It’s schedule SZN! It is the time of the year when most bookmakers release their NFL win totals for the upcoming season. Predicting future records in sports is as hard as it gets. Teams may change drastically from one season to another. Regression to the mean, positively or negatively, plays a significant role, too. However, year in year out bettors across the world try to predict future wins by placing their hard earned money on season win totals.
Roughly two weeks before the 2018 NFL season, I posted the implied strength of schedule based on Pinnacle win totals. The published number by the bookmaker doesn’t always represent the exact win total. Due to the juice, the numbers usually shade towards either side. For instance, the Arizona Cardinals sat at 5.5 wins odds of -193 in favor of the over. This price tag equals 6.3 wins. It’s the normalized win total.
During the off-season, public opinions on NFL teams are widely positive. Every organization is better than the year before, addressed its weaknesses during free agency, and every veteran is in the shape of my life. Beat writers are supposed to be biased and to cover their teams in a positive manner.
If ESPN's beat writers were betting on the Vegas win totals for their teams, 26 would bet the over, five are predicting their teams to exactly hit the Vegas win total, and only one would bet the under (Dolphins beat writer @CameronWolfe) https://t.co/bYu5L0RPVw
— Michael David Smith (@MichaelDavSmith) April 18, 2019
The betting markets tend to judge NFL teams more positively than negatively, too. With 16 games on their schedule, NFL teams can win a collective 256 games per season. The sum of the normalized win totals from Pinnacle last year was 260.7. That’s 4.7 more wins than it’s mathematically possible. Twenty-two win totals shaded towards the over. However, the Over/Under was 13-17-2 last season. If you had bet the Over blindly 32 times for $100 as the base amount, your result would have netted -$820 on the year (-19.8% ROI). In contrast, if you had bet all 32 Unders blindly, your profit would have been +$620 (+17.7% ROI). Distribution-wise, you would expect the Over/Under to be around 50/50 over time. But with the heavy shade towards the over side, it’s not truly a winning proposition to bet overs blindly.
Last season, the difference between actual wins and Pinnacle’s normalized win totals was 2.1 per average. The betting markets predicted 13 win totals to be within 1.5 wins of the actual team record. Ten times they fell outside the range of 2.5 wins. The correlation between Pinnacle win totals and the real team wins was 0.27 (R^2).
Win Totals and Strength of Schedule
With the schedule release, many football fans and media outlets try to predict the future strength of schedule (SOS) based on team wins from the season before. Here’s a friendly reminder: don’t do this! There is no predictive nature. It doesn’t matter how many games a team wins in the season before. Predicting future SOS is hard. Based on our team projections we can calculate a rough estimate, but it cannot be perfect. If Deshaun Watson got hurt in week one in 2018, it would have changed the SOS landscape entirely. The SOS for teams like the Colts or Jaguars would change because the Texans were suddenly much weaker than predicted before the season.
As mentioned before, the correlation between Pinnacle win totals and actual team wins was 0.27 in 2018. However, the relationship between 2017 wins and 2018 wins was just 0.12. The calculated SOS based on 2017 wins correlated 0.02 with the actual 2018 SOS. It’s simply noise. Wins from the season before have no predictive value for the victories the next season.