Judging the accuracy of the model in predicting a game total means to determine the difference between the projected totals and the actual total scores of as many games that have been played; and, for the entire season thus far, the average delta for my model has been 9.4 points. That sounds like a lot; however, compared against the Vegas line – itself a projection - and this season’s games, it’s not that bad, as Vegas books have been 10.3 points off for the season. Moreover, the model’s delta is now down to 6.8 points as of Week 5, when only current data began to be used, while Vegas has been off by 9.7, which is not much of an improvement.
Now, as I’ve said, ‘I’ve been to this movie before’. I’ve never had a statistical model for NFL game totals that was consistently accurate over the long haul; and, despite its good start, it’s only prudent to point out that game totals are far more difficult to project than game spreads, and the Limper hasn’t, thus far, been exactly lights-out against the spread. Nevertheless, I’ve been putting my shekels down on my totals projections since week 5 and I’ve turned my season around; so, until that delta starts to climb, I’ll keep posting O/U picks.
Naturally, the posted deltas, lines and picks, are all based on the most current data available; so, beware of major player injuries, sudden blizzards, and wild line movements – because things can change, and they usually do.