Denver went 14-3 last season. Kansas City went 6-11. They meet on Monday night to close Week 1, and the market that has to put a real number on it makes Kansas City something close to a coin flip at home.
That is not an oversight. It is a knee. Patrick Mahomes tore the ACL and LCL in his left knee on December 14, and twelve days before kickoff he still has not been cleared for game action, though he has not missed a camp practice and told reporters on Tuesday he is on track to start. Everyone pricing this game is pricing the same uncertainty.
So we ran an experiment. We gave a panel of AI models every fact we could fetch about this matchup, withheld every price and every point spread, and asked each one to price the entire Kalshi board for the game from scratch: the winner, all 25 winning-margin questions, all 19 total-points questions. Forty-six contracts, priced blind, then laid next to what the market actually charges.
The result was not what we expected, and it is not really about Mahomes' knee. It is about a pattern that runs through every tightly quoted line on the board, in the same direction on all of them, and about one contract in the middle of the board where the market and the machines describe two completely different football games. That contract is not the winner market, and it is the last thing in this piece.
New to event contracts? Start here → how to trade NFL games on Kalshi.
How This Board Was Priced
Every model received one identical fact card, built at run time on September 2, 2026. Nothing in it was written from memory.
The card carried both teams' complete 2025 game-by-game scoring ledgers, their point differentials, every meeting in this rivalry since 2018 with the final score and margin of each, Kansas City's home record and average margin at Arrowhead since 2021, and league-wide base rates from 1,359 regular-season games covering how often football games actually land on each winning margin and each total. It carried both 2026 rosters, the live league injury report, and a dated status record for both quarterbacks sourced to the reporting behind it.
It carried no prices. No point spread, no total, no moneyline, nothing that would let a model reverse-engineer what the market already believes. That is the whole design: a model that can see the price will drift toward the price, and then it has told you nothing.
All eight seats priced the blind round, and seven of them came back with a coherent revised board; those seven are the numbers below. Each one had to state two things before pricing anything: a distribution for the final margin, and a distribution for the total points. Every one of the 46 answers then had to be derived from those two distributions, so a seat could not contradict itself across the board. Answers that failed that check were rejected and re-run.
Then the seats read each other. After the blind first round, each model was shown all eight sets of first-round numbers and reasoning, stripped of model names, and given one chance to revise. Revising was optional, and holding a number against the group was explicitly written into the instructions as a legitimate answer. The revised round is the board below. In the quotes further down, the seats refer to each other as Forecaster A through H: those letters are the anonymous labels each model saw, not model names.
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.
The Board: Every Line On Broncos-Chiefs
Kalshi prices below are the midpoint between the best bid and the best offer, as of September 2, 2026 at 7:11 PM ET. "Our AI" is the middle number of the seven seats. On a Kalshi contract the price in cents is the market's probability directly: 55 cents to win a dollar means the market thinks it happens 55% of the time, so the two columns compare cleanly with no conversion.
A note on which lines are here and which are not. Kalshi lists 256 contracts on this one game across 38 separate market menus: quarter winners, halves, first team to score a touchdown, races to seven and fourteen and twenty-one points, overtime, total touchdowns. Three of those menus are quoted end to end today, and they are the three that decide the game itself: the winner, the 25-rung winning-margin ladder and the 19-rung total-points ladder. Those 46 contracts are the board below. Across the other 35 menus exactly one contract carries a real two-sided quote, a first-half tie at 15 bid and 16 offered; the remaining 209 sit at placeholders, a penny bid against a ninety-eight-cent offer, with no volume and no open interest behind them. Those menus fill in the hours before kickoff, not twelve days out. Listing a contract is not the same as quoting it, and neither is the same as trading it, so the placeholders are excluded rather than dressed up as a board.
