|
PureBytes Links
Trading Reference Links
|
"Paul M. Zislis" <pzislis@xxxxxxxxx> asks:
>I'm trading using end of day data. On 1/25/99 I got a signal from my
>system which gave me an executed order on 1/26. On 1/26, Chart Scanner
>properly reflected the open position and I also got another signal from
>the same system which gave me an executed order on 1/27. After that
>trade I was doing some system testing, and found that Chart Scanner
>running on 1/27 no longer reflects the trade I executed on 1/26 nor the
>signal it gave me on 1/26.
>...
>Also, the problem can't be with my data because Chart Scanner last
>night gave different results than it did this morning, without any
>intervening data updates.
TS seems to be inconsistent, to put it kindly, in its handling of data.
I've never been confident what has to do with the feed and what is
plain mishandling of feed data by TS. For example, since BMI switched
its indices to mostly-all MIC, I haven't collected a single correct
opening price for those indices; they're all quoted as last Friday's
close. Is BMI not tranmitting opening prices, or is TS not collecting
them? I dunno...
In the past, when I was using Signal, the daily volume for the S+P
futures would be transmitted near the opening time of the next day's
RTH. Most of the time, TS would pick it up, but sometimes it
wouldn't; I assumed that sometimes Signal didn't actually transmit
it, and that that accounted for the inconsistency. Also, that volume
data wouldn't actually be available when received, except in a quote
window, but would magically appear later. The point of all this
rambling is that TS and your feed can interact in mysterious ways.
Also, there's a difference between real-time data sent directly to
a chart and stored historical data; when you reload data into a
chart, say, by changing the number of days of data on it, you load
stored historical data. I could easily believe that this phenomenon
explains most of Paul's problem. A similar problem has to do with
"off" days, like weekends - supposedly TS4 fixed a known problem
with the server making up its own idea of data for those days if
you started the server on a weekend, duplicating prior data or
somesuch. Perhaps this problem has not been completely fixed.
Does your system use the opening price of the day in its logic?
Maybe TS is doing something odd with that opening price for you.
If I were experiencing this problem, I'd get very suspicious of my
data, as a first step. Try printing out the data your chart is
actually using; you may be surprised. Try to identify the sorts
of times you experience problems - is it usually, say, Monday's?
When do you collect your EOD data? Do you see a difference near
that time? Maybe you'll gather enough info this way to devise
a workaround for the problem. Have fun...:)
If you do, be sure to let us all know!
And Tony Haas <sptradr@xxxxxxxxx> adds:
>But isn't that a TS problem if the math functions as defined by the
>core program are incorrect? Wouldn't single vs double precision come
>into play here? Round off error would be less a problem with double
>precision math. Will TS2000 have double precsion math?
No, according to testing by several folks on this list some months
back. Remember all the ravings by "Omega Man" and friends, that
turned out to be relying on TS2000's ability to call VBA appendages?
I'd be very surprised if Omega changed that part of the program since
then. And indeed, depending on the system details, lack of arithmetic
precision could make a difference, but wouldn't explain Paul's
inconsistency from day to day, on supposedly the same data.
>Great, let's all break out the hand calculators and verify each
>signal! That's why TS2000 scares me more than a buggy TS4 with its
>known faults. You can usually work around them.
>
>TH
And some of us stay with trusty old TS 3.5.10, which had probably the
best ratio of bugfixes to new features of any of Omega's products.
Peter2150@xxxxxxx wrote:
>Not sure it's just Tradestation. I used to work in the Data processing of
>oilfield siesmic and regular well log data, using DEC Vax computers...
>It was always wise to have an idea of the answers expected,
>to verify what the computer said made sense.
Absolutely. If you're relying on TS-generated signals for trading,
you've got to verify that they're reasonable. If you're doing
backtesting, it probably doesn't make sense to use a system that's
very sensitive to subtle differences in your data.
Jim
|