Bring Back the Air Raid:
The Case for Running Less and Passing
More
Bring Back the Air Raid:
The Case for Running Less and Passing More
The 2024-2025 NFL season was, by many accounts, the year of the running back. Free agent stars Derrick Henry and Saquon Barkley dominated headlines and highlight reels, with Barkley nearly breaking the single-season rushing record while powering the Eagles to a Super Bowl victory. Even at the collegiate level, Boise State’s Ashton Jeanty nearly claimed the Heisman Trophy over Travis Hunter and a field of talented quarterbacks. The football world seemed to be experiencing a renaissance of ground game glory.
This apparent resurgence reinforced one of football’s most sacred tenets: that establishing the run is essential to offensive success. Coaches, analysts, and former players consistently preach the virtues of a “balanced” attack and the necessity of “wearing down” defenses with a physical running game. It’s an axiom so deeply ingrained in football culture that it’s rarely questioned.
But what if this conventional wisdom is fundamentally wrong?
Using modern NFL play-by-play data, this analysis challenges the run-first orthodoxy and reveals a counterintuitive truth: NFL teams are systematically under-utilizing the passing game, leaving significant win probability—and ultimately, victories—on the table.
Win Probability Added (WPA) measures how much each play increases or decreases a team’s chances of winning. When we compare passing and rushing plays from the 2024 season, the contrast is striking: the average pass increases a team’s win probability by 0.22%, while the average run by a running back actually decreases win probability by 0.08%.
| Play Type | Avg. WPA (%) | 95% CI (WPA) | Avg. EPA | 95% CI (EPA) | Number of Plays |
|---|---|---|---|---|---|
| Pass | 0.22% | (0.16%, 0.28%) | 0.0690 | (0.0486, 0.0894) | 22429 |
| RB Run | -0.08% | (-0.13%, -0.03%) | -0.0726 | (-0.0893, -0.0559) | 13303 |
| Note: Analysis includes all passing plays and running plays by running backs only (QB runs excluded). WPA represents Win Probability Added per play (as percentage), while EPA represents Expected Points Added per play. 95% CI represents the 95% confidence interval. |
This pattern extends to Expected Points Added (EPA) as well, where passing plays generate nearly 0.07 points on average, while running plays cost teams about 0.07 points. The data paints a clear picture—passing is substantially more efficient than running.
The player-level statistics tell the same story. Among starting quarterbacks, 24 of 32 (75%) generated positive win probability with their throws. For running backs, the numbers flip dramatically: only 11 of 32 starters (34%) added win probability when carrying the ball. Even the season’s celebrated star rushers barely outperformed middle-tier quarterbacks in terms of value added per play.
| Quarterback | Avg. WPA | 95% CI (WPA) | Avg. EPA | 95% CI (EPA) | Plays |
|---|---|---|---|---|---|
| J.Burrow | 0.72% | (0.36%, 1.08%) | 0.19 | (0.08, 0.3) | 770 |
| J.Allen | 0.67% | (0.37%, 0.97%) | 0.31 | (0.2, 0.42) | 684 |
| P.Mahomes | 0.64% | (0.35%, 0.93%) | 0.14 | (0.04, 0.24) | 815 |
| J.Daniels | 0.61% | (0.21%, 1.01%) | 0.20 | (0.09, 0.31) | 773 |
| L.Jackson | 0.59% | (0.22%, 0.96%) | 0.32 | (0.19, 0.45) | 632 |
| T.Tagovailoa | 0.57% | (0.17%, 0.97%) | 0.22 | (0.08, 0.36) | 450 |
| B.Mayfield | 0.5% | (0.15%, 0.85%) | 0.22 | (0.11, 0.33) | 714 |
| J.Goff | 0.49% | (0.15%, 0.83%) | 0.25 | (0.12, 0.38) | 638 |
