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

Re: S&P Systems ELA


  • To: John Cappello <jvc689@xxxxxxx>
  • Subject: Re: S&P Systems ELA
  • From: "Gary Fritz" <fritz@xxxxxxxx>
  • Date: Mon, 14 Jun 1999 13:15:54 -0400 (EDT)

PureBytes Links

Trading Reference Links

> A few weeks ago I purchased 2 systems from www.callnetuk.com
> 
> Since there was no non-disclosure or license agreement I have attached 
> both systems for your perusal.They both use close stops and 50% 
> trailing stops once profit is above $1000.Any ideas on usefulness of 
> these systems would be welcome.They are set up for Trade Station.

John, the systems in the ELA you sent did not use any stops.  Did you 
forget something?

If I remember correctly, this site came up on the Omega list a month 
or two ago.  The TS system reports show an average length of 0 bars 
in losers, and 1 bar in winners.

This is a major red flag.  Particularly in combination with the 
"tight stops" you mention (which are presumably built-in TS stops), 
this almost certainly indicates that the results are "cooked."  By 
using internal TS stops that are tight enough to create 0-length 
trades, you can create the appearance of fabulous results -- but you 
will find you can't replicate those results in realtime.  

The problem is that TS blithely reports it successfully closed out 
the 0-bar trade with e.g. the trailing stop -- but with no knowledge 
of the price action within the bar.  TS assumes the "best" and tells 
you the trade was profitable.  In reality you probably would have 
been stopped out for much less profit, and probably a loss.

TS's internal stops can safely be used if your trades are generally  
several bars long -- say, at least 4 to 6 bars.  That way you can 
usually be sure that the maximum favorable excursion (MFE) of the 
trade was hit at time X (on one bar), and the retreat from the MFE 
happened at some later time X+Y (on a later bar).  If the MFE *and* 
the exit happen on the same bar, you can't be sure that the MFE 
happened before the exit.

Bottom line, you CAN'T TRUST the TS system backtest reports for a 
system that reports 0- and 1-length trades.  Follow the system in 
realtime for a while and see how it *really* performs.  

The PDSIntra system is interesting.  It appears to be a version of 
the PDS system, but intended to run on intraday data instead of on 
daily.  That lets you test it much more realistically.

I tried running it on 30min SPX.  If you're willing to accept 
extremely low win% (e.g. 18%, not the 70% claimed on the website), 
you can run with a $250 MM stop and the system actually does fairly 
well.  In the last year, using the default parameters, it made $107k 
with only a $7357 drawdown, averaging $438 per trade even with the 
low win rate.  Higher MMstop values did worse.  However, remember 
that is on SPX, and you can't directly trade that.  It does much 
worse with real SP data.

So:  it looks to me as though the system does a pretty poor job of 
calling market moves.  The system report on the website seems to be 
bogus.  I'd love to be proven wrong -- it would be great to have 
systems with that level of accuracy!!  :-)

BTW, the correct URL is:
  http://www.callnetuk.com/home/tradingsystems/
Note, no space in "tradingsystems".

Gary