|
PureBytes Links
Trading Reference Links
|
Phil wrote:
> My real point is the absurdity of a
> fixed 6 point stop for a long-term SP backtest. Probably sometime in 1999
> when the system was developed it was just right, but now it's too much too
> tight and in the past it was way too loose.
Yup. The stop should be made adaptive to changing volatility. I got that
idea from Chuck (the king of ATR). Surprised his own systems don't use
it. :-)
> Now you could claim that coding your own MM stops avoids this problem
> since, as I said, your system couldn't enter its MM stops until the
> next bar. That means your MM stops wouldn't be in effect on the same
> bar as the entry, so there's no confusion. But it ALSO means you're
> "naked" on the entry bar, with NO MM stop. That's probably not what
> you want, given that the position DID get stopped out.
Well, coding your own can solve the problem of adjusting for volatility
but not the problem of the order of the ticks. When you are entering on
a stop, you don't have to be "naked". You can enter the stoploss order
on the bar before the entry bar. But you still won't know if your
stoploss was hit without using shorter timeframe data. Pseudocode....
if whatever then buy xxxx stop;
exitlong xxxx - y*averagetruerange(z) stop;
I'm a big believer in using intraday data, even on EOD systems, when
there is any possibility of two orders getting filled in the same day.
ps - I got a private note from Chuck saying he is aware of the problem
and the results on his web page, while hypothetical, do reflect what
happened intraday. In other words, if the system stopped out, his web
page shows a loss, even if TS later books it as a profit.
--
Dennis
|