| Question | Market | Our AI | Gap |
|---|---|---|---|
| Chiefs win the game | 54.5¢ | 50.8% | -3.7 |
| Broncos win the game | 45.0¢ | 48.9% | +3.9 |
| Chiefs win by 2 or more | 53.5¢ | 47.0% | -6.5 |
| Chiefs win by 3 or more | 49.5¢ | 43.7% | -5.8 |
| Chiefs win by 4 or more | 43.0¢ | 36.9% | -6.1 |
| Chiefs win by 5 or more | 40.5¢ | 34.5% | -6.0 |
| Chiefs win by 6 or more † | 37.5¢ | 32.3% | -5.2 |
| Chiefs win by 7 or more | 34.5¢ | 29.1% | -5.4 |
| Chiefs win by 8 or more | 29.5¢ | 25.4% | -4.1 |
| Chiefs win by 10 or more † | 23.5¢ | 22.3% | -1.2 |
| Chiefs win by 11 or more † | 21.0¢ | 20.0% | -1.0 |
| Chiefs win by 14 or more † | 16.5¢ | 16.8% | +0.3 |
| Chiefs win by 15 or more † | 14.0¢ | 14.8% | +0.8 |
| Chiefs win by 17 or more † | 11.5¢ | 13.0% | +1.5 |
| Chiefs win by 18 or more † | 10.0¢ | 11.2% | +1.2 |
| Chiefs win by 21 or more † | 7.5¢ | 8.5% | +1.0 |
| Broncos win by 2 or more | 39.5¢ | 46.5% | +7.0 |
| Broncos win by 3 or more | 37.5¢ | 43.8% | +6.3 |
| Broncos win by 4 or more | 30.5¢ | 36.6% | +6.1 |
| Broncos win by 5 or more † | 27.0¢ | 34.1% | +7.1 |
| Broncos win by 6 or more | 25.0¢ | 31.9% | +6.9 |
| Broncos win by 7 or more | 23.0¢ | 28.5% | +5.5 |
| Broncos win by 8 or more | 18.0¢ | 24.7% | +6.7 |
| Broncos win by 10 or more | 15.0¢ | 21.7% | +6.7 |
| Broncos win by 11 or more † | 12.5¢ | 19.5% | +7.0 |
| Broncos win by 14 or more † | 9.5¢ | 16.4% | +6.9 |
| Broncos win by 15 or more † | 8.5¢ | 14.5% | +6.0 |
| Over 21.5 total points | 96.0¢ | 95.0% | -1.0 |
| Over 24.5 total points | 92.5¢ | 91.3% | -1.2 |
| Over 27.5 total points | 89.5¢ | 86.7% | -2.8 |
| Over 30.5 total points † | 84.5¢ | 80.9% | -3.6 |
| Over 33.5 total points | 77.5¢ | 73.5% | -4.0 |
| Over 36.5 total points | 70.0¢ | 65.4% | -4.6 |
| Over 39.5 total points | 61.5¢ | 56.3% | -5.2 |
| Over 40.5 total points | 57.5¢ | 53.0% | -4.5 |
| Over 41.5 total points | 55.5¢ | 50.0% | -5.5 |
| Over 42.5 total points | 51.5¢ | 46.8% | -4.7 |
| Over 43.5 total points | 48.0¢ | 43.7% | -4.3 |
| Over 44.5 total points | 45.5¢ | 40.6% | -4.9 |
| Over 45.5 total points | 43.0¢ | 37.6% | -5.4 |
| Over 48.5 total points † | 32.0¢ | 28.9% | -3.1 |
| Over 51.5 total points † | 24.5¢ | 21.5% | -3.0 |
| Over 54.5 total points | 19.0¢ | 16.0% | -3.0 |
| Over 57.5 total points | 14.0¢ | 11.0% | -3.0 |
| Over 60.5 total points | 9.0¢ | 7.2% | -1.8 |
| Over 63.5 total points | 8.0¢ | 4.5% | -3.5 |
These numbers are model estimates, not predictions of fact and not financial advice. Event contracts involve risk, and you can lose money trading them. Kalshi is a CFTC-regulated event-contract exchange (18+; availability varies by state, as of September 2026), and trading fees apply.
Three housekeeping notes on the table. The fifteen rows marked † are quoted five to nine cents wide, which is a market maker holding a placeholder open rather than a crowd agreeing on a number; the midpoint shown for those rows is arithmetic, not a price anybody has to pay. The "over 45.5 total points" and "over 48.5 total points" rungs carry live quotes but have not traded a single contract yet. And the winner market is where the money is: the Denver contract alone has taken more than 113,000 contracts of volume, close to four times the traffic on the Kansas City side of the identical question.
