The Baltimore Ravens beat the Cleveland Browns 41-17 in Cleveland on a Sunday afternoon in mid-September, Lamar Jackson finished with 273 passing yards and four total touchdowns, and Ravens fans across the secondary markets spent the rest of the day arguing on social media that the team was a Super Bowl favorite. The same afternoon, the San Francisco 49ers moved to 2-0 against the Saints despite losing both George Kittle and Brock Purdy to injury and getting one of the worst statistical performances by a starting quarterback in Mac Jones’s career. The same afternoon, Dak Prescott threw for 361 yards in a Cowboys loss, Bryce Young completed 34 of 55 passes for the Panthers in a one-score game, and Drake Maye produced a Patriots win that nobody had picked. Week 2 was a 7-3 weekend for road favorites against the spread, the kind of week that produces confident takes about which teams are “real” and which teams are “exposed.” None of those takes will look the same by Week 6.
The analytical question after Week 2 is always the same and almost always answered wrong by the public coverage: which 2-0 teams will end up being real, and which 0-2 teams will end up being mirages of variance. The honest answer is that the sample size is small enough that neither classification can be made confidently. The historical data shows that Week 2 records correlate with rest-of-season records at a coefficient of approximately 0.21 — meaningful but weak. A 2-0 team becomes a playoff team about 58% of the time. A 0-2 team becomes a playoff team about 28% of the time. The numbers move modestly with each additional week of evidence. Through two weeks, the standings are noise.
What follows is what Week 2 actually told us at the structural level, where the trench-level numbers point in different directions than the W-L records, and which 2-0 teams the underlying numbers actually support.
The 2-0 teams and what the underlying numbers say
Six teams entered Week 3 at 2-0: Ravens, 49ers, Cowboys (1-1 but with strong underlying numbers), Patriots, Eagles, and Texans. Of those, the underlying metrics — pass-block win rate, pressure rate, EPA per play, success rate — point at very different stories.
The Ravens at 2-0 are real. Their pass-block win rate is 72% (top-three), Lamar Jackson’s EPA per play sits at +0.31 (third-best in the league), and the defensive EPA-allowed is -0.08 (top-eight). The structural profile matches a 12-13 win baseline if health holds. The two wins are the cleanest version of “real” the 2-0 group produces.
The Patriots at 2-0 with Drake Maye are real and surprising. Pass-block win rate of 63% (top-twelve), Maye’s EPA per play of +0.18 (top-ten), defensive numbers improved from 2024 by a meaningful margin. The Patriots were preseason picked for 6-7 wins. The Week 2 underlying data points at a 9-10 win team, with Maye’s individual development being the swing variable.
The 49ers at 2-0 are structurally compromised. Their pass-block win rate is 51% (bottom-twelve), Mac Jones produced -0.04 EPA per play against the Saints, and the injuries to Kittle and Purdy will affect the next month. The 2-0 record is real but the underlying performance is closer to a 7-8 win pace. The next four games will compress the gap.
The Eagles at 2-0 are also real but for a different reason than the Ravens. Their pass-block win rate is 68% (top-six), the defense has produced a 51% pressure rate without blitzing in two games (echoing the Super Bowl LIX template I covered in February). The structural profile is the same as the championship roster. The record is sustainable.
The 0-2 teams and what the underlying numbers say
| 0-2 team | Pass-block win rate | Defensive EPA | Real or mirage? |
|---|---|---|---|
| Browns | 48% (bottom) | +0.09 (bad) | Real 0-2; structural problems |
| Saints | 56% (mid) | +0.04 (mediocre) | Real 0-2; should be 1-1 |
| Chiefs | 61% (top-15) | +0.02 (close to avg) | Mirage 0-2; should be 1-1 |
| Giants | 52% (bottom-15) | +0.07 (poor) | Real 0-2; expected |
| Bengals | 67% (top-eight) | +0.10 (poor) | Mirage 0-2; defense will improve |
The Chiefs and Bengals at 0-2 are the structural mirages. Kansas City’s pass-block numbers and Mahomes’s expected EPA both point at a team that is structurally a 10-12 win club but has played two unfortunate games against quality opponents. The Bengals’ offensive line numbers are top-eight; their defensive EPA-allowed is bad in a way that the underlying personnel suggests will correct. Both teams will be 5-3 or 6-2 by Week 8, with very high probability. Treating either as “exposed” is the kind of two-week-sample mistake the analytical community has been documenting for fifteen years.
