Kalshi has 404 margin-of-victory contracts open on the first week of the NFL season, and 1,179,178 of them have already changed hands. Every one of the sixteen Week 1 games carries a full ladder, from a team winning by more than a point and a half all the way out to winning by more than twenty. It is a far bigger board than the winner market, and almost nobody writes about it.
We handed the sixteen contracts at the center of that board to nine AI models and told them nothing about what anything cost. What came back was not a scatter of opinions. It was a pattern pointing one direction, and the direction is the one that should make you most suspicious of it.
New to event contracts? Start here → how to trade NFL games on Kalshi.
Every contract on this page is a CFTC-regulated event contract, open to U.S. adults 18+ in the states where Kalshi operates, and each one settles at $1 or $0 against the final score. Prices refresh on this page as the week moves, and every number below carries the timestamp it was fetched at.
Free: The Weekly PM Market Brief
The 8-model panel's graded record, the week's biggest market-vs-model gaps, and what's spiking next. One email, Sundays.
Unsubscribe anytime, one click. We never share your email.
What Kalshi's Spread Board Actually Asks
A spread contract is not a winner contract wearing a different hat. It pays only if the named team wins by MORE than a stated number, so a team can win the game outright and the contract still settles at zero. Every line carries a half point, which means a push is impossible and there are only two outcomes.
Each game is listed as a ladder of about twenty-five of those contracts, one per half-point step in each direction. For the Seahawks and Patriots opener, that runs from Seattle by more than 1.5 all the way to Seattle by more than 20.5, and the same range for New England.
Out of each ladder we took one rung: the one Kalshi itself prices closest to a coin flip. That rung is the market's own line, the number where the exchange says the two outcomes are even money. It is worth being blunt about what that choice does to the table below. Because we picked the coin-flip rung in every game, the Kalshi column sits between 47 and 54.5 cents by construction. The market did not independently decide sixteen games were toss-ups. We asked the panel to judge the sixteen lines the market had already declared to be toss-ups.
That makes the question on this page a narrow and answerable one. Kalshi says this number is a coin flip. Is it?
How Nine Models Priced The Board Without Seeing A Price
Each of the nine models got one call covering all sixteen contracts at once, with an instruction to price the week from a single view of the league rather than sixteen unrelated judgments. Asking a model about one rung at a time on a margin ladder produces numbers that contradict each other, so the whole board goes in one prompt or it does not go in at all.
What they saw was a fetched data card rather than their own memory: every team's complete 2025 season converted into a game-by-game margin profile, both teams' records and point differentials, the full quarterback room on every current roster, every non-Active designation on the league's injury list, the 2026 preseason results, the exact settlement language of each contract, and the league-wide base rates for clearing each threshold. What they did not see was a single price, spread or implied probability.
The card also carried what the feeds cannot know. The league's official Week 1 injury report has not been published yet, because teams file those in the days before kickoff. So the card added dated reporting from opener week. Atlanta still had not named a starting quarterback as of September 2. Kirk Cousins was confirmed as the Raiders starter over the rookie taken first overall. And both Cincinnati starting receivers were limited in practice, Ja'Marr Chase with a hyperextended knee and Tee Higgins with a heel contusion.
After that blind round, each model read the other eight boards anonymously and could revise once. All nine returned a board in both rounds, and all nine changed at least one number in the second. The figures below are the revised round, and the blind first round is stored either way.
All 288 calls went into our public grading ledger the moment they were made. When these games settle, every seat's Week 1 spread record posts to the full graded scoreboard. That ledger is also the reason to read what follows as an argument rather than an instruction: across 487 graded cases where this panel has disagreed with a market price by ten cents or more, the market has been right 71.3% of the time.
The Board: Every Week 1 Spread Contract
Prices are the midpoint of the best bid and best offer on Kalshi as of 5:59 PM ET on September 3, 2026. The Panel column blends the nine seats by taking the median of their probabilities rather than the average, so one outlier cannot drag the number, and converts it into the same unit, so the two columns are directly comparable: a panel figure of 47 cents means the models think a fair price is 47 cents. The board is sorted by the size of the disagreement, widest first.
Contracts are written the way the scoreboard reads them. Kalshi lists the Chargers contract as winning by more than 9.5 points, and because scores are whole numbers that is the same thing as winning by 10 or more, so the table says "Chargers by 10+".
