The League Still Punts When the Math Says Go

The through-line for August 2026 is a familiar one: NFL decision-makers are leaving value in plain sight, on fourth down and on the cap sheet, and the teams that hoover it up are the ones worth watching. We graded 32,731 decisions from the 2025 season. The picture that comes back is of a league that still trusts its gut over the arithmetic — and pays for it in wins.

On the contract side, the August ledger carries 80 verdicts: 16 fair prices, 10 overpays, and 54 entries we're watching because the public data can't settle them yet. The overpay problem is concentrated, as it almost always is, at quarterback — where the gap between what teams are paying and what they're getting is widest and most expensive. The bargains, meanwhile, are hiding in plain sight on rookie deals and the occasional veteran who quietly outran his market.

One number anchors the whole month: the Cowboys are carrying $249.9 million in future prorated cap charges — money already spent, now just waiting to land. That's not a crisis today. It is a constraint that will shape every decision that front office makes for the next several years. The bill always arrives; the only question is whether you planned for it.

In 2025, NFL teams went for it on fourth down 18.9% of the time in competitive situations. The model — built on play-by-play since 1999, grading each decision at the moment it's made, blind to what happened next — wanted them to go 36.1% of the time. That 17-point gap is the go-gap, and it represents real wins left on the field.

"Expected wins lost" is just what it sounds like: if you take every fourth-down decision a coach made and compare the win probability of the choice he made against the win probability of the optimal choice, the difference accumulates into fractions of a win per game. Brian Daboll lost the most — 0.551 expected wins on the season, or 3.24 points of win probability per game across 136 graded decisions. He went for it on 24.3% of fourth downs when the model wanted 42.7%, and his overall decision agreement rate was 70.9%. DeMeco Ryans was nearly as costly at 0.525 expected wins lost, and his go rate was the most conservative of the costly group: 9.8% against a model recommendation of 35.2%. Kevin Stefanski rounds out the bottom three at 0.433 expected wins lost across 151 decisions — the most in the group — with a 71.8% agreement rate.

The best coaches in 2025 weren't dramatically bolder; they were more precisely calibrated. Nick Sirianni led the group in agreement rate at 88.8% across 144 decisions, losing just 0.129 expected wins on the season — 0.76 points of win probability per game. Dan Campbell went for it more than anyone in the top tier (26.3% actual against a 36.4% model recommendation) and still posted a 84.8% agreement rate. Mike McDaniel, with 111 decisions graded, had the lowest expected wins lost of the three at 0.111. The gap between Sirianni's 0.76 points per game lost and Daboll's 3.24 is the distance between a coach who's reading the situation and one who's following a feel.

The single costliest call of the season came in Week 6: Jonathan Gannon, down 4 in the fourth quarter, fourth-and-7 at the opponent's 9-yard line, chose to go for it instead of kicking the field goal. The go gave Arizona a 20.7% win probability; the field goal was worth 33.3%. The decision cost 12.6 points of win probability in one snap. The model doesn't know whether the conversion attempt succeeded — it doesn't need to. At fourth-and-7 from the 9, down 4, a field goal cuts the deficit to one score and keeps the game alive. That's the arithmetic, and it was available before the play was called.

The ledger this month carries 80 contracts across all positions. Every verdict comes with a falsifier — a stated condition that would prove the verdict wrong — and a re-grade date. That discipline is the point. A verdict without a falsifier is just an opinion; we're trying to do something more useful than that.

Patrick Mahomes: $64 million a year, the 98th percentile of the quarterback market, against 92nd-percentile production over the graded window. The price and the output live on the same block, so the verdict is FAIR PRICE. What makes it wrong? Simple: top-5 production flips it to a bargain; production that falls out of the top 30 flips it to an overpay. We re-grade in March 2027. The exit math is worth noting — walking away after 2029 costs Kansas City $31.3 million in dead money, so this deal has real teeth if the production curve bends earlier than expected.

Daniel Jones lands differently. Indianapolis is paying $44 million a year — 94th percentile of the quarterback market, 56% guaranteed — for 69th-percentile production over the past two seasons. The verdict is OVERPAY, and the falsifier is narrow: a top-5 finish in quarterback production in 2026 would make this price fair and this verdict a miss. That's the bet Indianapolis is making — that a career year is coming. The problem is that the contract is priced like the career year is already here, not on the way. If Jones doesn't finish top-5, the Colts owe $35.2 million in dead money to exit after 2027. That's a lot of rent on a gamble.

