|
PureBytes Links
Trading Reference Links
|
FWIW:
Carelessly designed systems and systems testing protocols frequently allow
unattainable trades to be made; they then happily log all of the wonderful
performance of these trades into the statistics. For example, some systems
get into and out of price shocks that, by definition, are untradeable.
Also, some systems will trade every minor turn in price, showing wonderful
profits. Obviously, anyone who trades such a system is in for a rude
awakening.
One cure that I use in system testing is a worst case slippage model. Many
testers use extraordinarily inappropriate slippage numbers for trading a
bucking bronco like the S&P (it's unusual for most system tests to allow
more than 1/2 big point per RT for slippage). The fairly brutal slippage
model that I use (day trading) will kill most of these systems. At entry
AND again at exit, the worst price in the 3 minute period for that side of
the trade is used. This usually amounts to multiple points of slippage per RT.
Computers can be used successfully in system design and testing. It's just
that they're also very easy to misuse.
Allan
______________________________
At 07:26 PM 8/29/98 -0400, MRLYNNG@xxxxxxx wrote:
>In a message dated 98-08-29 16:54:27 EDT, trade_jack@xxxxxxxxx writes:
>
><< here's my contribution to the snake oil system of the month club.
> beats the hell out of rjp's. nothing more than a momentum crossover
> system with a stochastic like filter to keep you in the trend
>
> i thought i'd found the grail, but the smell quickly reminded me that
> i had stepped into a cow pie instead >>
>****************************
>TJ:
>I am a discretionary trader so I am not familiar with the details of systems.
>However, I would like to have some fun with your system, if you don't
mind. I
>assume you do not trade the Cow Pie system so maybe we can analyze it to (1)
>show the deficiencies of system testing and (2) get suggestions about how to
>correct these deficiencies or (3) learn how to avoid problems with our own
>systems now and in the future.
>
>If you can post the code (and list the deficiencies if you wish), we will
have
>the list point out as many deficiencies as possible for testing systems as
>well as for the Cow Pie in stage one. Stage 2 will focus on how to avoid or
>correct problems. TJ, if you trade this system or don't want to disclose the
>code for any reason, just say so and the subject will be dropped.
>Lynn
|