Books here are tight. Fourteen of the sixteen contracts are quoted a single cent wide and the other two are two cents wide, so unlike a thin prop board, none of these gaps can be explained away as a market without a price.
| Contract | Game | Kalshi | Panel (fair) | Gap | Traded |
|---|---|---|---|---|---|
| Chargers by 10+ | Cardinals at Chargers | 50.5¢ | 36.0¢ | -14.5 | 2,058 |
| Raiders by 4+ | Dolphins at Raiders | 48.5¢ | 36.0¢ | -12.5 | 6,250 |
| Chiefs by 3+ | Broncos at Chiefs | 50.5¢ | 40.0¢ | -10.5 | 11,944 |
| Cowboys by 3+ | Cowboys at Giants | 54.5¢ | 45.0¢ | -9.5 | 58,593 |
| Lions by 7+ | Saints at Lions | 53.5¢ | 46.0¢ | -7.5 | 12,513 |
| Bengals by 4+ | Buccaneers at Bengals | 50.5¢ | 44.0¢ | -6.5 | 12,021 |
| Ravens by 4+ | Ravens at Colts | 49.0¢ | 43.0¢ | -6.0 | 30,635 |
| Bills by 2+ | Bills at Texans | 47.0¢ | 44.0¢ | -3.0 | 9,339 |
| Seahawks by 4+ | Patriots at Seahawks | 49.5¢ | 47.0¢ | -2.5 | 136,364 |
| Rams by 4+ | 49ers at Rams | 50.5¢ | 48.0¢ | -2.5 | 53,758 |
| Bears by 3+ | Bears at Panthers | 53.5¢ | 53.0¢ | -0.5 | 23,339 |
| Titans by 2+ | Jets at Titans | 50.5¢ | 50.0¢ | -0.5 | 1,807 |
| Eagles by 5+ | Commanders at Eagles | 51.5¢ | 51.0¢ | -0.5 | 13,928 |
| Jaguars by 8+ | Browns at Jaguars | 51.5¢ | 53.0¢ | +1.5 | 24,020 |
| Steelers by 4+ | Falcons at Steelers | 47.5¢ | 50.0¢ | +2.5 | 6,619 |
| Vikings by 2+ | Packers at Vikings | 48.5¢ | 53.0¢ | +4.5 | 15,968 |
On four of these contracts the favored team is the road team. Those are Baltimore, Buffalo, Chicago and Dallas.
One Pattern: The Panel Prices Almost Every Favorite Below The Market
On thirteen of the sixteen contracts the panel's number sits below Kalshi's. On three it sits above. The average gap across the whole board is 4.25 cents to the low side.
Remember what the Kalshi column is. Every one of those rungs is a line the market has priced as a coin flip. The panel is not saying these favorites will lose. It is saying they will win by less than the number, which on thirteen of sixteen games puts its fair price below what the contract costs.
The models named the mechanism themselves, and it is the same one every time: they treated a margin contract as a margin contract instead of as a strong opinion about who is better.
This board's biggest systematic error is treating margin contracts like winner contracts. The big-number games all deserve base-rate gravity. — Claude Fable
The base rates in the card are what gives that teeth. Across all 272 games of the 2025 season, a team won by more than 3.5 points in 36.6% of team-games. Push the ask out to more than 6.5 points and that falls to 30.0%. Push it to more than 9.5 and it falls again, to 22.8%. The median winning margin all season was 8 points. Those are unconditional numbers and a real matchup should move well off them, but they set the gravity a favorite has to climb against, and the panel kept finding that the climb was steeper than the market had priced.
Now the honest part. A gap that points the same way on thirteen of sixteen games is exactly as consistent with a biased panel as with a mispriced market. If nine models share a habit of shading margin contracts downward, that habit would produce this board. Two things argue against it being pure bias, and neither is conclusive.
The first is that the panel's own numbers are internally coherent. We checked every contract against this same panel's winner-market probabilities for the same games. On all sixteen, the probability of covering came in below the probability of winning, which it must. The ratio between the two also tracked the size of the line. It ran about 0.94 where the ask was a point and a half, and about half that where the ask was 9.5. A panel applying a flat pessimism discount would not produce a gradient that clean.
The second is that the three contracts where the panel went the other way are specific rather than random. It likes Minnesota by more than 1.5 at home because Green Bay lists both Micah Parsons and Josh Jacobs as unavailable. It likes Pittsburgh by more than 3.5 because Atlanta has no named quarterback. It likes Jacksonville by more than 7.5 because Jacksonville won by 14 or more in seven games last season, a total only the Rams beat. Those are reasons, not noise.
