NFL Week 5 Card Grading: The First Football Card Goes 2-1 for +3.425u, the Teaser Cashes, and the Floor Rule Passes Its First Live Test 2-0

The house's first-ever NFL desk card is graded. It was published Sunday at 10:30 a.m. MT with three legs, a declined machine leg, a promoted fourth, and a printed promise: everything grades at the published prices, wins and losses, no retroactive claims. Here is everything.

The desk card, graded

Two things about that top leg. Green Bay closed a 2.5-point underdog and won by 24 — a 26.5-point cover, one of the widest gaps between line and score on any ticket this desk has carded in any sport this October. And the 2.0-unit stake was the card's largest, which is what the edge said to do — +18.9 points of model probability against a 43.6% market is a big number, and big numbers get big units. The scorecard is the quality of logged decisions. Sunday logged three, and the sizes were logged before kickoff.

The deviation, scored both ways

The machine's card had Miami 1.5u @4.10. The desk declined it — edge inside the model's known error band, quarterback-blind to a Willis-versus-Burrow game, zero NFL calibration games behind the number — and promoted Seattle instead. Both teams lost. Cincinnati beat Miami 31-24, and San Francisco beat Seattle 32-29.

The swap saved 0.5u ex post. That sentence gets its luck label in permanent ink: a swap that saves money because both legs lost is not evidence the process was right, it is evidence the outcomes were kind. The process defense stands on its own terms — the Miami edge was the model's largest printed number in a sport where the model has never been graded, and "biggest number in the least calibrated sport" is a named error on this desk's shortlist. Seattle was the promoted alternative, and it graded −1.0u like any honest leg. The machine's own Miami leg settles at −1.50u in the ledger. Nobody here is confused about which side of the card did the work.

The teaser, as placed

The loop's Sunday sheet floated one teaser before the window: Cleveland +8.5 and Green Bay +8.0 at an assumed 1.91 payout, marked pending. It was placed. Both legs covered — the Browns lost to the Jets by exactly five, and the Packers won by 24 — so the ticket grades a win at the assumed price: +0.91u. The fill price was not on the slip in front of us when this published; it grades at the sheet's 1.91 and the log records the fill when the slip surfaces, per the standing convention from the division series. If the window paid worse, the grade gets corrected in the open.

The machine's day

The loop's full-day card stamped Sunday morning was Miami 1.5u, Green Bay 2.0u, Milwaukee 0.25u — the baseball leg bumped Pittsburgh on the full-day board, a swap disclosed in the log rather than retro-edited into the article. Graded: −1.50, +3.30, +0.19, for +1.99u, with the Milwaukee leg already folded into the Game 1 grading piece. The NFL-scoped version printed in the lean card — Miami, Green Bay, Pittsburgh — would have graded +2.925u. Either scope, the machine finished the day green on the strength of the same Packers ticket.

And the humility number, printed because it does not change after a winning Sunday: the loop's NFL backtest runs a 0.2279 Brier across 816 games against the market's 0.2126. One card does not close that gap. Twenty cards might tell us something.

The floor rule's first live test: 2-0

The house rule printed Sunday morning — raw model probability under 35% means no bet, whatever the capped number says — faced its first two live edges that afternoon. Both fake edges lost. The model's Tennessee number (a +8.9-point printed edge off a 24.4% raw) lost 26-3 to Houston. Its Arizona number (+3.5 printed, 23.0 raw) lost 45-21 to Detroit. Two passes, two losses, zero units donated to the concept that a cap can manufacture a probability. The floor rule earned its ink.

The leans we didn't bet

Minnesota was the day's strongest raw opinion and the framework could not capture it — a home favorite priced past the card's staking rules. It won anyway. A lean is a logged decision with no stake attached; it grades right or wrong in this column and nowhere else. Philadelphia and Cleveland were both named and both lost, which is the leans column doing its job too: showing the reader the misses that a card-only record would hide. Baltimore voided itself when Lamar Jackson was ruled out Sunday morning; Atlanta won the night game, and a void that would have lost is graded a void, not a win.

Tickets, as placed

Every leg of the desk card was placed at the published prices, owner confirmed. The teaser was placed, fill price to log. The two National League dinger slips from Sunday were placed and graded that night in the Game 1 piece — Muncy and Contreras, 0-2, −0.50u, house dingers 0-3 for October with the count carried forward to tonight's card. The portfolio framing stays honest because every ticket above is either graded or named as pending, and none is retroactively claimed.

The survivor card — New England and Detroit, both winners, graded Monday with the re-based prices and the phantom-line autopsy — lives in its own piece and is not double-counted here. Process over outcome, but outcome pays the ledger: desk NFL +3.425u, teaser +0.91u, machine +1.99u across two sports. Bring your own pencil.