| S.Darnold | 0.48% | (0.13%, 0.83%) | 0.08 | (-0.04, 0.2) | 722 |
| J.Hurts | 0.42% | (0.02%, 0.82%) | 0.18 | (0.05, 0.31) | 585 |
| M.Stafford | 0.37% | (0.02%, 0.72%) | 0.11 | (0, 0.22) | 665 |
| D.Prescott | 0.36% | (-0.14%, 0.86%) | -0.02 | (-0.18, 0.14) | 338 |
| G.Smith | 0.34% | (-0.05%, 0.73%) | 0.06 | (-0.06, 0.18) | 702 |
| B.Purdy | 0.34% | (-0.05%, 0.73%) | 0.19 | (0.06, 0.32) | 554 |
| K.Cousins | 0.34% | (-0.18%, 0.86%) | 0.06 | (-0.09, 0.21) | 516 |
| J.Herbert | 0.33% | (0.02%, 0.64%) | 0.11 | (-0.01, 0.23) | 652 |
| J.Love | 0.33% | (-0.05%, 0.71%) | 0.15 | (0.01, 0.29) | 523 |
| B.Young | 0.17% | (-0.26%, 0.6%) | 0.00 | (-0.14, 0.14) | 477 |
| C.Stroud | 0.14% | (-0.19%, 0.47%) | 0.01 | (-0.1, 0.12) | 739 |
| B.Nix | 0.11% | (-0.18%, 0.4%) | 0.04 | (-0.07, 0.15) | 704 |
| D.Maye | 0.09% | (-0.32%, 0.5%) | 0.02 | (-0.13, 0.17) | 453 |
| A.Rodgers | 0.08% | (-0.31%, 0.47%) | 0.04 | (-0.08, 0.16) | 681 |
| K.Murray | 0.08% | (-0.26%, 0.42%) | 0.13 | (0.02, 0.24) | 653 |
| R.Wilson | -0.02% | (-0.44%, 0.4%) | 0.04 | (-0.11, 0.19) | 440 |
| T.Lawrence | -0.05% | (-0.59%, 0.49%) | 0.02 | (-0.14, 0.18) | 329 |
| J.Winston | -0.06% | (-0.63%, 0.51%) | -0.05 | (-0.24, 0.14) | 345 |
| C.Rush | -0.07% | (-0.46%, 0.32%) | -0.08 | (-0.24, 0.08) | 349 |
| D.Jones | -0.09% | (-0.54%, 0.36%) | -0.07 | (-0.21, 0.07) | 410 |
| C.Williams | -0.14% | (-0.49%, 0.21%) | -0.03 | (-0.13, 0.07) | 735 |
| A.Richardson | -0.17% | (-0.71%, 0.37%) | -0.12 | (-0.31, 0.07) | 315 |
| G.Minshew II | -0.22% | (-0.72%, 0.28%) | -0.08 | (-0.25, 0.09) | 367 |
| W.Levis | -0.68% | (-1.26%, -0.1%) | -0.13 | (-0.31, 0.05) | 383 |
| Note: Analysis includes the 32 QBs with the most passing plays in 2024. WPA (Win Probability Added) represents the average change in win probability per play. EPA (Expected Points Added) represents the change in expected points per play. Both metrics indicate the value added by each quarterback on passing plays. Green values indicate positive contributions, red values indicate negative contributions. |
| Running Back | Avg. WPA | 95% CI (WPA) | Avg. EPA | 95% CI (EPA) | Plays |
|---|---|---|---|---|---|
| D.Henry | 0.43% | (0.03%, 0.83%) | 0.11 | (0, 0.22) | 376 |
| J.Gibbs | 0.43% | (0.03%, 0.83%) | 0.13 | (0.01, 0.25) | 269 |
| C.Hubbard | 0.41% | (-0.07%, 0.89%) | 0.04 | (-0.07, 0.15) | 253 |
| S.Barkley | 0.29% | (0%, 0.58%) | 0.07 | (-0.04, 0.18) | 443 |
| J.Cook | 0.2% | (-0.06%, 0.46%) | 0.06 | (-0.05, 0.17) | 269 |
| R.White | 0.14% | (-0.53%, 0.81%) | -0.13 | (-0.29, 0.03) | 149 |
| K.Williams | 0.13% | (-0.21%, 0.47%) | -0.10 | (-0.2, 0) | 359 |
| K.Hunt | 0.12% | (-0.15%, 0.39%) | -0.01 | (-0.1, 0.08) | 235 |
| B.Irving | 0.07% | (-0.33%, 0.47%) | 0.06 | (-0.06, 0.18) | 229 |
| J.Jacobs | 0.06% | (-0.21%, 0.33%) | -0.10 | (-0.21, 0.01) | 335 |
| D.Montgomery | 0.06% | (-0.5%, 0.62%) | -0.06 | (-0.21, 0.09) | 197 |
| J.Mixon | 0.04% | (-0.3%, 0.38%) | -0.08 | (-0.19, 0.03) | 298 |
| B.Robinson | 0% | (-0.28%, 0.28%) | 0.00 | (-0.07, 0.07) | 540 |
| A.Jones | -0.04% | (-0.28%, 0.2%) | -0.08 | (-0.17, 0.01) | 275 |
| R.Dowdle | -0.04% | (-0.34%, 0.26%) | -0.05 | (-0.16, 0.06) | 244 |
| J.Williams | -0.04% | (-0.47%, 0.39%) | -0.11 | (-0.28, 0.06) | 163 |
| D.Achane | -0.09% | (-0.45%, 0.27%) | -0.07 | (-0.2, 0.06) | 211 |
| J.Dobbins | -0.09% | (-0.45%, 0.27%) | -0.01 | (-0.15, 0.13) | 207 |
| J.Mason | -0.09% | (-0.54%, 0.36%) | -0.07 | (-0.22, 0.08) | 160 |
| C.Brown | -0.14% | (-0.46%, 0.18%) | -0.04 | (-0.15, 0.07) | 233 |
| J.Conner | -0.16% | (-0.46%, 0.14%) | -0.03 | (-0.14, 0.08) | 250 |
