>At 17:51 08/06/2006 +0200, Magne Aldrin wrote:
>>Those of you that are interested in football in addition to
>statistics
>>may have a look at the web-page http://vm.nr.no containing winner
>>probabilities for each team during the FIFA World Cup in Germany. The
>>probabilities will be updated daily during the tournament. The
>>probabilities are based on a Poisson model. The model is estimated by
>>maximizing a penalized and robustified Poisson-likelihood, based on
>>expert opinions before the tournament starts and the match results
>>during the tournament.
>
>Why not simply use betting odds? http://www.oddschecker.com
>
>Regards
>
>Martin
Well, admittedly as a non-expert, I would say because the probabilities
at http://vm.nr.no/indexEng.html are free of several sources of bias
present in the betting odds:
1. Bookmakers odds are determined partly by bets placed, which are
influenced by subjective probability beliefs, but also other factors
such as patriotism. This is particulary apparent in a tournament between
nations. So, for example, if I wanted to bet on England, I would expect
to get a better price from a German bookmaker than an English one!
2. When betting odds are converted to probabilities, they sum to more
than 1, in order to allow the bookmakers to have an expected profit.
There is no information on how the adjustments from probabilities to
odds are distributed over probabilities of different size, so reliable
recovery of the probabilites is not possible.
Finally, suppose you are in the situation where you want to place a bet
on the outcome of the World Cup, but you are out of touch with football,
how do you assess which bets are good!? The site recommended by Magne
seems to be exactly what you need.
Regards,
Dave
|