|
PureBytes Links
Trading Reference Links
|
I got very weird results at times in testing systems in TS6. Sometimes
backtest results would change, on the same data, if you opened and
restarted.
I have no confidence in the accuracy of any version of TradeStation.
It's great that they set up TradeStation World and users can report bugs in
version 7 and read the reports of others, and it's great that they seem to
be diligently trying to fix them. Unfortunately, since they have in effect
outsourced the quality assurance function to the user community, there is no
guarantee that serious bugs will be found and reported; it's kind of a
random process, and users may not catch bugs that are hard to detect, like
incorrect results in backtesting.
The right way to do QA is with rigorous, thorough test scripts that exercise
all functions in the program and do test all calculation functions. Although
users can check features, checking calculations is difficult for end-users,
so I wonder just how anyone will ever know whether TradeStation is
calculating results accurately.
Another important aspect of good QA is "regression testing" -- after bugs
are fixed, run test scripts again to make sure you didn't introduce new bugs
and make things that worked before now not work correctly.
There are tools that can automate some of this. This whole QA process is
well known in at least segments of the financial software industry and is
not a mystery -- TradeStation just isn't doing it, presumably because they
don't want to spend the money.
If as a pragmatic matter they need to outsource much of QA to users, I
suggest that they at least do good internal QA in areas that end users are
unlikely to be able to cover for, particularly the calcuations.
> -----Original Message-----
> From: Mike Eggleston [mailto:mikee@xxxxxxxxxxxx]
> Sent: Wednesday, April 16, 2003 3:19 PM
> To: omega-list@xxxxxxxxxx
> Subject: money management in ts6?
>
>
> I have written a program external to ts6 to test some things.
> This program is generating (to me) really good results.
> I'm moving the logic back to ts6 to verify the results.
> Here's the rub. I have coded stops in my program and in ts6.
> Sometimes the numbers match. Sometimes it seems as if ts6
> stops the strategy too soon. In non-optimized mode for strategies
> of historical data does ts6 include the 'jitter' to make things
> appear more realistic? Might there be something going on that
> I've not considered?
>
> Mike
>
|