| T.Bigsby | -0.16% | (-0.6%, 0.28%) | -0.01 | (-0.18, 0.16) | 171 |
| J.Taylor | -0.17% | (-0.48%, 0.14%) | -0.09 | (-0.2, 0.02) | 320 |
| R.Stevenson | -0.22% | (-0.72%, 0.28%) | -0.18 | (-0.32, -0.04) | 215 |
| T.Pollard | -0.24% | (-0.53%, 0.05%) | -0.13 | (-0.24, -0.02) | 276 |
| N.Harris | -0.24% | (-0.51%, 0.03%) | -0.09 | (-0.19, 0.01) | 275 |
| B.Hall | -0.28% | (-0.65%, 0.09%) | -0.11 | (-0.23, 0.01) | 219 |
| A.Kamara | -0.3% | (-0.65%, 0.05%) | -0.08 | (-0.19, 0.03) | 236 |
| K.Walker | -0.33% | (-0.84%, 0.18%) | -0.11 | (-0.24, 0.02) | 159 |
| T.Tracy | -0.4% | (-0.97%, 0.17%) | -0.12 | (-0.27, 0.03) | 195 |
| D.Swift | -0.44% | (-0.75%, -0.13%) | -0.17 | (-0.28, -0.06) | 262 |
| T.Etienne | -0.66% | (-1.03%, -0.29%) | -0.20 | (-0.34, -0.06) | 154 |
| Note: Analysis includes the 32 RBs with the most rushing plays in 2024 (QB rushes excluded). WPA represents the average change in win probability per rushing play. EPA represents the change in expected points per rushing play. Both metrics indicate the value added by each running back on rushing plays. Green values indicate positive contributions, red values indicate negative contributions. |
These raw numbers alone don’t prove teams should pass more. One could reasonably argue that effectiveness stems from strategic balance—that the threat of the run opens up the pass, and vice versa.
If we plot team passing rates against their performance, we observe something curious: teams that pass more frequently tend to perform worse overall. This might seem to vindicate the run-first philosophy, but there’s a critical confounding factor: game situation.
NFL teams dramatically alter their play-calling based on score and time remaining. Teams trailing by 14+ points pass on 81% of plays, while teams ahead by the same margin pass just 43% of the time. Similarly, teams with low win probability (<10%) pass at nearly double the rate of teams with high win probability (>90%).
This situational bias creates a misleading correlation. Teams don’t perform poorly because they pass too much; they pass more because they’re already performing poorly. To untangle this knot, we need a more sophisticated approach.
To isolate the true relationship between passing tendency and performance, we can use Pass Rate Over Expected (PROE)—a metric measuring how much teams pass relative to what would be expected in their specific game situations.
When we reexamine the relationship using PROE, a starkly different picture emerges. Teams that pass more than expected don’t suffer any decline in passing effectiveness. In fact, passing more correlates with better overall offensive efficiency, suggesting that most NFL teams are leaving value on the table by running too often.
But team-specific factors could still be skewing our results. Teams blessed with elite quarterbacks naturally pass more and achieve better results. Could the relationship be driven by these talent disparities rather than play-calling choices?
To address this concern, we employ a fixed-effects regression analysis that essentially compares teams against themselves. Rather than comparing the Chiefs to the Bears, we’re examining how the same team performs when it passes more or less than its own typical tendencies.
The results are unambiguous: even after controlling for team quality, passing more than expected leads to better outcomes. When teams increase their passing rate above their own norm, their win probability and expected points both rise significantly.
