Cuidado con la falsa teoría de la probabilidad. Y con las intuiciones engañosas

Cullen Roche nos avisa: la probabilidad aplicada a los mercados financieros es falsa. Falla más que una escopeta de feria. Es mejor no sacar conclusiones probabilisticas de las consecuencias de la subida de los tipos de interés por la FED. Tampoco es recomendable sacar conclusiones sobre el resultado de las elecciones.

In my book I tell a story about a friend in college who claimed to have a foolproof gambling strategy. We would simply watch a roulette wheel and identify a trend in specific spins and then bet against it until we won. The thinking here is that a roulette wheel has about a 50% chance of landing on red or black (it’s actually a bit lower than that depending on where you play and the number of green slots the casino uses to tilt the odds in their favor). So, if you play long enough and identify a time when lots of blacks or reds are hit then you can bet against the trend. (See nota bene below).

Of course, this is all wrong. But we see patterns where we want to and not where they are. In fact, each spin of a roulette wheel is its own totally unique event. One spin has no connection to what happened in any previous spin. So, the betting plan is doomed from the start since it’s based on a flawed premise.

Repito en castellano: toda apuesta en la ruleta es un evento único, no está correlacionado con el pasado. No está relacionado con lo que ha pasado en la jugada anterior.
La teoría de la probabilidad dice que la probabilidad de que salga rojo o negro es un 50%. Esto verdad y mentira. Es verdad solo en el plano teórico. Pero puedes apostar siempre rojo y puede que siempre salga negro. Que antes de que apuestes haya salido durante dos días seguidos rojo no garantiza que la próxima salga negro.

As the Fed raises rates today we are seeing all sorts of similar comparisons in the financial markets. Many pundits are citing past returns or rate hike environments as a guide to the future. But we should be clear about this. Not only do we have totally insufficient data with which to work here (a handful of hikes in the post-war era), but this analysis implies that the current environment has some connection or similarity to some past environment when we know for a fact that this environment is its own totally unique occurrence.

It is probably best not to read into today’s hike too much, but if you must it is useful to think of this event as its own unique environment as opposed to falling into the trap that says this environment will resemble the past just because a thin data set tells you so.

NB – I quickly identified the error in this logic and this friend still owes me several cases of Busch Light.

Aplicamos esto a las elecciones. La gente piensa - y se la conduce a pensar- que un parlamento "equitativo" es más rerpesentativo, traerá un gobierno más justo, las cosas irán mejor. Totalmente ilusorio.