|
PureBytes Links
Trading Reference Links
|
Well done Phil and my hat's off to you.
I've recently gotten a system almost up to 70% and I think that is "hot
stuff".
But the difference between 70% and 90% is HUGE....I don't think I will ever
get there.
One last important issue:
are the results based on 5 min / 20 min / or daily bars ?
How much data was involved in the back-testing ?
> -----Original Message-----
> From: Phil Lane [mailto:patterntrader@xxxxxxxxxx]
> Sent: Thursday, January 13, 2000 11:32 AM
> To: M. Simms; MarketMayvin@xxxxxxx; omega-list@xxxxxxxxxx
> Subject: Re: Pattern Trader - Performance of 90+ %
>
>
> ----- Original Message -----
> From: M. Simms <prosys@xxxxxxxxxxxxxxxx>
> To: <MarketMayvin@xxxxxxx>; <omega-list@xxxxxxxxxx>
> Sent: Wednesday, January 12, 2000 1:43 PM
> Subject: RE: Pattern Trader - Performance of 90+ %
>
>
> > I wonder if Phil might help "legitimize" his systems validity by:
> >
> > 1) indicating how long they took to develop and test
> >
>
> Well... I've been focusing and refining this stuff for at least 6-8 years.
> Started off slowly. Ramped up the effory considerably when I got TS. Spent
> 1000s of hours trying everything in the book. Tried all the traditional
> indicators and systems - couldn't come up with anything solid. Developed
> carpal tunnel syndrome that won' t go away.
>
> But eventually I got on to a couple things that worked. This was
> a few years
> back. I'd recommend "how to make money in stocks" by William O'Neal. The
> chapter on how to read the market is very significant. Something
> commodites
> traders would not normally get into.
>
> Found another thing that worked one day just by thinking about statistics.
> Just hit me out of the blue. Was thinking about "How frequently does (fill
> in the blank) occur, and when it does what are the odds of (fill in the
> blank) happening".
>
> > 2) how much coding was involved; were external DLLs required ?, etc.
> >
>
> The current entry rules are suprisingly simple. Exits are a bit more
> tricky - has to be done in such a way as to minimize "loopholes" i.e.
> missing a target by a hair and having the trade turn into a loser. Or
> anything else that might not have shown up in a backtest. I
> learned this the
> hard way, running earlier (proto) versions of the systems.
>
> > 3) methodology - pattern rec, trend following, scalping extremes
> > (countertrend), statistical,
> > or maybe a combination ??
> >
>
> pattern rec (not the high powered math stuff, just simple ohlcv stuff,
> counters, time windows, etc) for entries, statistical profile stuff for
> exits. Intermediate-term moves on daily charts.
>
> > 4) self adjusting ? neural net, etc.
> >
> no -
>
> So I've got the stuff to a workable level by now, but certainly am not
> FINISHED! After all this time I have only three things that work
> well enough
> to make a system out of. Also note that past performance, simulated
> performance in particular, is no guarantee of future results!!!!!!! And
> because of this fact I still have my doubts every time I take a trade.
>
> regards,
> Phil
> http://www.patterntrader.com
>
>
>
|