Every Seat's Number
The full per-seat numbers on the winner market, the margin rungs either side of the key number, and the two totals closest to the market's own. This is the revised round, after each model read the others.
| Question | Fable | Opus | Sonnet | GPT | Gemini | Grok | Kimi | Middle |
|---|---|---|---|---|---|---|---|---|
| Chiefs win the game | 48.9 | 52.8 | 50.8 | 51.0 | 51.3 | 47.9 | 48.5 | 50.8 |
| Broncos win the game | 50.8 | 46.9 | 48.9 | 48.7 | 48.4 | 51.8 | 51.2 | 48.9 |
| Chiefs win by 3+ | 43.7 | 47.3 | 45.7 | 43.6 | 46.5 | 43.0 | 43.5 | 43.7 |
| Broncos win by 3+ | 45.5 | 41.5 | 43.8 | 40.5 | 43.5 | 46.5 | 46.0 | 43.8 |
| Chiefs win by 4+ | 36.8 | 39.3 | 38.2 | 36.9 | 39.5 | 35.9 | 36.4 | 36.9 |
| Broncos win by 7+ | 29.6 | 25.4 | 28.5 | 26.0 | 28.5 | 30.5 | 30.1 | 28.5 |
| Over 42.5 total points | 44.0 | 42.5 | 46.8 | 47.0 | 47.0 | 47.5 | 46.8 | 46.8 |
| Over 40.5 total points | 50.5 | 49.0 | 53.2 | 53.0 | 53.0 | 53.5 | 53.2 | 53.0 |
Every seat on this panel is graded against real market settlements — records to date: Claude Fable 87% on 1,529 graded calls · Claude Opus 87% on 1,601 graded calls · Claude Sonnet 85% on 1,577 graded calls · GPT 85% on 7,849 graded calls · Gemini 86% on 6,535 graded calls · Grok 84% on 5,491 graded calls · Kimi 84% on 3,113 graded calls. Recomputed daily; the full scoreboard is public.
These numbers are model estimates, not predictions of fact and not financial advice.
Every number in this piece gets graded in public once this game settles; the running tally lives on the full graded scoreboard.
More live boards from the same panel: AFC West odds — the same two teams for the division, priced almost dead even right now at Kansas City 33.5¢ and Denver 31.5¢ · AFC Championship, where Kansas City trades at 11.5¢ and Denver at 9.5¢ · the full Week 1 AI board on all 16 games. Prices fetched September 2, 2026.
One Pattern Runs Through The Whole Board
Read the gap column from top to bottom and the thing that jumps out is not the size of any single number. It is that every number points the same way.
Thirty-one of the 46 lines are quoted tighter than five cents, which is the range where a Kalshi quote is telling you something. On all thirty-five the panel lands on the same side of the price: below the market on Kansas City winning, below it on every Kansas City margin rung, below it on every total-points rung, above it on Denver winning and above it on every Denver margin rung. Add the eight wide-quote rows where the panel still finishes outside the quote altogether and the count is thirty-nine lines pointing one way and zero pointing the other.
The seven rows left over are the deep Kansas City blowout rungs, Chiefs by 10 or more through Chiefs by 21 or more, and there the panel's number lands inside the bid-ask spread. On a contract quoted twelve bid at twenty-one, a model saying 16.8 is not disagreeing with anyone. It is filling in a number the market has not bothered to have. Those seven are marked on the board and left out of the count, because counting them as disagreements would be counting the width of an empty market as a finding.
What is left is not seven models each finding a different quarrel with the market and landing on the same side by coincidence. It is one disagreement about the football game showing up thirty-nine times.
The disagreement has two components, and they are independent of each other.
The first is about who is better. The models kept coming back to a fact that the standings hide. Denver won 14 games and outscored its opponents by 90 points. Kansas City won 6 games and outscored its opponents by 34. Those two facts do not fit together the way records usually do, and the reason is one-score games: Denver went 11-2 in games decided by eight points or fewer, and Kansas City went 1-9. Run both point differentials through a Pythagorean estimate, the standard way of asking how many games a team's scoring "should" have won, and Denver comes out around 11 wins rather than 14, Kansas City around 9.5 rather than 6. The eight-game gap in the standings is closer to a game and a half in the scoring. Close-game records are famously unstable from one season to the next, which is exactly why the models refused to treat 14-3 against 6-11 as a description of the two rosters.
