Eight months ago the New England Patriots won the AFC East by two games, went 5-1 inside it, beat Houston in the divisional round at home, won the AFC Championship Game in Denver, and lost Super Bowl LX. Then they traded a 2028 first-round pick for A.J. Brown. Buffalo, meanwhile, fired the head coach who had won this division five years running.
The exchange where people put real money on the question has Buffalo as the favorite anyway, and it is not particularly close.
We handed eight AI models the whole division without letting any of them see a price. All eight came back with the board upside down.
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.
New to event contracts? How to trade NFL on Kalshi covers how these contracts price and settle before you read a board like this one.
These are market prices and model estimates, not predictions of fact and not financial advice. Kalshi is a CFTC-regulated event-contract exchange (18+; availability varies by state).
The Champion Nobody Will Back
Start with what New England actually did, because the market and the models are reading the same season and reaching opposite conclusions.
The Patriots went 14-3 with a point differential of plus 170, the best in the division and the best in the AFC. They scored 490 points, second in the league, and allowed 320, fourth. Underneath the standings the play-level ledger says the same thing louder: New England finished first of 32 teams in expected points added per play on offense, and first in EPA per dropback, which is the passing-game version of the same measure. Drake Maye was voted MVP runner-up behind Matthew Stafford, leading the NFL in completion percentage, yards per attempt, passer rating and QBR in his second season. Mike Vrabel won AP Coach of the Year and offensive coordinator Josh McDaniels won Assistant Coach of the Year, for a turnaround from back-to-back 4-13 seasons.
Then they went and got a receiver. New England acquired A.J. Brown from Philadelphia for a 2028 first-round pick and a 2027 fifth. Brown played for Vrabel in Tennessee from 2019 through 2021, which is part of why the Patriots were on his list.
One real caveat sits inside all of that, and every model on the panel found it. New England's offense was the league's best through the air and the 24th-best on the ground by rushing EPA per attempt, on a 7.8 percent sack rate. That is the more variance-exposed shape of an elite offense: it is a great passing team, not a great running team, and it now carries the tax every team pays the season after a Super Bowl. The schedule makes that point on the first night. New England opens at Seattle on Wednesday, September 9, against the team that beat it 29-13 in February, in the first Wednesday game in franchise history.
Buffalo Fired The Man Who Won It Five Times
Two days after losing 33-30 in overtime at Denver in the divisional round, Buffalo fired Sean McDermott after nine seasons. His Bills had won the AFC East in each of the five years from 2020 through 2024; 2025 was the first season since 2019 they did not. Eight days later the Bills promoted offensive coordinator Joe Brady to head coach on a five-year deal. It is his first NFL head-coaching job.
The bull case for Buffalo is one name and one number. Josh Allen is 30 and entering his ninth season, and Buffalo's offense finished third in the league in EPA per play and second in rushing success rate, the highest floor in the division. Brady was Allen's own coordinator, so the offense the Bills are keeping is the offense that produced those numbers, which is more continuity than "first-time head coach" implies. They also sent a second-round pick to Chicago for DJ Moore and a fifth, a receiver who played under Brady in Carolina and who is coming off a career-low 682 receiving yards.
The bear case is more specific than "new coach." Buffalo's defense ranked 31st of 32 in rushing EPA allowed last season, and nothing in the offseason record establishes that the leak has been closed. Its 2026 calendar is also the harder one: seven standalone national windows against New England's four, two four-day weeks, and a Christmas Day game at Denver. New England drew the easiest slate in the division measured by opponents' 2025 net EPA per play, and gets extra rest before four separate games.
The two teams split last season, and they split on the road. New England won 23-20 at Buffalo on October 6; Buffalo won 35-31 at New England on December 14. Because they split, the division record that separated them came from the other games. Buffalo lost 30-13 at Miami in November. The Bills finished 4-2 inside the division; the Patriots went 5-1.
What 24 Seasons Of This Division Say
The AFC East has produced four different champions since 2002, and the defending champion has repeated 18 times in 23 chances. That is 78 percent, and by the end of the panel's second round all eight models had cited it.
The number needs one correction before you lean on it, and the correction is not the one you would expect. All 18 of those repeats belong to exactly three runs: New England 2003-2007, New England 2009-2019, and Buffalo 2020-2024. Take only the champions who had just taken the division away from somebody else, which is New England's exact situation now, and the sample shrinks to four resolved cases. Three of those four went on to win it again the following season. New England took the division in 2003 and held it the next year, then took it back in 2009 and held that one too. Buffalo took it in 2020 and held it. Miami took it in 2008 and lost it straight back. Three of four is 75 percent. Four cases is not a statistic and should be read as one. But unlike most base rates that get conditioned this way, this one barely moved.
