The League Still Leaves Wins on the Field

The 2025 season confirmed one thing above all others: the two biggest levers a franchise controls — fourth-down decisions and contract pricing — are still being pulled the wrong way by most of the league, most of the time.

On the field, coaches went for it on 4th down at roughly half the rate the math preferred. Off the field, the league's contract database shows ten outright overpays among the 80 deals graded this cycle, zero classified as genuine good value, and 54 flagged for ongoing monitoring. The board is not bullish.

These two problems compound each other. A team that overpays at the top of the roster has less margin for error everywhere else. A team that also leaves wins on the table through conservative fourth-down decisions is spending a dollar and collecting sixty cents. That is the condition of roughly a third of the league right now.

In 2025, NFL teams went for it on 19.8% of competitive fourth downs. Play-by-play data since 1999 puts the mathematically correct rate at 36.5%. The gap — 16.7 percentage points — is not a rounding error. It is a structural habit, and it has a price.

That price is measured in expected wins lost, or EWL. Think of EWL the way you think of a job that pays $20 an hour when the market rate is $37: every hour you work, you lose $17 in potential wages. EWL converts fourth-down mistakes into the same currency as the standings. Every suboptimal call chips away at a team's expected win total, one decision at a time. We graded 32,731 decisions in 2025. The ledger is long.

The three least costly coaches illustrate what discipline looks like in practice. Nick Sirianni in Philadelphia lost an average of 0.78 points of win probability per game — the lowest figure among head coaches — while agreeing with the model on 91.6% of his 144 graded decisions. He went for it 12.1% of the time against a model recommendation of 18.7%, the tightest gap on the list. Dan Campbell in Detroit posted 0.73 points lost per game and agreed with the model 83.8% of the time across 139 decisions. Mike McDaniel in Miami was the most conservative of the three in absolute terms — a 14.3% go rate against a 29.8% recommendation — but kept his per-game cost to 0.66 points, the best mark in this data set. None of these coaches are perfect. They are simply the least expensive.

The costliest coaches show what the other side looks like. Brian Daboll in New York cost the Giants 3.37 points of win probability per game — more than five times McDaniel's figure — while agreeing with the model on only 71.8% of his 136 decisions. DeMeco Ryans in Houston went for it 13.1% of the time when the model wanted 37.7%, the largest gap among any coach in the data, and posted 3.18 points lost per game. Ben Johnson, in his first year in Chicago, was graded at 2.70 points per game with a 71.7% agreement rate across 145 decisions. The decision-time grades here are about the arithmetic at the moment of the call, not what happened on the next play. The outcome is beside the point. The math is what it is before the ball is snapped.

Eighty contracts graded. Sixteen fair prices, ten overpays, fifty-four watches, and zero good values. Every verdict comes with a falsifier — the specific condition that would prove us wrong — and a re-grade date. Verdicts without falsifiers are not analysis; they are opinions. We do not publish opinions.

Patrick Mahomes at $64M a year is a fair price. That figure sits at the 98th percentile of the quarterback market. His production over the period measured sits at the 92nd percentile. The price and the output live on the same block. The falsifier runs in both directions: top-five production makes this a bargain and this verdict too conservative; production outside the top 30 makes it an overpay and this verdict too generous. We re-grade in March 2027. Fair deals rarely make headlines, which is exactly why they are worth naming when they appear.

Daniel Jones in Indianapolis is the sharpest case in the overpay column. At $44M a year, he is priced at the 94th percentile of the quarterback market, with 56% of the deal guaranteed. His realized production — injury included, because that is what the team actually received — sits at the 69th percentile over the last two seasons. A 25-percentile gap between the bill and the output has to be closed by a career year. The contract prices that career year as a near-certainty. The falsifier is precise: if Jones finishes top five among quarterbacks in 2026 production, this verdict is wrong and we say so in March 2027. That bar is not impossible. It is just not what the evidence predicts.