Layered on that is the mechanical matchup. Denver recorded 68 sacks last season. Kansas City recorded 35. That is the widest single statistical gap between these two rosters, and it points at the softest thing on the field: a Kansas City offensive line breaking in new pieces, with left tackle Josh Simmons listed questionable with a back injury and guard Trey Smith working through a hip. Andy Reid described the group on Wednesday as "just a matter of them getting adjusted to everything, from the snap count to the plays to the calls to the pass protection." A quarterback playing his first live football since December, behind a line still learning its calls, against the league's most productive pass rush, is a stack of three risks rather than one.
The second component is the one almost nobody talks about, and it is unanimous. The panel is under the market on every single total-points rung, all nineteen of them, by as much as 5.5 points. The market's numbers imply this game lands somewhere between 42.5 and 43.5 total points; the panel's imply between 41.5 and 42.5. About a point lower, all the way up and down the ladder. The models' evidence for it is the rivalry itself: these teams have averaged 41.2 points a game since 2018, and just 33.7 across their last six meetings. Both defenses allowed fewer than 20 points a game last season. Week 1 games are sloppy. None of that is exotic, and all of it points the same direction.
Where The Panel Changed Its Mind
The revision round is where this gets interesting, because reading the other seats moved the panel toward the market, not away from it.
The two seats that came into the blind round highest on Kansas City both came down after reading the others, one of them by nearly seven points.
The group's median (KC ~47-48% to win, total ~41.5) matched my independent read, so I held my core position: Denver as a half-point road favorite rather than D/F's KC-by-2. D and F lean on Kansas City's long-run +6.6 home ledger, but that figure is built mostly on prime-Mahomes seasons; applying it at full strength to a Week 1 game with a QB twelve days from clearance off an ACL double-counts home field against QB risk. What the revision round genuinely changed: F and G's explicit mixture framing (85% Mahomes-starts world vs 15% Fields world) convinced me to widen my margin SD from ~12.5 to 13.5, since the bimodal QB outcome fattens the tails even when the mean barely moves … — Kimi
[Editor's note: that is the largest revision on this board, and the seat does not appear to know it. "D/F's KC-by-2" was Kimi's own opening position. It came into the blind round at Kansas City by 2 points and 55.2% to win the game, and left the revision round at Denver by half a point and 48.5%. Panel rationales are published as written, flagged rather than tidied.]
I moved toward the Forecaster G/F/D framing rather than the stronger Denver-leaning cluster: the evidence that Mahomes took first-team reps throughout camp and is tracking to start deserves real weight, while the card's instruction requires discounting two of Denver's recent series wins. I kept the total low because nearly every independent rationale identified the same under drivers: Week 1, Denver's pass rush, Kansas City's offensive-line transition, and recent low-scoring head-to-heads … — GPT, moving Kansas City from 56.3% to 51.0%
Then the seat that would not follow. Opus finished with the highest Kansas City number on the board, and it is the only one that answered the panel's Denver case head-on instead of absorbing it.
Held my center at KC +1.0 rather than moving to the group's slight Denver lean … The group's Denver case rests almost entirely on one number — 68 sacks vs 35 — and nobody prices the other side of the same trade: Denver's interior line is arguably in worse shape than Kansas City's (Deiter on IR and Wattenberg questionable means BOTH centers are compromised, plus Gargiulo out with an ACL), and Jonathon Cooper is Out, which trims the 2025 sack figure that is doing all the work. I also think "not cleared for game action" twelve days out is close to the modal state for a player who has practiced every camp day with the ones — I put ~90% on Mahomes starting, not the 85% several used … — Claude Opus
That is the honest counterargument to this entire article, from inside the panel, and it is worth holding onto. Denver's pass rush is the load-bearing wall of the Denver case, and one of the players who built it is currently listed as out.
