|
PureBytes Links
Trading Reference Links
|
IVT wrote:
>I do not know where you learned statistics, but I bet
>your low-lag smoother represents the expected price
>value far worse than statistically un-biased estimate:
>
>price tomorrow equals price today
>
>can't beat that.
Can easily beat that. With a biased estimate that assesses the
presence of noise contribution in today's price.
As I wrote earlier:
>> My own tests, for example, of statistical probability bands above
>> and below an expected price value do indicate that a smoother
>> with low lag and no overshoot represents "expected" price better
>> than anything else, and the indicator bands that result would not
>> be used as an entry/exit signal, but as a "reversion to the mean"
>> bias indicator.
In THAT specific application, you can't use actual prices as
estimates of expected price because the actual prices include noise
in their movements. At least, that is an _assumption_ one can make
when developing an indicator.
Is it "better" in terms of better profitability? The question
has no meaning. You can't develop a strategy around one thing
(like a statistical smoother) and expect to make a valid comparison
after substituting something else that the strategy wasn't designed
for. A strategy developed around prices without smoothing wouldn't
even use the same decision algorithms. For example, if you have a
strategy built around trendlines, it's meaningless to delete the
trendlines and substitute a moving.
-A
|