These findings hold when examining team performance across the full 2024 season. The data shows that 26 teams generated positive value from passing plays, while only 12 teams achieved positive results from rushing plays. Most striking is the potential gain teams could realize by increasing their pass rates to match the league’s most pass-heavy teams (90th percentile PROE). Our model suggests that teams like Philadelphia, Baltimore, and Green Bay could gain nearly a full win over the course of a season simply by adjusting their passing tendency. Even successful offenses appear to be leaving significant value on the table by running more often than optimal.
| Team | Rush WPA (%) | Pass WPA (%) | Rush EPA | Pass EPA | PROE | Plays to Shift | Wins Gained | Points Gained |
|---|---|---|---|---|---|---|---|---|
| MIA | -0.44 | 0.25 | -0.25 | 0.09 | -0.024 | 128 | 0.88 | 42.8 |
| ATL | -0.09 | 0.37 | -0.01 | 0.08 | -0.075 | 185 | 0.86 | 18.1 |
| BAL | 0.25 | 0.64 | 0.08 | 0.33 | -0.077 | 202 | 0.79 | 50.4 |
| BUF | 0.19 | 0.62 | 0.05 | 0.30 | -0.026 | 146 | 0.62 | 36.1 |
| LV | -0.74 | -0.15 | -0.28 | -0.06 | -0.004 | 103 | 0.61 | 23.2 |
| SF | -0.16 | 0.30 | -0.06 | 0.14 | -0.021 | 118 | 0.55 | 23.6 |
| PIT | -0.24 | 0.08 | -0.12 | 0.06 | -0.051 | 162 | 0.52 | 29.4 |
| WAS | 0.22 | 0.67 | 0.01 | 0.22 | 0.007 | 111 | 0.50 | 23.6 |
| GB | 0.10 | 0.37 | -0.03 | 0.16 | -0.076 | 187 | 0.49 | 35.5 |
| LAC | -0.16 | 0.28 | -0.09 | 0.09 | -0.008 | 107 | 0.48 | 18.4 |
| TB | 0.06 | 0.51 | 0.00 | 0.22 | -0.001 | 106 | 0.47 | 24.0 |
| PHI | 0.25 | 0.44 | 0.05 | 0.19 | -0.080 | 235 | 0.45 | 30.9 |
| LA | 0.03 | 0.35 | -0.09 | 0.11 | -0.027 | 138 | 0.44 | 27.3 |
| SEA | -0.27 | 0.34 | -0.13 | 0.04 | 0.021 | 73 | 0.44 | 12.4 |
| JAX | -0.34 | -0.02 | -0.07 | 0.00 | -0.030 | 122 | 0.39 | 8.6 |
| DAL | -0.16 | 0.15 | -0.13 | -0.07 | -0.017 | 123 | 0.38 | 7.4 |
| NE | -0.30 | 0.00 | -0.19 | -0.02 | -0.023 | 121 | 0.36 | 20.2 |
| MIN | -0.09 | 0.51 | -0.11 | 0.09 | 0.043 | 54 | 0.33 | 11.1 |
| DET | 0.27 | 0.47 | 0.04 | 0.24 | -0.034 | 147 | 0.29 | 29.4 |
| DEN | -0.17 | 0.14 | -0.08 | 0.06 | 0.010 | 90 | 0.28 | 12.4 |
| HOU | -0.08 | 0.15 | -0.11 | 0.01 | 0.008 | 100 | 0.24 | 11.7 |
| KC | -0.03 | 0.62 | -0.07 | 0.13 | 0.062 | 35 | 0.23 | 6.9 |
| NYG | -0.34 | -0.17 | -0.12 | -0.07 | -0.025 | 125 | 0.21 | 5.3 |
| NYJ | -0.16 | 0.07 | -0.11 | 0.05 | 0.029 | 63 | 0.14 | 10.1 |
| NO | -0.09 | 0.00 | -0.04 | -0.07 | -0.047 | 147 | 0.13 | -5.3 |
| IND | -0.09 | -0.02 | -0.07 | -0.03 | -0.075 | 176 | 0.12 | 8.0 |
| CHI | -0.26 | -0.18 | -0.14 | -0.04 | -0.016 | 117 | 0.09 | 12.4 |
| CLE | -0.39 | -0.32 | -0.14 | -0.20 | 0.009 | 94 | 0.07 | -5.2 |
| CIN | -0.15 | 0.75 | -0.08 | 0.19 | 0.089 | 0 | 0.00 | 0.0 |
| ARI | 0.10 | 0.08 | -0.02 | 0.13 | -0.018 | 116 | -0.02 | 17.2 |
| TEN | -0.29 | -0.33 | -0.17 | -0.06 | -0.068 | 171 | -0.07 | 20.3 |
| CAR | 0.21 | 0.11 | -0.04 | -0.04 | -0.014 | 106 | -0.11 | -0.1 |
| Note: WPA shown as percentages (e.g., 0.50% = 0.005 win probability per play). ‘Plays to Shift’ shows how many plays would need to be converted from run to pass to match CIN’s PROE (0.089). Teams with better rushing than passing effectiveness show negative gains. Calculation directly uses the difference in effectiveness between each team’s passing and rushing plays. |
To ensure these findings aren’t merely a product of one unusual season, we extended our analysis across 19 years of NFL data (2006-2024). The evidence is remarkably consistent: in both regular season and playoff contexts, passing more than expected is associated with significantly better outcomes.
