Jan

7

One of the interesting fallacies that one comes across in markets is the part whole fallacy. If you correlate the whole with a part of the whole + a random number, you come up with amazingly high numbers. For example, the first quarter change in a year or an earnings is correlated about 45% with the whole year change by randomness assuming each quarter has the same variation and there is no correlation between the first quarter and the last 3 quarters. An exact formula is given in Biometry (page 573 on google) by Robert Sokal (no relation to the man framed by his boss). In any case, the relation between Jan and the whole year is mainly a part whole correlation, although it has ceased working in the last 10 years and is in the graveyard except for the oldest seasonarians. However, if the correlation between the first month and the last 11 months is positive, then by a modification of the formula the correlation between the first week and the next 3 weeks must be negative.

Ralph Vince writes: 

Vic,

Doesn't this get to the heart of the matter, that being that good minds get sidetracked into boobey-traps all over the place in our endeavor here?

On a planet where camouflage is the dress code du jour, where predation and it's avoidance often depend on deception, we end up — as cognitive beings, reflexively and relentlessly seeking patterns and relationships — looking for things in prices that ultimately deceive us (or at best, work until they do not, a cruel form of deception, longer-term).

Markets, as man-made constructions, are particularly adept it this. I am again reminded of Nabokov's Lolita, one of the greatest pieces of English literature in this opinion of this amateur critic (second only to his Speak, Memory), as one that most will not consider because it alludes to sex with an adolescent — the common disdain for that, a perverse ruse in itself keeping many from ever enjoying the novel.

In similar fashion, I think one must must must must MUST assume randomness, however unpalatable the idea of trading randomly-generated data may be. In fact, as a strategy, rather, as concern or description of the underlying character of what we are working with, one must craft their strategy under the assumption that randomness belies the data flow, yet, should it go into periods where the data becomes non-random-like, to have that accrue further to our benefit.

It's more difficult to do than it seems at first glance, but, from a personal standpoint, it has been the most beneficial realization in my trading life.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search