|
PureBytes Links
Trading Reference Links
|
Hi there,
A.Lo is indeed a 'financial engineer' @MIT
(http://web.mit.edu/alo/www/). He doubts that 'econophysicists'
discovered smth substantially new, and so do I. People studying the
data in question knew about 'fat' tails in distribution of returns
for ages.
Markets are obviously not 'efficient', how they can be if a market
cap of large companies can change over 50% is a matter of months?
However, I have surely not implied that TA can 'predict' them (you
should specify what 'predict' means). On the other hand, statistics
of the data and forecasting are legitimate empirical methods applied
to 'quantify' finance. The insurance and mortgage cos do not use TA,
they use 'econometrics', they do not use MACD, they use band-pass
filters, MAs in forms of ARMA and ARIMA, and datamining/forecasting
with neural nets, etc.
Whether all this constitutes 'science' is open for debate
(http://minneapolisfed.org/pubs/region/00-12/review.cfm).
For instance, it is a result with high statistical probability that
the global warming is linearly linked to the number of pirates
(http://www.venganza.org/).
And with neural nets I can easily fit my own signature,
so tools should be used within reason.
In any case, what we have is empirical datasets, and one should make
ones best to optimize odds via statistical analysis (which you may
call TA) in common situation with limited information. In this, FA
just adds another datastream.
The _hypothesis_ is that one can estimate ones odds of a given method
of investing, timeframe, and instrument used with some method of
forecasting/extrapolation.
If it is 'better' than fixed income, you may use it, otherwise you
must not (Sharpe:: http://www.stanford.edu/~wfsharpe/art/art.htm).
Forecasting usually works better for the past than for the future
(N.Bohr), and yet one is better off using it than listening to gurus
who always know where the market will be tomorrow, or using some
magic blackboxed 'indicator' of the same. We're talking about
probabilities, and even if the estimated probability of the event is
zero, it may happen. Therefore, I would be in the learning camp.
PS. As for 'econophysics', I do not think it's a good term, but
talking about random matrix theories, turbulence, and replica methods
may certainly help those folks to collect consultance fees.
--- In amibroker@xxxxxxxxxxxxxxx, "loveyourenemynow"
<loveyourenemynow@xxx> wrote:
>
> Hi Alex,
>
> thank for the interesting link.
> Not random walk just means that models based on the random walk
> hypothesis (gaussian distribution of the stochastic component) are
not
> accurate. It does not mean technical analysis is successfully
> predicting market evolution, but that other models (not random walk,
> not necessarily and I would add quite likely not technical analysis)
> can be more successful.
> By the way the two authors are not Princeton Professors
(MIT,Pennstate
> I think), and are not physicist but economists , and looking at the
> Nature article you link to, they do not seem to like econophysics
that
> much ...
> Econophysics papers are freely available on http://xxx.lanl.gov,
but i
> guess economist do not even read them, I personally like them
>
> Thanks
>
> Ly
>
Content-Description: "AVG certification"
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.430 / Virus Database: 268.15.6/568 - Release Date: 12/4/2006 3:20 PM
|