|
PureBytes Links
Trading Reference Links
|
At 11:53 AM 9/10/2005, Donald Beck wrote:
>The issue of the chart re-adjusting the bar or signal
>takes place when a price gaps, not because of tick
>time sequence.
>
>If your active order was a buy at 101.10, and trades
>took place at 101.01, .02, .03, .04, .05, .06, .07,
>.08, then traded at 101.12, the signal would be placed
>at 101.12 (the traded price) so long as you did
>nothing to refresh the chart data. If you refresh the
>chart data, then the signal would move to 101.10 (the
>active order). Therefore when doing historic
>evaluations and or optimizations for strategies,
>erroneous results are present.
You do not specify the data compression but if each trade were a bar
(1-tick bars) then the buy signal would occur at 101.12. If all of the
ticks are inside of one bar, then the buy signal would occur at
101.10, as you say, since the application uses only the OHLC data
after the bar has closed. You can back-test "within the bar" to
whatever resolution you want in TS8 (and maybe TS200i - I don't
recall.)
>In 2000i, the order in which the ticks come in are not
>used in the making of the bar. The only time it is
>meaningful is on a 1 tick.
That is true on historical data using only the OHLC values, obviously.
Testing inside the bar, you get whatever resolution you chose.
>Omega uses "bouncing ticks" (which is adjustable and
>silly in my opinion) with Omega developed assumptions.
>I know this because I was one of the Beta testers and
>had many conversations with Omega Staff including Bill
>Cruz. This is one issue of many that I pushed on very
>hard (having bought 14 4.0 packages and submitted many
>unknown bugs to them for years on a number of products
>bought me a tiny bit of clout)for change. It took a
>lot of pushing and in the end they added a
>non-functional feature in the optimization section. My
>last towel thrown in, I began to try and work around
>issues rather than point them out to them.
The "bouncing ticks" was an attempt to "model" what happened within a
bar when you only have OHLC data. They described it accurately and you
can chose whether or not to use it. The only perfectly accurate way to
back-test is to test within the bar to the 1-tick level, which is
extremely time consuming. Everything else is an approximate model of
what happened.
I never could figure why people care about this anyway since a
back-test is only an approximation of what MAY happen in the future.
The errors from extrapolating past performance into the future TOTALLY
SWAMPS and errors due to when your trade got filled within a bar.
>Because my use of 4.0 goes back many years, I don't
>remember how they handled the data bar issue.
Mine also (about ten years) and I switched from using TS4 and TS2000i
to TS8 years ago so it is hard to remember much about them. I cannot
imagine why anyone would still be using TS4 with all it's limitations.
Bob Fulks
|