Financial Forecasting Competition

[tenorio@ecn.purdue.edu]

At the risk of clogging everyone's mailboxes, and as someone involved with
the competition I feel the need to answer Bill's concerns:

>Motivated by the announcement on this list of the "First
>International Nonlinear Financial Forecasting Competition",
>I would like to raise a concern that has troubled me for
>some time.  I wonder whether it is really socially responsible
>to work on these things, or to support such work.
>
..

>
>What concerns me, is that the financial forecasting techniques
>I have seen are not oriented toward predicting the failure
>or success of individual enterprises, but rather toward
>identifying and predicting global trends in the flow of 
>money. 

The capacity of predicting the failure, or success of an enterprise is more
related to the analysis of fundamentals of the economy than technical
analysis. In technical analysis, a supposition is made that price reflects
all the information available to the market participants, as well as their
expectations. I have not seen all the entries of the competition itself,
but it is safe to assume that the predictors are not based on identifying
"global trends in the flow of money." As a matter of fact, the competition
is restricted to time series prediction processes where a single price
series exists, apart from the reality of the world that generated it. You
could say that this is the ultimate technical analysis test: a single price
series alone. The problem posed here seems to be harder than the problem
addressed by real traders, with the benefit of "market sideviews" and all
the rumors, etc. Which brings us to the motivation behind this competition.

This competition was meant as a follow  up on to the work of Weigend,
colleagues and collaborators. The Santa Fe Competition results, although
not geared towards economic time series, were basically all negative
towards their predictability. Meanwhile several software houses were making
claims of great predictability accuracy, as well as commercial systems of
great value based on non-linear techniques. It seemed very confusing to
know for sure whether or not such techniques are viable. The charter of the
competition, in spite of its limited scope, is to give the sketch of an
answer whether there is validity to the use of non-linear techniques on
this type of time series. There is a lot of smoke and mirrors, folklore,
and even more bizarre tales when it come to predictability of the markets.
Our goal is to shed some light on this, at least in this restricted
viewpoint.

Second, a number of researchers have looked at the economy as the result of
the interaction of many independent agents, operating with similar
policies. A sort of emergent behavior in a parallel distributed soup. Not
very much different from other systems such as biological, ecological, and
neurological networks. If this is the case, predictability of one has some
implications to predictability of the others, at least in a gross sense.

Now, commercial investment firms use all forms of prediction schemes in the
market: from fundamental analysis to looking at the planets. It is all done
in the name of making money. We do not claim to have an answer to the
"ethics" of different methods. Besides they all seem to have the same
(irk...) objective. For the market to work, there must be a seller and a
buyer. Therefore someone must always be on the wrong side of the issue. So
if company A did succeed and some investor lost a lot of money by not
believing in its success, after his/her loss, is this lack of belief to be
blamed as "promoting the destruction of a companies future?" Is that
judgment changed by the fact that we know that the investor used
fundamental analysis, whereas the "other guy" used a technical analysis?
How about if the roles were reversed?



 I am skeptical that this is possible in the first
>place, but even if it is possible, it seems to me that to
>make money this way is to be a parasite upon the financial
>system, rather than to serve it.

Some of us are skeptical too, thus the competition. But what is the reason
for the fear of its results? Again, I do not claim the authority to make
ethical judgments on our financial system, which is plagued by both
fundamental and technical analysts. From a strictly economic view point,
all market players are buying and selling risk. This form of financial
insurance is just a more sophisticated form of insurance on goods, which
has promoted the great prosperity of  western civilization in the last 500
years. Without insurance, ship owners would not venture the oceans after
merchandise, etc. Without it today, we would not have simple things like
the post office, UPS and the like, and farmers could not guarantee crop
prices, etc. (For lack of health insurance some voters elected our
President). It is all about minimizing risk, no matter how we do it. The
problem is that the person who assumes your risk, does not do it for any
one's blue eyes, but to make a living, and therefore will use whatever
means available. 

Our role is to minimize the risk of investment in technologies that will
not promote a true predictive gain. But I would like to remind you that
science cannot verify negative assertions, the scientific method is not
adequate for that. Further our resources are limited. The best we can hope
for is to shed some light, and that is our charter as scientists. And the
question is: "Where is the Truth concerning the performance of these
methods?"