Where this gets weird
The clean “Ravens-Patriots-Eagles are real, Chiefs-Bengals will rebound” reading misses three things that complicate the projections.
The first is that the Drake Maye case is genuinely uncertain. Sophomore quarterbacks who post elite Week 1-2 numbers have, historically, regressed at a meaningful rate. Trevor Lawrence in 2023 was a clean example — strong start, mid-season collapse to ordinary play. Bryce Young in 2024 the inverse. Whether Maye looks like an elite Year 2 quarterback by Week 8 depends on factors the Week 2 sample cannot reveal: offensive line consistency, defensive coordinator adjustments by future opponents, his own emotional resilience under pressure. The structural ceiling is real. The structural ceiling does not guarantee the season’s path.
The second is that the 49ers’ injury situation could collapse the underlying numbers further. If Purdy misses 6+ weeks and Kittle’s recovery extends past Week 5, the structural profile shifts from “compromised 2-0 team” to “stretch-run wild-card bet.” The 2-0 record will be irrelevant by Week 12 if the injuries compound. The Week 2 data is the cleanest sample of the healthy team; subsequent samples will be different.
The third is the Chiefs specifically deserve a longer reading than the 0-2 record suggests. Their losses were to teams with playoff-level talent. Mahomes’s underlying numbers are within his career range. The defense is the same defense that produced the Super Bowl LIX pressure rate. The structural profile of a healthy Chiefs team in 2025 is closer to 11-12 wins than 8-9. The market line on Kansas City’s season win total has moved up slightly since Week 2, not down. The market is reading the underlying numbers correctly.
What to track in Weeks 3-6
- Mac Jones’s EPA per play against starting defenses. Week 1-2 sample is against weaker units. If he posts +0.10 or above against top-15 defenses, the 49ers’ 2-0 is sustainable. Below 0, the structural compromise asserts itself.
- The Bengals’ defensive EPA-allowed. If it drops from +0.10 toward 0.00 by Week 6, the 0-2 mirage corrects. If it stays at +0.10, the structural problem is real.
- Drake Maye’s pass-block-adjusted EPA. The cleanest projection signal for sophomore quarterbacks. If Maye produces +0.10 with pass protection at 60%+, the Patriots are real. If he produces +0.10 only when protection holds at 70%+, the ceiling depends on the line.
- The Ravens-Lions game in Week 4. Two top-three teams meeting in a regular-season test of who has the better Super Bowl chance. The outcome will not decide anything but the underlying numbers from the matchup will reveal more than any other early-season game.
The callback
That Sunday afternoon in mid-September when Lamar Jackson dropped 41 on the Browns and the Eagles defense pressed Mahomes into a difficult loss and the 49ers won despite injuries and Drake Maye’s Patriots made the most-unlikely 2-0 of the league — that afternoon produced exactly the kind of confident weekend that NFL analytics writing has been getting wrong for fifteen years. The records of all sixteen teams will move meaningfully between Week 2 and Week 8. The team that looks “exposed” today will be a playoff lock by Halloween in roughly a third of cases. The team that looks “real” today will be a 4-4 disappointment in roughly a quarter of cases. The structural numbers — pass-block, pressure, EPA — are the only data from Week 2 that survives into October with any predictive force. The records do not. The W-L column is noise. The trench-level metrics are signal. The next six weeks will reveal which 2-0 teams earned the standing and which 0-2 teams were paying the variance tax of a tough early schedule. The small samples piece covers the broader version of this. Week 2 happened. The takes will follow. The math is unchanged.
Pass-block and pressure data via ESPN Stats and Information; EPA per play via Pro Football Reference; advanced metrics context via PFF.
