|
PureBytes Links
Trading Reference Links
|
At 08:15 AM 10/19/98 -0700, Gabe Hanover wrote:
>I would like to expand the discussion regarding indicators. In addition
>to the problem of lag, there is a more fundamental problem. There is a
>philosophical argument which I think is applicable here: You can not
>predict A with A.
>
>For example, take the syllogism: All beagles have brown eyes. Bob has
>brown eyes, Therefore, Bob is a beagle. What went wrong? We are using
>beagles to predict beagles. This is a logical fallacy. Likewise, prices
>and indicators based on prices, contain no inherent predictive
>information and cannot be used a price predictors. Why do they seem to
>work? Two reasons are: 1) They can identify, but not predict the future
>of trends; and, 2) When a sufficient proportion of market players act
>according to a favored indicator, it becomes a self fulfilling prophecy.
>While, I'm sure you can think of additional reasons, I would argue that
>none of them support the argument that indicators are valid price
>predictors.
>
>Where does that leave us? If you were fortunate enough to take
>Statistics 101, you know that multi-linear regression analysis is
>perhaps the best tool for predicting A. It does not use A to predict A,
>but rather, it uses independent variables b, c, d, e. etc. For us, this
>means using indexes, market statistics, sentiment indicators, interest
>rates, another stock, future or group and any other independent numeric
>indicator to predict the future of A, except A itself.
>
>Does anybody doubt this? Am I going over old territory?
>
It is certainly true that the more relevant information that one can bring
to bear on a problem, the better the prediction will be. It should also be
observed that restricting the solution to multi-linear regression may yield
an inferior solution to using non-linear, non-parametric multiple regression.
However, can an entity contain predictive information in and of itself? How
about an object coasting in a frictionless forceless environment. Its
future position can be predicted with certainty. Doing this requires
knowledge of the state of the system; i.e., that it is operating in a
highly predictive mode.
I would argue that the major problems with Gambler Indicators include not
only excessive lag, but also the non-existence of discriminatory
information that reveals whether the predictive state is high or low.
Allan
"There are no facts, only interpretations." - Friedrich Nietzsche
|