There is a fact that cuts the other way, and two seats reached for it and named the wrong pair. Claude Sonnet wrote that this title had "changed hands to a bad prior-year team only twice in 24 years (2002 NYJ, 2008 MIA)." Kimi put the same idea as "a sub-.500 prior-year team has won this division twice in 24 years." The count is right and one of Sonnet's two cases is not: the 2002 Jets came off a 10-6 season. Check every champion in the window against its own prior year and the two are Miami in 2008, off 1-15, and New England last season, off 4-13. Miami did not repeat. New England is the second case, and it has not resolved.
The Board
The AI blend is the median of eight independent models after a revision round in which each one read the others' reasoning without knowing which model wrote what.
| Team | 2025 record and point differential | Kalshi | AI blend |
|---|---|---|---|
| Buffalo Bills | 12-5, +116 | 56.0¢ | 37.0% |
| New England Patriots | 14-3, +170 | 39.5¢ | 56.5% |
| New York Jets | 3-14, -203 | 3.5¢ | 3.8% |
| Miami Dolphins | 7-10, -77 | 2.5¢ | 3.0% |
This is not a disagreement about the top two. It is the same two numbers with the names swapped. The market pays 56 cents for Buffalo, where the panel's middle answer is 37. It pays 39.5 cents for New England, where the panel's middle answer is 56.5.
How To Read This Board
A price in cents is roughly the market's implied chance. Buffalo at 56¢ means traders collectively give the Bills about a 56 percent chance. Every contract that resolves yes pays exactly $1, so in payout terms $100 on Buffalo comes back as about $179 if the Bills win the division, and nothing if they do not. The same $100 on New England at 39.5¢ comes back as about $253, on the Jets at 3.5¢ about $2,857, and on Miami at 2.5¢ about $4,000.
Because the NFL always names exactly one division champion, these four contracts cover every outcome between them, and the league's published tiebreaker sequence starting with head-to-head decides any tie. They are still four separately quoted markets, so the midpoints need not total exactly 100 cents. Here the midpoints sum to 101.5. Buying every side would actually cost the four asks, which come to 104 cents, so a full sweep of this board is 4 cents underwater before anything happens. The blend column is a median of eight separate estimates rather than one distribution, so it need not total 100 either: as displayed it sums to 100.3, and 100.25 before the Jets number is rounded.
The book is real on the top two: Buffalo is quoted 55 bid, 57 ask, and New England 39 bid, 40 ask. The four contracts have traded about 245,900 between them, with roughly 196,900 still open. The share is lopsided in a way worth noticing. Buffalo's contract has traded 104,400 of that, the most on the board. New England's has traded 36,500, the least. The team the panel wants is the one the fewest people have bothered to price.
More live boards from the same panel: AFC North odds, where Baltimore leads at 46.5¢ and Cincinnati sits at 33.5¢ · AFC South odds, where Houston is out to 48.5¢ over a Jacksonville team that won 13 games · AFC West odds, where Kansas City, Denver and the Chargers sit inside two cents of each other. Prices fetched September 5, 2026.
Every Seat's Number
| Model | Buffalo | New England | NY Jets | Miami |
|---|---|---|---|---|
| Claude Fable | 37.0% | 55.0% | 3.5% | 4.5% |
| Claude Opus | 37.0% | 54.0% | 5.0% | 4.0% |
| Claude Sonnet | 37.0% | 56.0% | 4.0% | 3.0% |
| GPT | 36.0% | 57.0% | 4.0% | 3.0% |
| Gemini | 38.0% | 57.0% | 3.5% | 1.5% |
| GLM | 37.0% | 57.0% | 3.0% | 3.0% |
| Kimi | 38.0% | 57.5% | 2.0% | 2.5% |
| DeepSeek | 38.0% | 55.0% | 4.0% | 3.0% |
| Median | 37.0% | 56.5% | 3.8% | 3.0% |
Every seat on this panel is graded against real market settlements — records to date: Claude Fable 89% on 1,833 graded calls · Claude Opus 89% on 1,920 graded calls · Claude Sonnet 87% on 1,897 graded calls · GPT 85% on 8,328 graded calls · Gemini 87% on 7,014 graded calls · GLM 83% on 3,179 graded calls · Kimi 86% on 3,412 graded calls · DeepSeek 83% on 3,260 graded calls. Recomputed daily; the full scoreboard is public.
