The 2025 NFL season opened at 8:21 PM EST on Thursday, September 4 at Lincoln Financial Field, sixty-four seconds after Jalen Carter walked off the field in a defensive ejection that had nothing to do with the football play that had just been called. Carter had appeared to spit in Dak Prescott’s direction following the opening kickoff. The officials handled the discipline at the line of scrimmage before Philadelphia had run a single snap. The Eagles played the entire game without their starting interior defensive lineman, in conditions that had already produced a 73-minute weather delay, and they beat the Cowboys 24-20 with Jalen Hurts producing the kind of dual-threat line that the analytical community had been pricing into the team’s 2025 ceiling since the spring.
Hurts finished 19 of 23 for 152 yards, plus two rushing touchdowns and 62 yards on the ground. He threw two passing touchdowns. Saquon Barkley added a rushing score. The Cowboys’ Javonte Williams added two rushing touchdowns of his own in a first half that ended at halftime with the score tied at 20. Philadelphia’s defense pitched a shutout in the second half. The game was decided by a Brandon Aubrey 53-yard field goal that closed the first half, the Eagles’ second-half defensive adjustments, and Hurts’s three-and-out conversion rate in the third quarter being above the league average for top quarterbacks against quality defenses.
What follows is what made the Eagles 24-20 win analytically defensible despite the Carter ejection and the weather, where the early-season numbers from the opener actually predict and where they do not, and how to read Week 1 without the cycle of overconfident projection that produces NFL Twitter’s most-deletable content of the next two months.
What the Eagles actually did in the second half
The first half was a back-and-forth offensive game between two top-eight offenses operating at roughly their season-long capabilities. The Cowboys produced four scoring drives, the Eagles produced three, and the halftime tie at 20 reflected the underlying matchup quality. The second half was the structural story. Philadelphia held Dallas to zero points across six possessions. The Cowboys produced three punts, two turnovers on downs, and a missed field goal. Their EPA across the second half was -8.4 — bottom-five in any team-half from the entire 2025 season opener weekend.
The defensive adjustments that produced the shutout were not exotic. Vic Fangio’s halftime change was to drop the defensive line into a slightly wider alignment that took away the Williams cutback runs that had produced both Cowboys rushing scores in the first half. The pressure on Prescott went from 22% in the first half to 41% in the second. Without Carter on the field — and against an offensive line that had been getting beaten one-on-one — Philadelphia generated pressure with linebacker blitzes that the Cowboys’ offensive coordinator failed to anticipate. The 41% pressure rate is the number that decided the game.
Hurts’s line — 152 passing yards on 19 of 23 — looks pedestrian until the context is added. He had no interceptions, took zero sacks, and added 62 yards on the ground with two scores. His combined EPA for the game was +0.31 per play, which would have ranked top-three among quarterbacks in Week 1. The pedestrian passing line was a function of game plan (run-heavy approach with Barkley) and weather (the 73-minute delay produced wet field conditions that limited the deep passing game). The advanced numbers had Hurts performing at MVP level despite the box score that suggested otherwise.
What Week 1 NFL data actually predicts
| Week 1 metric | Correlation with rest-of-season |
|---|---|
| QB EPA per play | 0.31 (weak) |
| QB pressure rate allowed | 0.42 (moderate) |
| Defensive EPA-allowed | 0.28 (weak) |
| Pass-block win rate | 0.51 (moderate-strong) |
| Offensive penalty rate | 0.18 (very weak) |
The pattern is consistent across years. The trench-level metrics — pass-block win rate, pressure rate — correlate with rest-of-season performance at meaningful levels. The outcome metrics — EPA per play, points-scored differential — correlate weakly. The Week 1 cycle that obsesses over scoring outputs is reading the wrong category of data. The cycle that ignores trench-level performance is missing the actual predictive information.
Applied to the Eagles-Cowboys opener: Philadelphia’s pass-block win rate of 67% in Week 1 is a moderate-strong signal that their offensive line is going to be a top-ten unit for the season. Dallas’s pass-block win rate of 51% is a moderate-strong signal that their offensive line is going to be bottom-twelve. Those two numbers will compound across the season much more meaningfully than the 24-20 final score will.
Where this gets weird
The clean “Eagles won, here’s what it means” reading misses three things that the typical Week 1 cycle gets wrong every September.
The first is the weather impact. The 73-minute delay and wet field conditions affected both teams’ deep passing games. Hurts’s 152-yard passing total is partly a function of the conditions; the same game in dry weather probably produces 220-240 passing yards. The Cowboys’ first-half 20 points came mostly on the ground for a similar reason. The variance the weather added is real and is not in the predictive correlations above. Any Week 1 number from a weather-affected game should be discounted by 15-20% in either direction.
The second is the Carter ejection. Carter is Philadelphia’s most-disruptive interior defender. The Eagles played 60 minutes without him and still produced a shutout in the second half. That is a meaningful tell about the defensive depth — the unit can function without its top piece. But the predictive read for the rest of the season requires assuming Carter will be on the field, which adds defensive upside that the Week 1 game did not measure. Both teams are probably stronger than their Week 1 performance against each other suggested.
The third is the early-season narrative cycle that the next two weeks will produce. The Eagles will be the favorites for the NFC after a Week 1 win against a divisional opponent. The Cowboys will face a quarterback-blame cycle that ignores the structural offensive line issues. By Week 8, the narratives will have updated three times. The data that drives the actual season-long story will be the offensive line and defensive pressure numbers, and almost none of the next-two-weeks coverage will be about those numbers.
How to actually read NFL Week 1 results
- Look at pass-block win rate before EPA. The trench-level metric correlates with rest-of-season performance at almost twice the rate of the outcome metric.
- Discount weather-affected games by 15-20%. The Eagles-Cowboys numbers should be read with this adjustment.
- Suspend judgment on coaching narratives until Week 4. The first-month sample is too small to validate or reject any preseason coaching question. The cycle that produces hot takes in Week 1 will be apologizing for them by October.
- Watch the next-week game more carefully than the Week 1 game. Week 2 against a different opponent is the cleanest test of whether the Week 1 numbers reflect team quality or matchup-specific variance.
The callback
Those 64 seconds at the start of the 2025 NFL season opener, when Carter walked off the field for spitting at Prescott and Philadelphia’s defensive depth chart got tested before a single offensive snap, set the tone for a game that the underlying numbers will eventually treat as one of the most analytically defensible Week 1 results of the year. The Eagles won because their pass-block win rate held, their second-half defensive adjustments produced a shutout, and Hurts’s combined EPA was top-three for the week despite a passing line that the box-score readers ignored. The Cowboys lost because their offensive line was structurally compromised against pressure and their second-half adjustments did not arrive. Both teams are probably better than the 24-20 score suggested. Neither team is exactly what the next-two-weeks narrative cycle is about to claim they are. The small samples piece covers why the Week 1 cycle keeps producing wrong predictions. The opener happened. The takes followed. By Week 8, half the takes will be wrong. The trench-level numbers from Thursday night will still be the most predictive thing the opener produced.
Game data via NFL.com; pass-block win rate via ESPN Stats and Information; EPA per play via Pro Football Reference.
