PureBytes Links
Trading Reference Links
|
It's natural to think about time when using tick charts. There is a
relationship between price and ticks (activity) to produce the bars.
There seems to be an indirect relationship between time and the behavior
of price tick bar volatility, since traders may play ball differently
(number of contracts per tick, liquidity, etc.) during a very active
trading session. But is there a relationship between time and ticks?
This is subtle but important when you design your strategies for tick
bar intervals, having time intervals in mind.
You can do stuff like employing variable period indicators and gaming
today's session tick count based on the last 3 days and so on in order
to force a relationship between tick and time intervals, but I think
this gets you into the dangerous world of curve-fitting. Unless you're
doing something with opening/closing session times, or tracking some
strange market phenomenon that produces profitable price patterns every
Tuesday between 11:01 and 11:06 am, I would avoid time when dealing with
tick charts. Tick charts are better at framing the price momentum and
volatility and I would design systems around that.
If you're not convinced, try to find another period in time where the
trading volume was similar to 09/2003 and test your system against it.
Finally, I think basing tick bar intervals on fib numbers is as good as
basing it on the tides. The actual bar number is as meaningless as it
is arbitrary. Your system should be robust at least for a good range of
intervals. But if your system is good when the interval is 300 and bad
when it is 250, then your system is producing a random event
masquerading as a profitable equity curve and is sure to fail.
When you find something that looks good, get into the habit of trying to
disprove it quickly by using a lot of data for statistical significance,
a range of variables to test for robustness, and forward test it for a
period into the future. You'll be amazed how many great equity curves
are in fact collections of random garbage.
-F
Abhijit Dey wrote:
When I first started writing strategies (intraday), I experimented
with tick charts. This was beginning of last year. Then I came up with
something that looked really good. I started trading it, and had one
month when I literally printed money (in % terms to my modest
account). I was ecstatic. This was 09/2003. Then a funny thing
happened. I not only didn't make money on that strategy, but lost 25%
of September's winnings in October. Although the losses weren't big, I
went into shock... my holy grail, not working!! It took me a while to
realize why... the volume in russell emini has almost doubled - all of
a sudden. I get double the number of bars for the same tick interval.
If you are thinking, well, of course - it wasn't very apparent to me
at that point of time. But I learnt my lesson, and threw away all my
tick strategies. My monkies run on minute charts now.
But, I still have a tick chart up (really like them), on which I base
my intraday discretionary trading. Trouble is, I am always tinkering
with the bar interval. I seem to like the tick interval best when I
get about 100-130 bars in the whole RTH.
Which brings me to the question - would any of you gurus like to share
some insights on how to get around this dillema? Like some kind of
analysis on the weekend to decide what interval to use for the entire
next week, maybe? OR maybe do this every nite? Doing this intraday
seems pretty disruptive.. I go like "Oh man.. big volume today.. bump
up the interval".
Also, I cut my teeth trading in Woodies chat room. I learnt a lot of
good stuff, and currently spending time unlearning some of what I
learnt. Over there, people seem to pick fib numbers for tick chart
interval, like 233, 377 etc. Is that fairly regular practice? Should I
bother? Or just pick nice round numbers like 300, 400 etc?
Thanks much!
Abhijit
|