How the football grades have done · 2026-09-09 – 2026-09-14
The registered read (declared 2026-09-08, before any game settled). Per market, the A leans: flat-stake return at the price the card showed, the same result at the close, and beat-the-close on the leans whose line moved. First look at 30 settled leans in a market, season-end look after week 18. Verdicts: edge-shaped = return above zero and beat-the-close above 52%; noise = return within five points of zero and beat-the-close 48-52%; wrong = return under -5% or beat-the-close under 48%, and the band comes off on the owner's word; overconfident = the model's mean P(lean side) exceeds the hit rate by more than 8 points, and the cut is re-derived. No cut, feature or blend is tuned on 2026 rows before the first look. Break-even at -110 is 52.4%.
Columns. RETURN is the flat-stake return per $1 at the price the card showed. AT CLOSE is the same result priced at the captured close. BEAT is how often the lean's side got a better number than the close, counted over the leans whose line moved. MODEL is the model's mean chance on the lean's side, to hold against WIN %.
OVERALL
| SEGMENT | N | W | L | PUSH | WIN % | RETURN | AT CLOSE | BEAT | MODEL | NOTES |
|---|
| Every settled grade | 48 | 27 | 20 | 1 | 57.5% | +11.7% | +11.7% | 36% of 14 | 53.4% | |
| of which graded — a lean | 3 | 1 | 2 | 0 | 33.3% | -37.7% | -37.7% | 0% of 2 | 51.4% | insufficient sample |
| of which no lean — coverage | 45 | 26 | 18 | 1 | 59.1% | +15.0% | +15.0% | 42% of 12 | 53.5% | |
The two indented rows sum to the first. Graded is the pick record — every card where the model claimed an edge. No lean is every card where it agreed with the market; settlement still writes a W or an L against its nominal side, which is bookkeeping, not a call.
BY MARKET AND BAND· the pick record, crossed · the verdict reads the A leans
| SEGMENT | N | W | L | PUSH | WIN % | RETURN | AT CLOSE | BEAT | MODEL | NOTES |
|---|
| Spread · A | 1 | 1 | 0 | 0 | 100.0% | +87.0% | +87.0% | — | 56.9% | first look at 30 · 1 so far |
| Spread · all graded picks | 1 | 1 | 0 | 0 | 100.0% | +87.0% | +87.0% | — | 56.9% | insufficient sample |
| Moneyline · A | 2 | 0 | 2 | 0 | 0.0% | -100.0% | -100.0% | 0% of 2 | 48.7% | first look at 30 · 2 so far |
| Moneyline · all graded picks | 2 | 0 | 2 | 0 | 0.0% | -100.0% | -100.0% | 0% of 2 | 48.7% | insufficient sample |
Verdicts. Spread A: first look at 30 · 1 so far · Moneyline A: first look at 30 · 2 so far
NO LEAN — COVERAGE· the market's right; not a track record
| SEGMENT | N | W | L | PUSH | WIN % | RETURN | AT CLOSE | BEAT | MODEL | NOTES |
|---|
| Moneyline | 14 | 9 | 5 | 0 | 64.3% | +27.1% | +27.3% | 40% of 10 | 54.2% | insufficient sample |
| Spread | 15 | 8 | 6 | 1 | 57.1% | +10.5% | +10.5% | — | 52.3% | insufficient sample |
| Total | 16 | 9 | 7 | 0 | 56.2% | +8.6% | +8.6% | 50% of 2 | 54.0% | insufficient sample |
PLAYER PROPS· the projection ledger, settled from the play-by-play
Same discipline, declared the same day. First look at 100 settled leans overall and 30 STRONGEST leans, read separately: return at the lean price, hit rate, and the calibration of every priced projection (the model's mean P(over) against the over rate). Same verdict table. The prior is the one prop lane run before this: MLB strikeouts, 49,559 priced closes, lost to the close. A lean is the projection's leftover after the vig-free price clearing 8 points; STRONGEST is the card's own top three at the last pregame build — firm, one per driver, best expected return — frozen in the ledger so this table reads what the card showed.
| SEGMENT | N | HITS | PUSH | HIT % | RETURN | MODEL | NOTES |
|---|
| Every lean | 371 | 192 | 0 | 51.7% | +15.2% | 67.2% | |
| of which STRONGEST — the card's top three | 39 | 22 | 0 | 56.4% | +11.4% | 70.4% | |
| of which firm — no flag at the build | 122 | 60 | 0 | 49.2% | -4.6% | 67.4% | |
| of which BEST OF THE BOARD — the sheet's top five on a day of two or more games | 5 | 3 | 0 | 60.0% | +15.6% | 81.3% | insufficient sample |
PROPS BY MARKET· every lean, split by what was projected
| SEGMENT | N | HITS | PUSH | HIT % | RETURN | MODEL | NOTES |
|---|
| anytime TD | 37 | 9 | 0 | 24.3% | +97.4% | 30.9% | |
| completions | 15 | 8 | 0 | 53.3% | +0.3% | 67.8% | insufficient sample |
| pass attempts | 16 | 7 | 0 | 43.8% | -16.2% | 68.3% | insufficient sample |
| passing yards | 13 | 8 | 0 | 61.5% | +16.4% | 67.2% | insufficient sample |
| receiving yards | 99 | 55 | 0 | 55.6% | +5.1% | 71.2% | |
| receptions | 102 | 60 | 0 | 58.8% | +17.8% | 69.6% | |
| carries | 42 | 22 | 0 | 52.4% | +0.0% | 76.1% | |
| rushing yards | 47 | 23 | 0 | 48.9% | -5.5% | 73.4% | |
STRONGEST BY MARKET
| SEGMENT | N | HITS | PUSH | HIT % | RETURN | MODEL | NOTES |
|---|
| completions | 2 | 1 | 0 | 50.0% | -7.0% | 63.4% | insufficient sample |
| pass attempts | 4 | 2 | 0 | 50.0% | -3.8% | 68.3% | insufficient sample |
| passing yards | 2 | 2 | 0 | 100.0% | +93.0% | 66.4% | insufficient sample |
| receiving yards | 2 | 0 | 0 | 0.0% | -100.0% | 69.9% | insufficient sample |
| receptions | 14 | 7 | 0 | 50.0% | +5.1% | 68.1% | insufficient sample |
| carries | 8 | 5 | 0 | 62.5% | +21.4% | 75.5% | insufficient sample |
| rushing yards | 7 | 5 | 0 | 71.4% | +34.7% | 73.5% | insufficient sample |
Calibration, every priced projection. The model said over 44% of the time on average; the over hit 46% (n 666). A gap of more than 8 points in either direction is the overconfident verdict for this lane.