Fwd: [Husker] Point spreads

[Steve Reichenbach]

> I guess my original point was that point spreads established a month prior  
> to the game taking place, no matter who's establishing them ... are  
> meaningless.  
>  
> Dave Norris  

And my point is that you are factually incorrect.  The differences from
actual outcomes of predictions by computer programs such as Sagarin's
are undoubtedly larger for games further in the future, but based on
historical evidence they the best predictors available.

The point spread is designed to get equal money on each side of the
line.  Effectively, the point spread is a weighted consensus prediction
where many predictions of individual bettors are weighted by how much
they bet and then combined.  The point spread is a pretty good
predictor.  If the bettors were, on average, better than Sagarin, then
the point spread would be a better predictor, but many betters are not
very good (even though the worst ones tend to be weeded out by their
ability to continue to lose money) and so it is possible for good
computer programs to do better.  The stength of such programs is the
ability to account in an empirical fashion for large amounts of actual
data in a fashion that would have been successful for past events
(i.e., building a statistical model and optimizing it).  The weakness
of these programs is that they do not really incorporate knowledge of
the game and cannot account for things that can't be quantified.

I won't confirm any names, but was told that some authors (with whom we
may be familiar) may have used computer-based predictors (with which we
may be familiar) to bet money (legally, of course) on games in which
the computer prediction was much different from the spread.  Of course,
even if the computer program is better than the line, you may have to
bet a large number of games to have a high likelihood of overall success
because less likely events do occur.  So, it is still risky business
unless you're the house and have even money on both sides of the line.