Jaxon Smith-Njigba is the cleanest entry in the ledger. Seattle is paying $42.15 million a year — the 100th percentile of the wide receiver market — for 99th-percentile production. At 25 years old, with an aging curve that still has room, this is the rare case where market-setting price meets market-setting output. FAIR PRICE. The exit after 2028 carries $21 million in dead money, so the Seahawks are committed, but the production record gives them reason to be.

The 54 WATCH verdicts — the largest single category — cover edge defenders and others where public data grades the price but can't grade the film. Will Anderson Jr. at $50 million a year is the top of that book: 100th percentile of the edge defender market, 67% guaranteed, age 25. We can tell you the price is at the very top of the market. We can't tell you whether the production justifies it from play-by-play data alone. So we say so, and we watch.

A rookie contract is a below-market lease with a known expiration date — the team locked in the price before the player proved what he was worth, and for a few years, both sides live with the terms. The teams that draft well and develop well are the ones collecting the difference between what they pay and what they're getting. This month's windfall list makes that concrete.

Harold Fannin Jr. is the headline. The Browns are paying $1.53 million against a cap — 0.5% of the total, the 16th percentile of tight end pay — for 88th-percentile tight end production. The surplus between what Cleveland is paying and what that production would cost on the open market is 72 points, the widest gap on the board. Drake Maye shows the same dynamic at quarterback: $9.99 million a year (2.8% of the cap, 55th percentile of QB pay) for 96th-percentile quarterback production, a 41-point surplus. New England is getting near-top-of-market quarterback play at a price that doesn't yet reflect it. That window closes when the rookie deal expires; the question is what the Patriots build around it before then.

The veteran bargain list is shorter and quieter, which is usually how bargains work. Devon Achane is at $3.15 million (1% of the cap, 49th percentile of running back pay) for 93rd-percentile running back production — a 44-point surplus on a veteran deal, which is genuinely unusual. The market hasn't caught up to him yet, and Miami is the beneficiary. Travis Kelce at $4.9 million (1.6% of the cap) for 95th-percentile tight end production is a 40-point surplus, though at 39 years old the aging curve is a real variable that the production percentile doesn't capture on its own. Kyle Pitts rounds out the group: $4.62 million for 86th-percentile tight end production, 40-point surplus, still just working through a market that underpriced him coming off an injury history.

The through-line across all four: the discount aisle is almost always about either a rookie scale that hasn't caught up to performance, or a veteran whose market price got depressed by something — injury, position scarcity, a down year — that the current production has moved past. Find those gaps before the market does, and you're building cap room without spending it.

A restructured contract is a balance-transfer credit card: you lower the payment today by spreading the balance over future months, and the balance doesn't shrink — it just moves. Future proration is that moved balance, sitting on next year's statement. The Cowboys are carrying $249.9 million of it, with Dak Prescott's deal accounting for 14.2% of the cap annually. The Texans are next at $238.1 million — though Houston's quarterback cap hit is only 6.6%, which means that proration is spread across the roster rather than concentrated at one position. The Bills ($204 million, 15.7% QB cap share) and the 49ers ($201.8 million, 8.5% QB share) complete the top four. These aren't emergencies — they're constraints. Every free agent signing, every trade, every extension those front offices negotiate happens in the shadow of those numbers.

The albatross list is where the bill has already arrived and the math has turned ugly. Deshaun Watson: $40.96 million against the cap — 12.1% of the total, the 90th percentile of quarterback pay — for 8th-percentile quarterback production. That's an 82-point negative surplus, the worst on the board. Cleveland is paying for a quarterback who isn't there. Brandon Aiyuk lands at the same negative-82 surplus from a different direction: $13.67 million (4% of the cap, 85th percentile of wide receiver pay) for 4th-percentile wide receiver production. Whatever happened between the contract and the field, the output hasn't followed the price.

Evan Engram ($14.14 million, 98th percentile of tight end pay, 23rd-percentile production, -74 surplus) and T.J. Hockenson ($15.61 million, 100th percentile of tight end pay, 28th-percentile production, -72 surplus) complete the group. Both are being paid as if they're at the top of their position; both are producing well below it. The position market set the price; the player didn't sustain the output that justified it. That gap is what an albatross actually is — not a bad person, not a bad player necessarily, just a contract where the price and the production ended up in different zip codes.