Totally agree with the diminishing the spread point as a way to go but I have had a chance to work it through yet, and it’s interesting to know Mr Winston feels the same way.

Re: early in the season, I don’t have too much faith in the numbers until about week 5 or 6 of most sports. I keep track of the running standard deviation to know where it sits in respect to historical values and make a judgment from there (I believe Mr Winston uses a different technique – possibly something Bayesian re: one of the recent comments on his site). Incidentally the least squares way would’ve had last nights USC game bang on (USC by 7.9 no weighting, USC by 9.6 at 90% weighting).

Re: the Levitt paper – interesting read – it has it’s flaws (due to how they got the data, I believe it was from something like the Hilton Super Competition) but it illustrated how bookmakers take advantage of gamblers biases due to preconceived ideas and effectively how they didn’t balance the books by setting the line to extract the maximum profit based on the bias of their clients and not to reflect the match-up or balance the sides up. As a result it’s my personal belief that bookmakers leave themselves open to least squares/srs style system players as the line isn’t a reflection of how they expect that match to end.