Against both of those sits the graded record quoted above, and it is not flattering to the panel. Read the four rows below as the argument the models make, with that record attached.
The Four Widest Gaps
The Chargers Are Being Asked To Do Something They Never Did
The largest disagreement on the board is the Chargers giving more than 9.5 points to Arizona, priced at 50.5¢ against a panel median of 36. The case against it is not about Arizona, who went 3-14, were outscored by 133 points across the season, and never once won by 14. It is about the Chargers.
Los Angeles went 11-6 last season and outscored opponents by 28 points across the whole year. That is an average margin of +1.6. Eight of their seventeen games were decided by 8 points or fewer. They are a good team that wins narrowly, and that is close to the worst possible profile for a contract requiring a ten-point win.
An 11-win close-game team is a bad vehicle for a 10-point cover regardless of opponent quality. — Claude Fable
This is also the second-lightest contract on the board at 2,058 traded, though it is quoted a single cent wide.
Kirk Cousins Did Not Rebuild The Roster
Las Vegas by more than 3.5 against Miami sits at 48.5¢ with the panel at 36. The Raiders won three games last season and were outscored by 191 points across the year. Their average margin was -11.2. They won by 4 or more twice in seventeen games and lost by 7 or more ten times.
The bull case is real: Klint Kubiak named Kirk Cousins the Week 1 starter over rookie Fernando Mendoza on September 2, and Miami counters with Malik Willis. One seat priced that at 55 in the first round. After the revision round all nine finished below 42. The seat that had been highest finished at 38.
Las Vegas won by 4+ twice all year and a stabilizer does not suddenly clear 3.5 against a live Willis offense. — Grok
Kansas City Is The Row The Panel Argues About
Kansas City by more than 2.5 against Denver is priced at 50.5¢ against a panel median of 40, and it is the row where the seats split into two camps rather than simply spreading out. Three sit at 34 to 36, three at 42 to 43.
The facts pull both ways and every seat agrees on what they are. Kansas City went 6-11 but outscored opponents by 34 points, the signature of a team that lost close games. Denver went 14-3 with 13 of its 17 games decided by 8 points or fewer, the signature of a team that won them. Both of those patterns tend to regress. Patrick Mahomes tore the ACL and LCL in his left knee last December, has not missed a practice since being cleared in July, played no preseason snaps, and is listed Questionable.
30 requires Mahomes to be functionally unavailable, and the card says he has not missed a practice. — Claude Opus, holding at 41
Denver's 13-for-17 in one-score games is luck, not identity, and Kansas City's positive differential at 6-11 is the inverse luck. The only genuine discount is Mahomes's knee. — Claude Fable, holding at 43
Dallas Is The Most Traded Disagreement
Dallas by more than 2.5 at the Giants is the busiest of the four at 58,593 contracts, quoted a cent wide at 54.5¢, with the panel at 45. It is also the row where the panel agrees with itself most closely of these four: all nine seats landed between 43 and 48.
Dallas went 7-9-1 last season and was outscored by 40 points, losing by 7 or more in seven games. The Giants went 4-13 but won by 14 or more three times. The panel is not arguing Dallas is worse. It is arguing that a team with a negative point differential laying points on the road is a thin proposition, and that the market's 54.5¢ is pricing the Cowboys brand rather than the Cowboys margin.
Here is every seat's number on the six contracts the panel prices lowest. Dallas is the next one out, at 45.
| Contract | Kalshi | ChatGPT | Claude Fable | Claude Sonnet | Gemini | Claude Opus | Grok | Kimi | DeepSeek | GLM | Panel |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Chargers by 10+ | 50.5¢ | 34% | 36% | 37% | 37% | 36% | 33% | 40% | 28% | 40% | 36.0% |
| Raiders by 4+ | 48.5¢ | 36% | 35% | 40% | 38% | 36% | 28% | 32% | 40% | 31% | 36.0% |
| Chiefs by 3+ | 50.5¢ | 39% | 43% | 40% | 34% | 41% | 34% | 43% | 42% | 36% | 40.0% |
| Ravens by 4+ | 49.0¢ | 42% | 43% | 43% | 35% | 43% | 33% | 45% | 48% | 39% | 43.0% |
| Bengals by 4+ | 50.5¢ | 44% | 43% | 46% | 41% | 45% | 40% | 48% | 45% | 43% | 44.0% |
| Bills by 2+ | 47.0¢ | 44% | 43% | 47% | 44% | 42% | 47% | 44% | 45% | 43% | 44.0% |
Every seat on this panel is graded against real market settlements — records to date: Gemini 86% on 6,624 graded calls · Grok 84% on 5,579 graded calls · GLM 82% on 2,941 graded calls · Kimi 84% on 3,119 graded calls. Recomputed daily; the full scoreboard is public.
