The Unemotional Football Bettor by Scott Kellen There have been several books published in the past few years on sports betting, and we will attempt to review them all at some stage. Stanford Wong's "Sharp Sports Betting" we looked at a couple of years ago, and that one caught our eye because of the mathematical intensity he usually brings to his writing. Scott Kellen also has his eye on the numbers...

Putting your trust in Z-Scores

The book doesn't exactly 'wow' with its aesthetic construction -- it's a spiral bound work with some 165 pages of content (plus the usual appendix, contents, etc) -- but as they say, don't judge a book by its cover...although as analytically obseessed as we are, a color picture on the front would do wonders to the book's ability to grab a passer-by's attenion.

In the first few chapters Kellen writes about his experiences migrating from a novice football bettor, relying on tout services, to learning to craft his own handicapping strategies, of which the key component is his fascination with Z-Factors.

Z-Scores are an attempt to quantify the statistical significance of a set of data, and the likelihood that the numbers seen are as a result of some genuine bias as opposed to a mere random fluttter. Mike Orkin has long pioneered similar analysis with his Optimizer programs (not to mention being a participant in our Head-to-Head Challenge), and there are certainly more than a few handicappers who do emphasize the importance of putting won-lost records in context. Stanford Wong in his book had a table that shows the "Rarity of Good W-L Records" such that you have a 1 in 100 shot of going 62-38 by flipping a coin, but only a 1 in 1000 shot of making it to 66-34.

Z-scores are easy enough to calculate, and the higher the number, the less likely the results you see are due to chance and more likely, it is presumed, to continue to hold true in the future. Kellen's underlying approach to handicapping is to seek out methods which display a 3.00 or higher Z-Factor, so as to "develop sound, winning situations" as he puts it.

And so he does, presenting us in the next sections of the book with "27 time tested strategies averaging 62.3% winners over the last 20 seasons."

The framing of a method in each case takes the reader through the steps to create the strategy, oftentimes beginning with a somewhat positive subset and then adding extra filters until the desired Z-Score level has been reached. For example, one NFL strategy begins with looking for matchups between 'positive rush teams' (take offensive yards per carry minus defensive yards per carry, so 4.0 offense, and 3.5 defense would be a +0.5 or positive rush team) and 'negative rush teams' in which case the positive rush team has covered 1128-998 (53.1%). Then he adds in the qualifying rules that the difference in net rush between the two teams is 1.0 or more yards (gets you to 604-524 or 53.5%), and then he adds in a line parameter to only play such teams as underdogs or when laying five points or less, which gets you to 409-302 (57.5%) and a Z-Score of 4.00 -- enough to count as statistically significant.

Now of course, it's almost impossible to create a method for predicting NFL games without to some extent relying on the knowledge of what happened in the past, but there are dangers with manufacturing approaches since you can take virtually any sample of data and backfit things to come to a subset that can pass the Z-Score test. It's very unlikely that methods devised in this way will continue to hit at the same rate since they have been optimized on the past history (eg why five points, not six? well that must be where the best slice is of the data). On the other hand, and this is important, they don't need to hit at the same rate to be good strategies to employ -- something may be 60% over the past five years and drop off to 55% in live usage over the next five, but that would still be very valuable information to know.

Kellen's work is so methodical, showing you year by year results that allow him to weed out findings that might be based on short periods of success, that you do gain confidence in his methods and their ability to be successful going forward. "Blind strategies" that arise from simply stumbling across and tweaking a set of data are to be avoided, but when the thinking behind an approach is solid and rational, as is the case with the ideas presented in this book, then there is every reason to believe some genuine nugget of truth about how football games play out has been discovered.

Indeed, more than a few chapters tackle subjects long promoted on TwoMinuteWarning, such as looking at Turnover Difference stats in a contrarian light.

Overall then, we can state this is an excellent buy at the $39.95 price, and ideal for someone looking to build a set of methods to use in handicapping the games for themselves. Of course there are the numerous obligatory references to Scott's own handicapping service for those who just want "the picks"...

The Unemotional Football Bettor can be ordered directly from Scott Kellen at SixthSenseSports.com