With a graduate degree in engineering, a love of statistics, and a love of NFL football I set out to create a predictive model for determining the outcome of NFL games.
There are two factors that go into my picks; the confidence (a complex function of the standard deviation between the predictions of several different predictive models), and the delta (the difference between the average casino line and the weighted average spread projected by the models).
The delta is used to determine a raw cumulative distribution function (z-statistic). I then take the raw z-statistic and adjust it on the basis of the confidence. This means that if the confidence is zero the adjusted z-statistic is 50.0%, i.e. a coin toss. This will not beat any juice (52.4% needed for most casinos US bettors have access to).
Generally events have an adjusted z-statistic (probability of winning) from 55% - 62% with weeks having a probability of winning from 62% - 72%. I am tracking my year to date results against expectation values. I'm running 58% on 59 YTD bet units (69% on best bets), right at the center of the distribution for expectation values.
Thanks for sharing this information here. It seems really very informative.
ReplyDeleteGold IPTV
Xtreame HDTV