Tuesday, January 31, 2012

Why No Model Can Beat the Market...

in the long run. And why that says little about the efficient markets hypothesis.

In a recent post, Scott Sumner says, "I’m still looking for the model that will tell me how to beat the stock market." He then goes on to claim: "Even the smart money can’t beat the market, except by luck."

My response is:

1) Of course there is no model that can beat the market steadily, over the long term; but
2) That does not in the least make the case for Sumner's claim.

And the key to seeing why both of these can be true is... historical understanding! Sumner's claim, in fact, rests on a form of scientism, which holds that, if you don't understand something by means of having a mathematical model for it, you have no understanding of it at all. There is only science and ignorance, and no other valid forms of knowledge. Which is self-refuting nonsense, since that claim itself is not a piece of scientific knowledge.

Here is a variation on the "Sumner hypothesis" to get us thinking about this the right way:

Sumner2: If you cannot give me a model that reliably generates symphonies on the level of Beethoven's, then, that means that composing good symphonies is a matter of pure luck.

I hope everyone can see that Sumner2 is sheer nonsense. (If you can't, well, I'm afraid I see no hope of us having any intelligent discussion on this topic at all, so I wouldn't bother posting a comment disagreeing with this.)

So why is it true that no model will beat the market in the long run? Because, the longer the model beats the market, the more people will start to use it, until it eventually becomes the market... and the market cannot beat itself! If the model says to buy stocks in January and sell in February, once everyone wants to do this, no one will be able to do it. Even earlier on in the model's rise in popularity, by the point that over half of all stock shares are being traded following the model, its anti-model (where a trader simply does the exact opposite of what the model recommends) will become profitable, and the model itself will make losses.

If that is the case, how can individual investors beat the market, except by sheer luck? They do so by judging the true nature of unique sets of historical circumstances better than do others (what Mises would call verstehen). Of course there is no model for doing that: to have a model means to have abstracted out certain common features of a variety of situations. To have a model is precisely to abstract away the truly historical. So of course there is no way to model Warren Buffet. (And, similarly, Beethoven's virtuosity consists in knowing, after precisely the specific notes that have come before, just which note to go to next, a judgment which cannot be generated from a model that deals only with generalities.)

And the fact that investors sometimes judge these circumstances well, and sometimes poorly, is no evidence at all against the cases of good judgment actually being good judgment rather than luck, unless one has already bought into the scientistic hypothesis that only scientific-mathematical knowledge is real knowledge. The fact that Kobe Bryant misses 50% of his shots does not mean that it is pure luck that he is in the NBA and I am not.


  1. Good one Gene. Check out this article if you have time.

  2. Market models are out and the prisoner's dilemma is in!


Career Fair Organization

A dark thought came into my mind. I imagined the hosts of career fairs plan eight hours worth of events for their participant companies. ...