Sebastian, you’re
dead right – this isn’t just meaningless academics for academic’s
sake. But I’ve never been convinced by that argument that you can make
something inherently random look like a stock price. I mean, sure, you can do
it…but then how much else out there looks like a stock chart? A patient’s
heart-beat? The soundwave-form of a Beatles song? The variation in birth-death
rates over time? None of those are purely random…
And what does “look
like” mean? Does it mean that you can model and design workable
(consistently profitable) trading systems on charts created from just pure
randomness?
Surely it’s
undeniable that because there are so many people (crowds) out there
trading/speculating (gambling?!) on the same thing then it sets up cycles? If
we presume that most TA tools measure the same thing and that there is a huge
element of commonality between various groups of trading strategies being used
(e.g. volume-based, oscillator-based, trend-following, contrarian, etc) then
one can assume that, for any given market, there are significant groups of
traders that will get excited at roughly the same time, given a certain ‘set-up’.
Then, when the price reaches a certain widely-perceived area of resistance/support
many start to bail out. Then they bail back in again when the price has
retraced 50% or some other Fibonacci level. That has to have some kind of
cylical influence…the duration and magnitude of which is dependant on the
degree of commonality between the traders (i.e. if everyone gets excited at the
same time then the move will be short and sharp, wheras if there is some ‘lead/lag’
time for everyone to get the message then it will be less so). Add to that the
fact that there are groups of traders acting on all kinds of different time
scales and you certainly have the recipe for creating something like what Jose
has represented in his “Synthetic Cycle Generator”. Only a minimal
amount of noise there (i.e. the randomness in the cyclic variation)…and
that looks like a stock chart!
Because it’s
so widely followed these days, TA has become a self-fulfilling prophesy to a
large extent. Therefore, (since most indicator-based strategies rely on cycles)
the cyclical element in the marketplace has, I would argue, actually increased since
Hurst wrote his
book in 1970. That’s why there’s so much more apparent volatility
and so much less ‘fundamental’ influence on prices. Unfortunately,
that hasn’t necessarily meant that TA works better because the pure noise
component has undoubtedly gone up with it, which also contributes to the
volatility.
That just
means that we need to constantly re-think our filters rather than ditch the
very basic principle of TA – i.e. that history repeats itself.
By the way,
that change in fundamental/cyclical /noise relationship over long periods of
time is why I also do not believe that it’s helpful in back-testing
systems over “massive historical data”. Markets change and so
should the systems we use.
From:
equismetastock@xxxxxxxxxxxxxxx [mailto:equismetastock@xxxxxxxxxxxxxxx] On Behalf Of sebastiandanconia
Sent: Thursday, September 08, 2005
11:18 PM
To: equismetastock@xxxxxxxxxxxxxxx
Subject: [EquisMetaStock Group]
Re: Synthetic cycles
"...This is the sensational bit! You can use random
noise, smooth it, and generate nice looking,
systematic effects. What
Slutzky did and what shocked the academic world at
the time was to
mimic an actual trade cycle using only random
noise..."
Burton Malkiel ("A Random Walk Down Wall
Street") describes an
experiment where the outcome of coin flips (+1 for
heads, -1 for
tails, and doing a running total) is displayed on
a stock chart.
After a few hundred coin-flips, the resulting
pattern of numbers looks
just like the activity of a "real"
stock.
These are not just meaningless egg-head, academic
thought
experiments. The implications are profound
for traders/investors
using most kinds of TA, including moving averages.
If your favored TA
method can't distinguish between
randomly-generated data and the real
thing, is it really measuring what's going on in
the market or is it
just measuring the characteristics of a data-set?
Since we know that most stocks travel together
("the rising tide
raises all boats"), can any indicator that
ignores the activity of the
overall market really be valid? Stocks also
rise and fall based on
earnings, dividends and valuations. Can any
indicator that ignores
these factors be considered valid? What
about economic factors?
Liquidity? Fed policy? Float
size? Short interest? Volume?
I stand by my original point that massive
historical back-testing
using the arbitrary mathematical formulas of the
vast majority of TA
methods only produces unimportant coincidental
correlations, and I
would welcome any logical argument or proof that
this isn't the case.
Luck,
Sebastian
SPONSORED LINKS
YAHOO! GROUPS LINKS