These numbers are model estimates, not predictions of fact and not financial advice. Event contracts involve risk, and you can lose money trading them.
More live boards from the same panel: the same panel on all sixteen Week 1 winners (Seattle 62.5¢ over New England) · Kalshi's Week 1 player props (Drake Maye 250+ passing yards 32.5¢) · where the Seahawks-Patriots line actually moved. Prices fetched September 3, 2026.
Where The Panel Moved, And What Would Move It Again
Where The Panel Changed Its Mind
All nine seats revised after reading the other eight anonymously. The useful revisions were the ones where a model went back to the card and found a personnel fact it had read but not priced.
Dropped Las Vegas, conceding to the panel that a -191 base rate is too catastrophic to lay 3.5 points despite the Cousins upgrade. — Gemini, cutting that contract from 55 to 38
Moved Green Bay at Minnesota from 43 to 52. Seven of eight peers priced this above me, and they specifically weighted Micah Parsons on PUP plus Josh Jacobs out as decisive for a 1.5-point number. That was dated news I had, and underweighted. — Claude Sonnet
Moved Chicago at Carolina up on the specific point that both Carolina starting tackles are out. I had not weighed the tackle pair. — Claude Fable
One seat held its ground where the pressure was heaviest, and said why:
I hold below consensus on Kansas City because Denver's low-variance profile and Bo Nix continuity still make a Kansas City cover unlikely. — GLM
What Would Change The Panel's Mind
Four things, each of them checkable and each with a direction.
Atlanta names a quarterback. As of September 2 the Falcons had still not said who starts. If Michael Penix Jr. is announced and comes off the injury report, the Pittsburgh contract gets harder for the home side and the panel's 50 comes down. If Tua Tagovailoa gets it, or the announcement waits until Sunday, it goes the other way.
The official Week 1 injury report lands. It does not exist yet. When teams file, the Cincinnati receivers, Sam LaPorta and Detroit's two starting safeties, and Mahomes all get a formal game status for the first time. The Detroit contract in particular is priced by a panel that can only see Kerby Joseph and Brian Branch listed as out, without knowing whether that holds.
Mahomes is cleared for game action, or is not. He is Questionable with no preseason snaps. A full clearance pushes the Kansas City number up toward the market. Anything short of it pushes the panel further from it.
A late line move on the Chargers. If Kalshi's coin-flip rung on Arizona at the Chargers slides from 9.5 down to 7.5 or lower, the board's widest gap closes without anyone changing their mind about the teams.
Settlement Timeline
| What | When |
|---|---|
| First contracts settle (New England at Seattle) | Wed Sep 9, after the final whistle |
| Melbourne game settles (San Francisco vs LA Rams) | Thu Sep 10 |
| Thirteen contracts settle | Sun Sep 13 |
| Last contract settles (Denver at Kansas City) | Mon Sep 14 |
| Panel's Week 1 spread record posts | Within a day of the last settlement |
Every call on this page is already in the public ledger, and this page gets re-scored when the story moves. Prices as of September 3, 2026 at 5:59 PM ET.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $210M traded
- 2027 Pro Football Champion · $79M traded
- Pro Baseball Champion · $65M traded
- 2028 U.S. Presidential Election winner? · $64M traded
- 2028 Republican presidential nominee · $62M traded
- Bitcoin price at the end of 2026 · $34M traded
Every market above links to our full AI model verdict; browse them all on the OddsShopper prediction markets hub, and see how every settled call actually scored on the full graded scoreboard.
The Bottom Line
Kalshi has told you, sixteen times, where it thinks a game is an even-money proposition. A panel of nine models that never saw those prices disagreed with thirteen of them, and disagreed in the same direction each time: the favorite wins, but by less than the number.
That is either a real disagreement in a market almost nobody covers, or it is nine models sharing one bad habit. The graded record says to lean toward the second. What makes the board worth reading anyway is that the arguments are specific and checkable. The Chargers really did average +1.6 a game. The Raiders really were outscored by 191 points across last season. Carolina really is missing both starting tackles.
Nine days from now the scores will settle every one of these, and the record goes up in public either way.