Will Anderson Jr. and Danielle Hunter, both with the Texans, carry WATCH verdicts rather than scored verdicts, and the reason is worth stating plainly. At $50M a year — the 100th percentile of the edge defender market — and $40.1M a year at the 99th percentile, both are priced at the very top of their position. But public play-by-play data grades pass-rush production the way it grades a receiver's statistics: it can count outcomes but cannot fully evaluate film. We price the deal, say so out loud, and watch the player. The falsifier for each is a market question, not a performance question: if the edge market's ceiling moves far enough past these numbers by 2027, they read as bargains in hindsight. We will know in March.

A rookie contract is a below-market lease with a known expiration date. The team locks in a rate set before the player has proven anything, and if the player outperforms that rate — which the best ones do dramatically — the team collects the surplus until the lease runs out. The names in this section are players still on those leases.

Harold Fannin Jr. in Cleveland is the largest windfall in the data. His rookie deal costs $1.53M against the cap — 0.5% of the Browns' cap space — and he has produced at the 88th percentile among tight ends. His pay sits at the 16th percentile of the position. The surplus, the gap between what he costs and what he produces, is 72 percentile points. Brock Bowers in Las Vegas is producing at the 98th percentile of tight ends on a deal that consumes 1.6% of the Raiders' cap. Drake Maye in New England is at the 96th percentile of quarterback production — realized value, not potential — on a deal worth $9.99M, or 2.8% of the cap. These numbers will not survive the re-signing. Enjoy them while they last.

The veteran bargain list operates differently. These players signed market-rate deals that their subsequent production has made look cheap. Devon Achane in Miami earns $3.15M — 1% of the cap, the 49th percentile of the running back market — while producing at the 93rd percentile. The surplus is 44 percentile points. Travis Kelce in Kansas City consumes 1.6% of the cap at the 53rd percentile of tight end pay while producing at the 95th percentile. His surplus is 42 points. Derrick Henry in Baltimore costs $5.59M — 1.8% of the cap — against 92nd-percentile production, a 23-point surplus. These are not flukes. They are the compounding reward of patience in roster construction: teams that signed players before the market fully priced them are now collecting the difference.

A restructured contract is a balance-transfer credit card: the monthly payment drops today, but the balance doesn't shrink. It migrates forward. The league contract database tracks the total future proration sitting on each team's books — the accumulated balance of every transfer made to keep this year's cap number manageable. The four teams carrying the most are Dallas at $249.9M, Houston at $238.1M, Buffalo at $204.0M, and San Francisco at $201.8M. Those numbers represent obligations already incurred that will arrive on future cap sheets whether or not the players are still on the roster.

The quarterback cap share figures alongside those balances matter. Dallas is carrying $249.9M in future proration while devoting 14.2% of its current cap to the quarterback position. Buffalo's $204.0M in future obligations sits alongside a 15.7% quarterback cap share — the highest among these four teams. Houston, by contrast, is running $238.1M in future proration with only 6.6% of the cap at quarterback. Different bets, different timelines, same eventual accounting.

Dead money is rent on an apartment you moved out of. Deshaun Watson in Cleveland illustrates the extreme. His deal consumes 12.1% of the Browns' cap — $40.96M — while his production over the period measured sits at the 8th percentile among quarterbacks. The surplus is negative 82 percentile points. Brandon Aiyuk in San Francisco is at the same negative-82 figure: 85th-percentile pay, 4th-percentile realized production. Evan Engram in Denver costs $14.14M at the 95th percentile of tight end pay against 23rd-percentile production. T.J. Hockenson in Minnesota is priced at the 100th percentile of the tight end market — $15.61M — against 28th-percentile realized production. In every case, the production numbers reflect what actually happened on the field, including injury. Talent and realized value are not the same thing, and the cap sheet only knows the difference after the fact. The bill for each of these deals is not coming. It is already here.