|
PureBytes Links
Trading Reference Links
|
> I think we are creating a problem where none exists. Don't think
> in terms of bars or time but rather samples. The smallest unit of
> measure is one sample so that's the base for any calculation such
> as standard deviation which is based on multiple samples. If you
> want to look at time periods other than daily, just change the bar
> interval.... weekly, 5 minutes, 10 ticks, whatever. The base unit
> for any calcs is still one sample.
No argument. But that misses the point I was making: whether that
one sample (day, week, tick, I don't care) is representative of a
random walk. Or, rather, whether that 1-bar sample is a good
normalizing value for the market so the resulting values are
meaningful.
E.g. let's say you have a market that trends strongly -- let's say
it increases by 1 on every bar, that's a pretty good trend. :-) The
StdDev of a 1-bar move is 0. Is that a good approximation of a 1-bar
move in a random-walk market? Not even close. Do we NEED a good
approximation of a 1-bar random-walk StdDev for this to work right?
Or is a 1-bar move in the market, even though it's definitely not a
random move, a good normalization value? I'm not quite sure.
Gary
|