NFL

Markets · Week 4

Our probability next to what the markets say, captured at every run from Monday to kickoff. "Open" = the first snapshot we captured, "close" = the last before kickoff. Sources are named; none of them is a model input. · predictions are published to show the model works. nothing here is betting advice.Last updated: predictions Oct 1, 11:02 AM ET · results graded Sep 28, 11:22 PM ET
W1W2W3W4W5 6,245 snapshots since Sep 12

Games to watch · where our pick differs from the market

These 3 games are our ones to watch coming into Week 4, where our model backs a different team than the market's market pick at kickoff and our pick is at least 52% (near coin-flips are left out). Both numbers are the probability that our pick wins; the market is a comparison, never a model input.
ARI @ NYGSun 1:00 PM
We back NYG · the market leans ARI
niftiNFL60%
market44%
PIT @ CLEThu 8:15 PM
We back CLE · the market leans PIT
niftiNFL56%
market43%
JAX @ CINSun 1:00 PM
We back JAX · the market leans CIN
niftiNFL52%
market43%
Share this: the card image, or post the Markets page and it unfurls with this card.

Season record vs the markets

Straight-up, on the 46 graded games. "Differ" = games where our pick was the other side of that market's pick, and how often we were right there.
MarketAtGamesMarket rightWe were rightAgreedDifferedOur record when we differed
DraftKingsclose4665.2%63.0% 3973–4
DraftKingsopen3259.4%62.5% 2932–1
Kalshiclose3363.6%60.6% 3031–2
Kalshiopen3357.6%60.6% 2853–2
nflverse priceclose4665.2%63.0% 3973–4
nflverse priceopen4660.9%63.0% 3974–3
Polymarketclose4665.2%63.0% 3973–4
Polymarketopen4660.9%63.0% 3974–3
Big divergences (our closing probability at least 8 points from Polymarket's): 11 games: we were right in 7, Polymarket in 8.

Where these prices come from

Four public sources, none of them a model input. We show them so our number can be judged against the market's, and we say plainly which links pay us.
SourceWhat it isHow we read it
PolymarketA prediction market: people trade contracts on each team, so the price is the crowd's probability.Public API, no account. Price of the home team's contract.
Kalshi · kalshi.comA prediction market regulated by the CFTC as an exchange (not a sportsbook). Same idea: the price is the probability.Public API, no account. Midpoint of the bid and ask, or the last trade.
DraftKings via ESPNA sportsbook price as displayed on ESPN's scoreboard.Market price with the bookmaker's margin removed; the market margin when no market price is posted.
nflverse priceThe consensus market's final price in the open nflverse dataset, the same price every backtest is graded against.Market margin, total and market price as published.
Disclosure. If you open a Kalshi account through this link and make a first trade, Kalshi pays us a referral bonus and gives you one too. That is the only money this site makes from any market, and it never changes a prediction. Kalshi is for U.S. residents 18 and over. Predictions are published to show the model works. Nothing here is betting advice. If gambling is a problem for you or someone you know, call 1-800-GAMBLER.

