Vegas is notoriously hard to beat. But as the NFL season gets started Thursday, there’s some promise for gamblers in betting on projected team win totals. Half-win totals from 8.5 on down seem to be very solid bets to hit the over, even when accounting for the vigorish, or house margin.
Money-line data is only available for a subset of the years for which we have win totals, but using a curated data set covering 2003 through 2018 provided by FiveThirtyEight contributor Greg Guglielmo, I calculated the implied probability that a team would hit the over for each half-win total below nine wins and compared it to the actual win percentage of those NFL teams. Surprisingly, there appears to be an edge.
Vegas typically wants a healthy margin on its action — the vigorish on win totals averages about 30 percent. But blindly betting the over on half-point totals from 5.5 to 8.5 wins since 2003 is nearly a push, with the house netting just a percentage point edge. And if you were to bet any of the half-point totals where the under is favored (indicated by a money line that is positive), you would actually come ahead, beating the house by 3 percentage points. Put in a way more familiar to gamblers: If you bet $100 on each of the 110 half-point overs in the sample with positive money lines, you’d come out ahead $330.
So why is there a systematic undervaluing of teams with half-point win predictions? One possible explanation might be that the public isn’t exactly sure how to value them; half-point totals aren’t particularly intuitive to bet. Additionally, perhaps there hasn’t been a widespread interest in surveying historical odds and win predictions, so bettors may not be particularly aware of the lack of calibration of the closing lines.
I ran this finding by professional bettor Seth Burn, and he pointed out that half-point win totals with the under favored don’t appear to exist currently, so perhaps the edge has already been squeezed out of the market. Or perhaps it’s a phantom edge that never really existed in the first place.