PureBytes Links
Trading Reference Links
|
>Okay, so help educate me. In the case of James P. O'Shaughnessy's work,
all we really know is that he had a couple of bad years at the same time the
market had some good years.
couple of bad years? i will show you below how pitiful this is....
We don't know if it will fail over the long haul. The work may be flawed,
but it looked pretty solid to me.
and what scientific test did you run to assume that his work looked pretty
solid? i guess you heard him speak once or read his book and you liked him
as a person so that qualifies his work to be pretty solid? maybe you played
golf with him or you are Irish too and that qualified your thinking that his
work was pretty solid. or maybe on the outside which i doubt - you run
zillions of iterations of on a powerful Unix machine with software that
would make ts look like a 3 year old in wet pants and that qualified your
thinking that his methods looked pretty solid?
>He used cleaner data for a longer period of time than most of the folks use
on this list use in their testing. Assuming that James is using his own
work, at what point does he walk away? If you were looking at his work,
what tests would you do to see if it was flawed?
clean data? in fact maybe that is one of the reasons he has failed (his
system is too data dependent) or maybe he scrubbed the CD under the sink
with running water before inserting into the coffee cup holder the rest of
call a CD-ROM drive? i have not looked at his work at all - so how then can
i criticize it you may ask? ez i know and understand thoroughly how to
build a real model that will hold up over time - more importantly i didn't
let my personal wants and desires influence the process. if this guy has
had a couple of bad years then HE HAS NOTHING! NOTHING! at all. he needs
to go back and learn to really build a model, starting first with the
data...
>> Something that works for a decade is nothing to sneeze at. Robyn
250 (trading days) x 10 (years) = (a lousy, pitiful and pathetic) 2500 (data
points) this anint sh_t when it comes to building a model!
250 (trading days) x 7 (trading hours) x 10 (years) = (a substantial,
credible) 17,500 (data points) which i have systems that are 60 to 70
percent accurate on!
i am making a bold assumption that whats his name is using daily data at
least and maybe weekly data to test on? or maybe even fundamental data?
whatever the case he has NOTHING!
>In the case of the ten good years, what happens in year 11, 12, and 13?
When would you walk away?
you walk away when the system exceeds its historical test drawdown and then
still trade it until it gives half of that back then you have permission to
stop! ok
mb
>
>== Rob ==
|