In many circumstances bookmakers' odds are likely to perform better than
model numbers since they can take into account factors that are too many
and varied to be captured by a model (team selection, weather conditions,
etc. etc.) However, the real reason for calculating match probabilities is
to try to beat the bookmakers, so it is pointless relying just on their
own calculations. The purpose of a statistical model is to try to identify
weaknesses in bookmakers' calculations, which may be present for the
reasons set out by previous discussants. The real skill though is in
adjudicating whether a substantial difference in bookmaker and model
numbers is due to a limitation in the model or a weakness in the market
price.
Now, so far as the allstat debate goes, it always strikes me as funny that
the messages that have the biggest effect of turning the list into a
discussion list are those sent by people complaining that a particular
thread has led to discussion (which usually just means there's been an
exchange of statistical views outside of that particular person's
interests). For what it's worth, imho, allstat would be a much more
interesting place if it did carry discussion of this type rather than
serve as an employment shop for pharmaceutical companies looking for
statisticians to carry out their t-tests.
But before anyone shouts at me for breaking list rules, going off-topic or
wasting precious bandwidth, don't worry, I'm leaving the list voluntarily.
Come on England, by the way...
Stuart
|