Week 4 · probability the home team wins

Each cell is open → close for that source, including ours. Bold = the source's pick at close. Market margin in brackets is home-minus-away (positive = home picked); ours is the model's projected margin. Polymarket and Kalshi prices are the market's own probability; sportsbook price are de-vigged from the market price. green = that source's market pick at kickoff is our pick; red = we differ.
KickoffGameniftiNFL (runs) PolymarketKalshiDraftKingsnflverse Result
Thu 8:15 PM PIT @ CLE 53% (+0.5) → 56% (+1.6) (38) 44% (-2.5) → 42% (-2.5) 44% → 42% 43% (-2.5) → 43% (-2.5) 42% (-3.0) → 43% (-2.5) pending
Sun 9:30 AM IND @ WAS 49% (-1.0) → 40% (-3.0) (27) 32% (-4.5) → 36% (-3.5) 34% → 36% 40% (-3.5) → 40% (-3.5) 36% (-4.5) → 37% (-3.5) pending
Sun 1:00 PM TEN @ BAL 86% (+14.3) → 86% (+14.2) (26) 84% (+10.5) → 86% (+10.5) 84% → 86% 80% (+11.5) → 80% (+11.5) 82% (+10.5) → 84% (+11.5) pending
Sun 1:00 PM NE @ BUF 65% (+5.1) → 76% (+9.3) (26) 70% (+5.5) → 72% (+6.5) 70% → 74% 69% (+6.5) → 69% (+6.5) 69% (+5.5) → 71% (+6.5) pending
Sun 1:00 PM NYJ @ CHI 66% (+5.1) → 74% (+8.1) (26) 63% (+3.5) → 62% (+3.5) 64% → 62% 60% (+3.5) → 60% (+3.5) 60% (+3.0) → 61% (+3.5) pending
Sun 1:00 PM JAX @ CIN 56% (+2.5) → 48% (-1.1) (26) 58% (+2.5) → 57% (+2.5) 58% → 57% 57% (+2.5) → 57% (+2.5) 60% (+3.0) → 56% (+2.5) pending
Sun 1:00 PM DAL @ HOU 60% (+2.4) → 62% (+2.9) (26) 57% (+2.5) → 60% (+2.5) 57% → 60% 57% (+2.5) → 59% (+3.0) 59% (+3.0) → 59% (+3.0) pending
Sun 1:00 PM ARI @ NYG 61% (+2.8) → 60% (+2.9) (26) 56% (+2.5) → 44% (-2.5) 55% → 44% 46% (-1.5) → 43% (-2.5) 55% (+2.5) → 45% (-2.5) pending
Sun 1:00 PM LA @ PHI 47% (-0.7) → 49% (-0.5) (26) 43% (-1.5) → 40% (-3.5) 43% → 40% 41% (-3.0) → 41% (-3.0) 45% (-1.5) → 41% (-3.0) pending
Sun 1:00 PM GB @ TB 52% (+1.0) → 42% (-1.5) (26) 48% (-1.5) → 38% (-3.5) 48% → 38% 40% (-3.5) → 40% (-3.5) 48% (-1.5) → 38% (-3.5) pending
Sun 4:05 PM MIA @ MIN 74% (+9.8) → 80% (+11.8) (25) 85% (+10.5) → 84% (+10.5) 83% → 84% 78% (+10.5) → 78% (+10.5) 84% (+10.5) → 82% (+10.5) pending
Sun 4:25 PM KC @ LV 38% (-3.9) → 36% (-4.7) (25) 32% (-5.5) → 34% (-4.5) 31% → 35% 37% (-4.5) → 37% (-4.5) 31% (-6.0) → 34% (-4.5) pending
Sun 4:25 PM LAC @ SEA 81% (+11.3) → 74% (+8.6) (25) 74% (+6.5) → 74% (+6.5) 74% → 76% 70% (+7.0) → 70% (+7.0) 71% (+6.5) → 76% (+7.0) pending
Sun 4:25 PM DEN @ SF 62% (+4.8) → 63% (+4.7) (25) 59% (+4.5) → 56% (+2.5) 60% → 57% 59% (+3.0) → 59% (+3.0) 61% (+3.0) → 57% (+3.0) pending
Sun 8:20 PM DET @ CAR 42% (-1.7) → 34% (-4.4) (24) 43% (+2.5) → 36% (-3.5) 44% → 36% 40% (-3.5) → 40% (-3.5) 42% (-3.0) → 38% (-3.5) pending
Mon 8:15 PM ATL @ NO 50% (+0.5) → 50% (+0.5) (17) 58% (+2.5) → 56% (+2.5) 57% → 56% 59% (+3.0) → 57% (+2.5) 57% (+2.5) → 57% (+2.5) pending
Sources: nflverse schedule feed (the market's final price used in every backtest), Polymarket public API (prices are the market's own probability; liquidity varies by game), ESPN's scoreboard feed (provider as named by ESPN). All free, none paid, none a model input.