| Game Type | Play Type | Metric | Coefficient | 95% CI | t-statistic |
|---|---|---|---|---|---|
| All Games | |||||
| All Games | Total | EPA | 0.1016 | (0.0418, 0.1613) | 3.57 |
| All Games | Rush | EPA | -0.5211 | (-0.6047, -0.4374) | -13.08 |
| All Games | Pass | EPA | 0.2407 | (0.1510, 0.3304) | 5.64 |
| All Games | Total | WPA | 0.0020 | (0.0004, 0.0036) | 2.57 |
| All Games | Rush | WPA | -0.0138 | (-0.0159, -0.0118) | -14.28 |
| All Games | Pass | WPA | 0.0060 | (0.0038, 0.0081) | 5.88 |
| Playoffs | |||||
| Playoffs | Total | EPA | 0.1065 | (-0.3183, 0.5313) | 0.53 |
| Playoffs | Rush | EPA | -0.4915 | (-1.0014, 0.0183) | -2.03 |
| Playoffs | Pass | EPA | 0.2161 | (-0.4425, 0.8747) | 0.69 |
| Playoffs | Total | WPA | -0.0088 | (-0.0198, 0.0021) | -1.70 |
| Playoffs | Rush | WPA | -0.0231 | (-0.0411, -0.0051) | -2.69 |
| Playoffs | Pass | WPA | -0.0054 | (-0.0219, 0.0110) | -0.69 |
| Regular Season | |||||
| Regular Season | Total | EPA | 0.1005 | (0.0353, 0.1658) | 3.24 |
| Regular Season | Rush | EPA | -0.5308 | (-0.6134, -0.4483) | -13.51 |
| Regular Season | Pass | EPA | 0.2455 | (0.1508, 0.3402) | 5.45 |
| Regular Season | Total | WPA | 0.0021 | (0.0003, 0.0039) | 2.41 |
| Regular Season | Rush | WPA | -0.0139 | (-0.0159, -0.0119) | -14.61 |
| Regular Season | Pass | WPA | 0.0063 | (0.0039, 0.0086) | 5.53 |
| Note: Fama-MacBeth analysis calculates the coefficient for each season separately, then averages across seasons. t-statistics are calculated using the standard error of the coefficient estimates across seasons. Playoff results may have higher variability due to smaller sample sizes. | |||||
The magnitude of these effects may appear modest at first glance—an increase of 0.0050 in win probability and 0.21 in expected points for each 0.1 increase in PROE. But in a league where games are often decided by a single possession, these margins translate to meaningful advantages over the course of a season.
The persistence of run-heavy strategies in the face of this evidence represents one of football’s most puzzling inefficiencies. In a hypercompetitive league where coaches work 100-hour weeks seeking every conceivable edge, why do teams continue to emphasize a demonstrably inferior offensive approach?
Cultural inertia and risk aversion likely play significant roles. Running plays produce fewer catastrophic outcomes like interceptions, making them feel safer. The physical nature of running also aligns with football’s traditional ethos of toughness and attrition.
But the data speaks clearly: establishing the run doesn’t open up the passing game—if anything, the opposite is true. Teams would benefit from passing more frequently in almost all game situations, even after accounting for defensive adjustments and team-specific factors.
The 2024 season’s celebration of running back excellence, while entertaining, masks this fundamental strategic truth. The most efficient path to victory in today’s NFL isn’t through a balanced attack or a dominant ground game, but through the air.
Perhaps it’s time to bring back the Air Raid after all.