|
PureBytes Links
Trading Reference Links
|
UG writes:
>
> It _might_, however...
>
> This sounds a bit to me like a classic chaotic system. (And not just
> me, many others have written texts to the same subject.)
>
> Having said that, and knowing that chaoitic systems are _extremely_
> sensitive to context, to the point of being Heisenbergy, would not doing
> something given a recognized pattern be enough to change it, thus
> rendering your analysis and pattern recognition moot?
Yes supposing that I was trading enough volume to move the market in a
significant way...I wish!! Or perhaps you are saying that ALL emerging
patterns "self destruct" because they are detected and acted on...that
conflicts with what I see in price action.
I'm certain (IMO) that the markets are "chaotic", and what this means to me
is that (a) the structure of price action is "similar" at all levels of
resolution, and (b) changes are extremely unpredictable because of the
non-linear nature of the system (small changes in initial conditions, etcetera).
However, given this, I STILL see periods in which price action has a
"recognizable" patterns of behavior (i.e, trending upward at such a slope
and modulating between two roughly parallel boundaries) which change
suddenly to new patterns (suddenly price is zooming downward very steeply
for N days (new pattern), only to then reassert the previous pattern at
a new absolute position on the chart). I.e., "chaotic" doesn't have to
mean that there aren't periods of consistency; the two ideas are not in
conflict (again, IMO). And I'm not suggesting "predicting" the next
pattern of behavior; I'm suggesting identifying it early and then exploiting
it.
Such patterns can be taken advantage of IF they last long enough to get at
least 2 (ideally 3 or more) successful trades in before the inevitable
loser as the price action phase shifts. OR, if an approach is taken of
"trade it once, wait for new pattern to emerge", i.e., try to avoid the
failures.
> I'm no scholar in such things, so I may be grossly over simplifying (or
> complicating) the problem, but it seems to be that phase behaviour is
> either too chaotic to try to 'capture', or it's just another name for
> something that chartists (et. al.) have been trying to do for years.
Indeed, taking advantage of the current pattern (i.e. trading a ranging
market by fading the extremes or going with a strongly trending market) is
a basic of trading. Question is: can such patterns be detected early,
computationally, and can the proper adaptive trading approaches be applied in
time to take advantage of them in a way that leads to consistent profits?
And how many such patterns are there that can both be detected and traded?
(I'm sure one mode of such a system is "I don't know what the heck is
happenning!", and indeed this could even be the predominate mode...and such
a system might still work very well).
Bill, thanks for your suggestion about inquiring about "Pattern Smasher",
I suspect they aren't interested in sharing any really powerful technology
they've got going though. As for tradestation not cutting it, I'm not
sure, I see the biggest problem as not being able to sequentially analyze
multiple contracts in one run (and no I won't work with continuous data,
it's too far from the real thing to work with IMO), although...perhaps I
could fake it by stiching the contracts together by hand and putting in
"markers" (i.e, price = 0 or something) between the contracts and writing
the s/w to detect the markers. Hmmm...
-Kevin
|