm technology to improve them by 
incorporating interesting new measurements, conditions, indicators, and 
information.  The package has a lot of value for the informed trader.

	This is not a trading system, but rather a platform containing toolboxes, 
analysis techniques and algorithms that allow the user to analyze trading 
systems correctly and to improve them.  It is the right approach from a 
statistical and logical point of view, but no available software analyzes 
systems this way, which is why I built it.  It is not a TradeStation add-on 
but a stand-alone application, although it will accept input from 
TradeStation or from any other trading program that can export ASCII files.

	This program has the potential to become a standard tool for serious 
traders. I would be interested in seeing the program marketed.  I am a 
statistician and a trader and not interested in marketing, but if anyone 
out there has such an interest and also the capability, please contact 
me.   Please, no individual inquiries -- the development costs have been 
sizable.


_| _| _| _| _| _| _| _| _| _| _|_| _| _|
_| pisani@xxxxxxxx

_|
_| _| _| _| _| _| _| _| _| _| _|_| _| _|
</x-flowed>