|
PureBytes Links
Trading Reference Links
|
In fact I test systems using CSI data, Genesis and TS data
and in every case of data inconsistencies it turns out that Tradestation
data is WRONG. I find the speed to be worse than 20% slower especially for
my long (ugly) system with 1500 lines of code - TS 8.1
is almost unusable- it's so slow
-----Original Message-----
From: Mike Symth [mailto:mqsymth@xxxxxxxxx]
Sent: Thursday, April 27, 2006 9:44 PM
To: omega-list@xxxxxxxxxx
Subject: RE: Optimization Speed TS2000i vs TS8.1
Bob Fulks Wrote:
>It was faster long before version 8.1.
No this is not true. TS2000i was 20% faster than all versions of
TS8.0. In addition TS8.0 could not use external ascii data. TS8.1
still can't access external ascii data as well as TS2000i
Bob fulks Wrote:
>Why anybody still uses TS2000i is beyond me..
I don't find all the so called "improrvements" in TS 8.1 over TS2000i
that worth while. I basically don't need all the "new" bells and
whistles. If I need speed I still have to write a C++ DLL, whether
TS2000i or TS8.1. While TS is clearly a superior trading and testing
platform it is still very slooow as my computation times have shown.
This slow optimization speed has not improved by very much from
TS2000i to TS8.1 .
Mark Simms wrote:
>What about the case of no DLL call, but Buy and Sell logic in place.
>This would isolate the DLL overhead penalty
That time was 9hrs 14min. This means that TS' s internal computations
take up to 90% of the optimization computation time. BTW the 1.5
years of 1min bars were from the 24hr JY currency.
|