Tuesday, February 26, 2008

Overconfidence

Prediction markets have been having a little bit of time in the sun recently because of the US elections. Generally they're illegal, but people are allowed to place small bets on various aspects of the election (in America). Pundits don't much like them and have written quite a few articles explaining that they're no good and delighting in cases where the market gets things "wrong". BUT

  • The limitation on bet size is bound to make the market less accurate.
  • When a market say that there's an 80% chance of event A and then event B actually happens, this does not mean the betting market was wrong.
  • The claim is that these markets will be more accurate than us, not that they are unfailingly perfect.

If prediction markets are very imperfect indicators of truth, and they are much better than even the best experts, the conclusion should be that you are not very good at developing true beliefs about the future.

2 comments:

Trevor Black said...

One thought about prediction markets that is different from polls.

If a poll says one person is at 55% and the other is at 45%...

I would expect the prediction poll to amplify or leverage the difference. So 55/45 poll at a guess should be at the very least70/30 in prediction markets.

I would think an 80/20 prediction market should acceptably be around the 55/45 mark too, so it is not as big a margin as people think.

Maybe its an occupational hazard, but I think of a 20% chance of something happening as VERY high.

That is more than the chance of rolling a 6 on a dice!

Prediction markets should be much more volatile.

stuart said...

really, I've been surprised by how quickly the candidates odds have been swinging.