Select Page

Vegas Over/Under betting lines are a key tool for the DFS player. They are remarkably accurate – for the 2012-2015 seasons, the average error was ~0.5 points per game. Given the pervasiveness of the statistic, we analyzed the predictions to find deeper meaning.

I’ll start by introducing a few terms that will be used throughout this analysis.  “Vegas over/under” is the amount of points that the biggest sports bookmakers predicted would be scored by both teams combined.  Actual total points is the number of points the two teams scored in reality.  “Vegas implied points” is the number of points bookmakers predicted would be scored by each individual team.  This number is derived from the over/under line and the spread.  Actual points refers to the number of points scored by each individual team.

The analysis of betting lines and actual points scored covers the NFL regular season from 2012-2015.

Below is a chart showing the actual number of points scored with respect to the over/under line. Using linear regression, we found that for each point that was added to the over/under, an extra 0.84 points were actually scored. In a perfect world, the slope of this line would be exactly 1. The fact that it is lower represents the fact that it is easier for games to outperform very low prediction and to underperform very high prediction.

A similar analysis of implied points with regard to actual points scored yields a similar result. Using linear regression, we found that for each additional implied point, an extra ~0.67 points were actually scored. For reasons similar to above, this number is also below 1. However, this metric for a single team is significantly lower than when looking at both teams in a game in aggregate. This too is explainable. Teams that are predicted to score a high number of points are generally large favorites. When they get ahead, they are incentivized to slow their pace of play, run the ball more, and therefore tend not to score as many points. Conversely, teams with low point predictions are typically large underdogs and tend to fall behind early. These teams are then incentivized to throw the ball more to score more points. This explains why the metric is less accurate for individual teams rather than games as a whole. Games with high over/under lines tend to occur when two offensive teams match up, where both teams are incentivized to score to keep up with each other. This leads to a higher slope for over/unders than implied points.

While it is good to know the relationship between Vegas lines and actual points scored, it is perhaps even more useful to know how well Vegas lines indicate how often a game will score a given number of points. Below are a table and chart illustrating this relationship. (Minimum 40 games)

 Points Scored O/U in (36, 40] O/U (40, 44] O/U in (44, 48] O/U in (48, 52] O/U in (52, 56] 0 1 1 1 1 1 4 1 1 1 1 1 8 1 1 1 1 1 12 1 1 1 1 1 16 1 0.994029851 0.989218329 1 1 20 0.966666667 0.964179104 0.973045822 1 0.978723404 24 0.966666667 0.91641791 0.959568733 0.993670886 0.957446809 28 0.883333333 0.853731343 0.911051213 0.955696203 0.957446809 32 0.833333333 0.776119403 0.854447439 0.905063291 0.914893617 36 0.683333333 0.659701493 0.749326146 0.85443038 0.893617021 40 0.533333333 0.534328358 0.654986523 0.765822785 0.829787234 44 0.45 0.405970149 0.51212938 0.64556962 0.744680851 48 0.333333333 0.31641791 0.404312668 0.550632911 0.595744681 52 0.2 0.217910448 0.293800539 0.424050633 0.531914894 56 0.133333333 0.167164179 0.210242588 0.335443038 0.382978723 60 0.066666667 0.12238806 0.148247978 0.278481013 0.319148936 64 0 0.089552239 0.094339623 0.183544304 0.234042553 68 0 0.059701493 0.053908356 0.113924051 0.127659574

As you can see from the chart, games with a higher over/under prediction have higher probabilities of achieving given scores, as expected. This is useful in determining which games are likely to go off. For example, a game with a predicted score in the range of (52, 56] is nearly five times as likely to have a game reach 60 points scored than a game that was predicted in the range of (36, 40]. However, it is also noteworthy that lower predicted scores are more likely to actually achieve their prediction.

Below is a similar table, this time using implied points.

 Points Scored (12, 16] Implied Points (16, 20] Implied Points (20, 24] Implied Points (24, 28] Implied Points (28, 32] Implied Points 0 95.45% 99.04% 99.05% 99.61% 100.00% 4 93.18% 96.39% 97.99% 97.48% 99.21% 8 77.27% 90.38% 93.73% 93.22% 97.64% 12 72.73% 80.05% 87.46% 89.15% 93.70% 16 47.73% 65.87% 75.03% 79.26% 83.46% 20 34.09% 44.95% 56.21% 64.92% 69.29% 24 18.18% 30.77% 37.63% 46.90% 56.69% 28 15.91% 17.31% 25.21% 33.53% 44.88% 32 11.36% 11.54% 15.74% 22.09% 30.71% 36 6.82% 6.73% 8.88% 12.79% 22.83% 40 2.27% 2.88% 3.91% 7.56% 17.32% 44 0.00% 1.20% 1.42% 4.07% 10.24%

The results for single teams are even more distinct than for game totals. Higher implied totals are much more likely to achieve a higher score total.

Vegas Data and Fantasy Points

Now you have a better understanding of how Vegas can help you predict the scoring of an NFL game, but we have to take that one step further if we are to use this knowledge effectively in DFS. A good portion of the DFS market knows to use Vegas lines, but do they use them enough or too little? Today we'll give you some concrete data to answer that question for yourself.

We'll be analyzing the relationship between both the implied and actual points of an NFL game, to the sum of offensive fantasy points scored by the teams in those games.  In our sample, each team plays an equal amount of home and away games.  Since teams do better at home, the implied point total on average and the average fantasy points per game is higher for home games than for away games.

I've defined the sum of offensive fantasy points as the sum of points across all players on a given team with these positions: QB, RB, WR, and TE.

Home Team

The home team scores an average of 94.61 total offensive fantasy points per game. The relationship between implied points and total offensive fantasy points is near linear. As such, the lower end of the implied points range will see an average 88.14 points, and each implied point after that is worth roughly 2.15 to the total of offensive fantasy points. It's also worth mentioning that games at the 32-35 range of the implied total range average 113.96 points. That's a 29.3% increase from games with implied totals of 20-23, and a 20.5% increase from the average.

Away Team

The away team scores an average of 89.37 total offensive fantasy points per game. This relationship is also very linear in nature, with each implied point adding about 3.35 total offensive fantasy points. The difference between the extremes in the implied totals for away teams is 29.45 points. You can expect a 41.2% increase in offensive fantasy points between teams projected to score 14.25-17.25 and teams projected to score 23.25-26.25

All things considered, it's obvious that home teams score more fantasy points than away teams, but it’s also true that implied points add more to the offensive fantasy total to away teams than they do to home teams. An implied point adds 1.2 more fantasy points to an away team than a home team. This would suggest that a DFS player should weight implied points heavier when they’re considering away teams, and weight them a little lighter when they’re considering home teams.