What Would Change The Panel's Mind
This board gets re-scored when the story moves. These are the specific things that would move it, with the direction stated in advance rather than after the fact.
- Kansas City clears Mahomes for game action, or does not. He is practicing fully but had not been cleared as of September 2. Formal clearance would push the panel's Kansas City number up several points, toward the market. A Justin Fields start would pull it down hard. Either way the board gets re-scored.
- Josh Simmons' back injury. Kansas City's left tackle missed Wednesday's practice. If he is out, the single mechanism the models leaned on hardest gets worse for Kansas City, and the panel's Denver margin numbers move up.
- Jonathon Cooper's status, on the other side. Denver's pass rush is the Denver case. Cooper being out is already the strongest argument against it, exactly as Opus said. His return would firm up the panel's Denver numbers across the board.
- Denver's centers. Deiter is on injured reserve and Wattenberg is questionable. A team starting a third-choice center against Chris Jones, himself questionable with a back injury, is a different offense than the one that averaged 23.6 points a game.
| Date | What happens |
|---|---|
| Sat Sep 12 – Sun Sep 13 | Final Week 1 injury designations published for both teams |
| Mon Sep 14, ~6:45 PM ET | Inactives confirmed, roughly 90 minutes before kickoff |
| Mon Sep 14, 8:15 PM ET | Kickoff at GEHA Field at Arrowhead Stadium |
| Mon Sep 14, late | All 46 contracts settle on the final score |
| Tue Sep 15 | Every number on this page posts to the public grading scoreboard |
Prices on this page carry their September 2, 2026 stamp and are re-captured when the story moves.
The Widest Gap Isn't The Winner Market
Here is the line this whole board was hiding.
The winner market is the one everybody looks at, and the disagreement there is real but modest: 54.5¢ against 50.8%, under four points. Go down one row on the board, to the margin question that decides who wins by three or more, and the picture changes completely.
That contract is the most-traded margin question on the entire game, with over 11,600 contracts through it, more than three times the volume on the next-busiest margin rung. Three is the most common winning margin in football by a wide margin, showing up in 14.6% of all regular-season games, which is why this rung is where serious money on a game like this actually goes.
The market prices it like this: Kansas City to win by three or more, 49.5¢. Denver to win by three or more, 37.5¢. A twelve-point spread between the two sides. In the market's picture of this game, if it isn't close, Kansas City is meaningfully the more likely team to be the one pulling away.
The panel prices it like this: Kansas City 43.7%, Denver 43.8%.
A tenth of a point apart. Not "close," but functionally identical, from seven models that never saw either price and derived both numbers from the same margin distribution. On the question of who is more likely to win this game by a field goal or more, the market has a twelve-point favorite and the panel has a dead heat. That is a much larger disagreement than the winner market shows, on a contract with real money behind it, and it is the reason this board was worth building at all.
One more detail that sharpens it. The winner market has barely moved in three weeks. Kansas City closed at 56¢ on August 13 and closed at 56¢ today, and on no day in between did it close outside the 55-to-58 band, through three straight weeks of Mahomes practice reports. The margin rung went somewhere. It closed at 52¢ on August 19, 52¢ on the 20th and 52¢ on the 21st, and it closed at 48¢ today. The money is not changing its mind about who wins this game. It is quietly changing its mind about by how much.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $210M traded
- 2027 Pro Football Champion · $78M 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
The market looks at Broncos-Chiefs and sees a home team whose quarterback is probably fine, which makes Kansas City a modest favorite. The panel looks at the same game and sees two teams whose 2025 records badly misdescribed them, a pass rush that produced almost twice as many sacks as the other one, and an offensive line still learning its own calls. It lands on Denver's side of every line the market genuinely prices, and under the market on every total.
Both readings are defensible, and Opus made the best case against our own conclusion from inside the panel. What is not in dispute is where the disagreement is biggest, and it is not the market everyone is watching. It is one row down, at the number three, where the money sees a twelve-point favorite and seven machines see a coin flip.
Every one of these numbers gets graded in public once the contracts settle late on September 14, win or lose. Compare the venues yourself on our Broncos vs. Chiefs Week 1 odds page, which tracks the sportsbook side of this same game, and the full Week 1 AI board covers the other fifteen games.
