NFL market analysis: Why are unders hitting at a historically high rate?
If you enjoy betting the over on NFL games, the 2023 season (so far) isn't for you.
Unders are hitting at a historically high rate through Week 6, with 61.3% of games failing to surpass oddsmakers' point total, per data from TeamRankings.
That marks the highest win rate for unders through six weeks since 1991, according to The Action Network's Evan Abrams.
Part of the reason for this surprisingly high rate is the fact scoring is down across the NFL.
An average of 43.4 points have been scored per game through Week 6, slightly down from last year's 43.8, continuing the downward trend since games averaged a whopping 49.6 points in 2020.
Thirteen teams are averaging fewer than 20 points a game. Remove the Dolphins (37.2 points) and the 49ers (30.7), and the remaining teams are averaging only 20.9 points, which would be the lowest since 2007.
Last eight seasons
| Year | Points/Game | Team Yards/Game |
|---|---|---|
| 2023 | 43.4 | 329.7 |
| 2022 | 43.8 | 340.1 |
| 2021 | 46.0 | 343.6 |
| 2020 | 49.6 | 359.0 |
| 2019 | 45.6 | 347.9 |
| 2018 | 46.6 | 352.2 |
| 2017 | 43.4 | 334.1 |
| 2016 | 45.6 | 350.4 |
(Data source: Pro Football Reference)
As seen in the data above, teams' offensive yards per game are also way down. A 329.7 yards-per-game average is the lowest since 2008.
A decrease in passing yards is the key reason for the overall dip in offense, with teams averaging only 218.6 yards through the air compared to 235.7 from 2012-2021.
Teams struggled to throw in 2022 as well, averaging 218.5 yards. However, they made up for it on the ground by rushing for 121.6 yards per game, the most since 1987.
But aren't over/unders supposed to factor in lower scores and lower offensive outputs? Yes and no.
The over/under set by a sportsbook's algorithm is not intended to predict the total points scored in a game. While they will often come close, the main objective of opening lines is to reduce one-sided action from sharp bettors.
If the total for a theoretical Chiefs-Bills game opens at 48.5 and sharp bettors believe the over is more likely to hit than the odds indicate, sportsbooks will adjust the total accordingly until one-sided action subsides.
But the fact that this theoretical line moved from 48.5 to 50.5 does not mean the sportsbook's algorithm was necessarily wrong, nor does it change the likelihood of a specific total occurring.
Rather, line movement only indicates the sharp-betting market disagreed with the sportsbook's opening line. In short: data and models generate an opening line, while sharp bettors' opinions (ignoring potential key injuries or weather) form the closing line.
With that in mind, let's look at the average closing line game total in the past seven years, and the average since 2003, to see if there are any answers as to why unders are hitting at 61% this season.
| Year | Avg. Total | Over % | Under % |
|---|---|---|---|
| 2023 | 44.1 | 38.7 | 61.3 |
| 2022 | 44.2 | 40.8 | 57.6 |
| 2021 | 46.5 | 44.6 | 55.1 |
| 2020 | 48.2 | 50.0 | 48.0 |
| 2019 | 45.2 | 52.1 | 47.6 |
| 2018 | 46.6 | 50.6 | 49.0 |
| 2017 | 44.1 | 41.4 | 57.7 |
| Since 2003 | 44.0 | 50.3 | 48.5 |
(Data source: TeamRankings)
Scoring in 2020 was historically high. It was also a bizarre year due to the pandemic, which forced many games to be played without fans and made it easier for offenses to operate.
That said, the betting market adjusted to the higher scores with a 48.2 average game total, which produced a relatively even split between overs and unders.
The three years that stand out from the chart above are 2023, 2022, and 2017. Each of these years has a scoring average of slightly over 43 points and an average game total of roughly 44. This combination resulted in under bets winning at a much higher rate than the expected 50-50 split.
The last time period when the points-per-game average was similar to the three above years was 2003-2013.
Here is the over/under data from that 11-year stretch.
| Year | Avg. Total | Pts/Gm | Over % | Under % |
|---|---|---|---|---|
| 2013 | 45.5 | 46.8 | 51.6 | 46.4 |
| 2012 | 44.8 | 45.6 | 48.3 | 50.7 |
| 2011 | 43.7 | 44.4 | 50.0 | 48.7 |
| 2010 | 42.8 | 44 | 55.8 | 42.6 |
| 2009 | 42.8 | 43 | 49.6 | 49.3 |
| 2008 | 42.3 | 44 | 47.1 | 40.3 |
| 2007 | 41.6 | 43.4 | 46.0 | 48.6 |
| 2006 | 40.5 | 41.4 | 40.2 | 57.3 |
| 2005 | 41.1 | 41.2 | 39.7 | 57.8 |
| 2004 | 41.3 | 43 | 51.3 | 48.3 |
| 2003 | 41.3 | 41.6 | 46.6 | 52.4 |
The points-per-game average from 2003-2013 was 43.5 and the average game total was 42.5. Overs hit 47.8% of the time and unders occurred at a 49.3% rate. (Pushes account for the remaining percentage.)
Let's compare that to the games played in 2017, 2022, and 2023 when overs hit 40.3% of the time and unders a staggering 58.9%. During these three years, the points-per-game average was 43.5 (the same as the 11-year stretch above) but the average game total was 44.1, which is 1.6 points higher than the average game total from 2003-2013.
Is that point differential the reason why unders are hitting nearly 60% of the time since 2022? It's too difficult to say that's the only reason, but it's definitely a significant factor.
Perhaps this data reveals a market inefficiency where total points scored begin to trend down and the betting market is slow to catch up.
As noted, the betting market was able to adjust quickly as more points were scored in 2020, as well as the stretch from 2010-2013 when points per game increased by nearly three while the over/under splits remained close to 50-50.
If NFL teams continue to struggle to put points on the board and the sharp betting market fails to adjust by lowering average game totals, then those who bet the over in hopes of exciting back-and-forth action will usually be handing over money to sportsbooks.