Model estimates generated September 5, 2026, price-blind from live fetched data. These are model estimates, not predictions of fact and not financial or trading advice. Models are frequently wrong; the market price reflects real traders' money. Kalshi is a CFTC-regulated exchange; 18+, availability varies by state.
Every number in this piece gets graded in public once the market settles, and the full graded scoreboard carries each seat's running record. All eight seats returned a board for this market.
Read across the New England column and the range is three and a half points, from 54 to 57.5. Eight models built eight separate views of this division and finished inside that band. The agreement is the finding, and so is what it is agreeing against: a market that has the same team 17 cents lower.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $211M traded
- 2027 Pro Football Champion · $81M traded
- Pro Baseball Champion · $67M traded
- 2028 U.S. Presidential Election winner? · $64M traded
- 2028 Republican presidential nominee · $63M traded
- Fed decision in Sep 2026? · $35M 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.
Where The Panel Changed Its Mind
After round one, each model read the other seven arguments anonymously and could revise. Nobody switched sides, but the two seats furthest from the group both walked back toward it, from opposite directions, and they met almost exactly in the middle.
Claude Sonnet started lowest on New England at 47 and finished at 56, the largest move on the board. Its first round treated the top two as nearly even on talent. Its second round did not change that reading and changed the conclusion anyway:
"But Buffalo's underlying numbers were close, not distant: #3 offense EPA (0.140 vs 0.157), #2 rushing offense, and a defense that was actually better than New England's against the pass (dropback EPA allowed -0.068 vs -0.038). Josh Allen at 30 remains the highest floor QB in the division. What pulls me off a near-even split is concrete and non-record-based: Buffalo fired the coach who won five straight division titles and replaced him with a zero-experience Joe Brady, and its 2026 calendar is materially harder."
— Claude Sonnet
DeepSeek came the other way, cutting New England from 62 to 55 while raising Buffalo from 30 to 38, and Claude Fable trimmed New England from 60.5 to 55. Fable spent its revision attacking the statistic the rest of the panel was leaning on:
"The 78% repeat rate is mostly Brady-era and McDermott-era dynasty seasons; it is weak evidence for a first-time repeat attempt in a division with two .700 teams… And New England's edge is dropback-driven (#1 pass EPA on a #24 rush offense, 7.8% sacks taken) — the more variance-exposed profile, in a post-Super Bowl season with a Munich trip."
— Claude Fable
The seat that moved least is the one that priced New England lowest at the end. Claude Opus went 57 to 54 and spent both rounds on the arithmetic under the record. A Pythagorean win total is simply the record a team's points scored and allowed would normally produce, and both of these teams beat theirs:
"New England's +170 over 17 games implies roughly 11.5-12 Pythagorean wins, so 14-3 outran its own differential by about two games; Buffalo's +116 implies about 10.5-11, outrunning 12-5 by roughly one… That is a one-win talent gap dressed up as a two-win standings gap. So I will not price NE near 60."
— Claude Opus
Kimi, the highest seat at the end, put the whole season on the two head-to-head meetings:
"New England wins by holding serve in its four games against Miami/New York, at least splitting Buffalo (Week 4 at BUF, Week 13 at home with extra rest), and letting Maye-to-Brown cash the easiest non-division slate into a one- or two-game cushion."
— Kimi
[Editor's note: the extra rest in that Week 13 meeting belongs to Buffalo, not New England. The Bills play Kansas City on a Thursday in Week 12 and then travel to Foxborough 10 days later on December 6; the Patriots come off a normal week following a Sunday night game at the Chargers. New England's own four extra-rest games are Weeks 2, 8, 12 and 15. The rest of the sentence holds: Week 4 is at Buffalo, Week 13 is at New England.]
What Would Change The Panel's Mind
Four triggers, each one checkable, drawn from what the seats named themselves.
- A Multi-Week Injury To Drake Maye Or Josh Allen. Every one of the eight named this first. Kimi sized it: Allen missing time pushes Buffalo toward 20 and New England toward 70, because neither backup room, Tommy DeVito and Behren Morton in Foxborough or Kyle Allen in Buffalo, is close to a division-race quarterback.
- The Week 4 Result At Buffalo, On October 4. The tiebreaker chain starts with head-to-head, so this game carries more than a win. Claude Opus put a Buffalo win plus a 3-1 start at moving the Bills toward 45; Claude Fable put a New England road win at pushing the Patriots above 60.
