> >
> > If you want serious anomalies, try ranking football teams that play a
> > dozen or less games ;-)
>
> Computer rankings are highly suspect in NCAA football because the data
> is WAY too sparse to be credible.  Heck, the college hockey schedule
> hardly provides enough games for a statistically sound ranking.
>

I have a lot more confidence in the rating that’s done in football than I do in the rating that’s
done in college hockey.   One reason is the amount of money that’s involved, and another is that
football has much more data available for the analysis.

One thing that football has going for it is the $ involved in rating teams and in predicting who’s
“better”, that is, who’s likely to win games and by how much.  I’m not talking about the financial
implications of the BCS computer rating systems.  I’m talking about the amount of money that is
bet on football.

Wagering on football is ubiquitous – a multi-billion dollar pastime.  It’s also a very Darwinian
system in which the people who are good at rating football teams survive, and the people who are
bad at it go bust and leave the system.  It doesn’t matter whether their selection method involves
sophisticated computer analysis or a dartboard, if it works you live to bet another day, if it
doesn’t work, you’re a dinosaur.  A million individual systems is more likely to yield an accurate
result than a few systems, regardless of who well thought out they are.

It’s true not only of the individual bettors, but also of the bookmaker.  If the bookie
consistently does a poor job of setting point spreads, he’ll eventually go out of business.

Football is also better suited to computer analysis, because there are a lot of statistics that
are kept and published that can be factored in to the rating system.  Not only do you know that
State U beat Ivy College 27 – 20, but you can take into account first downs, total yardage, yards
per passing attempt, yards per rush, and a host of other statistics that can supplement the score
and win/loss in evaluating the teams.

Not so with hockey.  The commonly kept and published statistics, SOGs , penalty minutes, etc.
don’t measure things that are terribly meaningful.  It may be true that over a large enough sample
space that a stat that’s commonly kept, like SOGs, has a strong correlation with a stat that’s
more meaningful, like scoring chances, but we’ll never know that, because “scoring chances” is so
subjective, rarely recorded, and almost never published.

Clay


__________________________________
Do you Yahoo!?
New Yahoo! Photos - easier uploading and sharing.
http://photos.yahoo.com/