|
PureBytes Links
Trading Reference Links
|
TJ wrote:
> if you've based your system on 30 min data, then you're hosed with
> the bouncing tick crap. the smaller the bar interval, then less
> likely will you get screwed by bouncing ticks.
Au contraire, mon ami. It's not the bar size that gets you. It's
the number of bars in the trade.
I assume you're talking about the problem of hitting multiple stops
in one bar. As far as I know, that's the only time the bouncing
ticks "feature" can impact system results. If one bar encompasses
more than one stop, you can't be sure that the TS backtesting
simulation chose the right stop. If you make sure you don't have the
possibility of hitting more than one stop within one bar, you don't
have to worry about it.
One way to avoid that is to use teensy bars. Another way is to
design your system to use long enough trades that they all span
multiple bars. My system averages 25 bars per win and 11 bars per
loss, so I have no problems. I've been trading it for over 5 months
now (on and off for almost 2 years in earlier incarnations), and it's
done MUCH better out-of-sample in real-time trading than it did in
backtesting.
> personally, i think the smallest effective bar interval should be
> no smaller than your response time to execute a trade (including
> feed delay too). since 4 tick spoo bars are roughly a 1/3 of a
> minute, then that gives ya 30 sec to work the trade.
I like a slower pace, and fewer and bigger trades. That way 30-90sec
BMI delays, or broker delays, or me fumbling the order delays, etc
aren't a big issue.
Gary
|