- Evidence Buffalo's Run Defense Has Been Rebuilt. Opus asked for the specific thing: a real front-seven overhaul or a healthy interior returning moves Buffalo up four to six points. Confirmation that the 31st-ranked leak is unchanged moves New England up, because the Patriots get two cracks at it.
- New England's Offensive Line Through September. Gemini flagged the 7.8 percent sack rate as the tell rather than the outcome. Early breakdowns behind that number lower New England and raise Buffalo; DeepSeek framed the same trigger as early-season regression from Maye.
Dark Horses The Panel Won't Dismiss
- New York Jets (3.5¢, Blend 3.8%): the panel is a fraction above the market and, more interestingly, above Miami, despite an eight-game gap in last season's records. The Jets were the worst team in football by point differential at minus 203 and last of 32 in EPA per dropback allowed. What moves them is the floor, not the ceiling: the Jets traded for Geno Smith ahead of his expected release in Las Vegas and named him the starter, with Frank Reich installed as offensive coordinator, in what is widely reported as a make-or-break second year for Aaron Glenn.
- Miami Dolphins (2.5¢, Blend 3.0%): the cheapest contract on the board, and the panel agrees with the price for a reason the standings hide. Miami won seven games last season. It then released Tua Tagovailoa in March, absorbing more than $54 million in dead money, and moved on from Tyreek Hill, Jaylen Waddle, Minkah Fitzpatrick and Bradley Chubb. Malik Willis starts. New head coach Jeff Hafley has said publicly that 2026 will not be measured in wins and losses, and Miami drew the toughest slate in the division. Six of eight seats put a 7-10 team below a 3-14 one on the strength of that.
- The Precedent Nobody Wants To Price: this exact contract paid out once on a team like these. Miami won the AFC East in 2008 the year after going 1-15. It is the only time in 24 seasons that a team went from a losing record to this division title in one step, until New England did it last season.
Settlement Timeline
| Date | What happens |
|---|---|
| September 9, 2026 | Week 1 opener. New England at Seattle, 8:20 PM ET, a Super Bowl LX rematch and the first Wednesday game in Patriots history |
| September 13, 2026 | Week 1. All three of the others open on the road: Buffalo at Houston, the Jets at Tennessee, Miami at Las Vegas |
| October 4, 2026 | Week 4. New England at Buffalo, the season's first division game and the first tiebreaker |
| December 6, 2026 | Week 13. Buffalo at New England, the rematch, with the Bills on 10 days of rest |
| January 10, 2027 | Week 18. Miami at New England and the Jets at Buffalo close the regular season |
| January 11, 2027 | Expected expiration. Contracts can close earlier once a champion is declared |
This board gets re-scored when the story moves, and prices on this page refresh during the season. Prices and panel numbers here are as of September 5, 2026.
The panel's calls are on the record. Our 8-model AI panel puts a graded, public verdict on the conference this division feeds into. See the live model verdict on the AFC Championship, where Buffalo leads at 15.5¢ and New England trades sixth at 9.5¢, or the full NFL board index for every market we track this season.
- NFL 2026 on Kalshi: Every Board We Track
- AFC Championship 2027 Odds: The Champion Nobody Is Paying For
- AFC North Odds: The Right Favorite At The Wrong Price
- AFC South Odds: The 13-Win Champion Isn't The Favorite
- AFC West Odds 2026: The 14-3 Champion Trades Third
The Bottom Line
Both sides of this board are reading the same two facts and weighting them opposite ways. New England was better than Buffalo last season on record, differential, points scored, points allowed, division record, offensive EPA and postseason result, and then added the best receiver either team acquired. Buffalo has Josh Allen and a first-year head coach.
The market has decided that Allen outweighs all of it, and that is a defensible read with 15 years of this division behind it. The panel, shown the same season and no prices at all, could not find the argument. Its widest disagreement is not with the Buffalo case; it is with the size of the discount attached to the team that just played in the Super Bowl.
If the models are right, the 39.5-cent contract is the one to have looked at in September. If the market is right, the first evidence arrives on October 4 in Orchard Park. Head-to-head is the first tiebreaker the NFL applies, and it is not decided until these two play again in December.
Hero illustration: OddsShopper, in the house collage style. Drake Maye photo by Tennessee Titans, licensed CC BY 3.0; Josh Allen photo by Erik Drost, licensed CC BY 2.0; photos cropped, toned, and composited.
