Last season the Dallas Cowboys scored 471 points, seventh-most in the NFL, and allowed 511, more than any other team in the league. They finished 7-9-1. Nobody in football was better at the first thing and worse at the second, and the two cancelled almost exactly: the scoring differential predicted 7.68 wins and Dallas banked 7.5.
Since late April, traders on Kalshi have been quietly walking that team up toward a winning season. The contract that pays if Dallas reaches 10 wins opened at 43 cents in April and has not closed below 52 since the middle of August. Nine AI models just priced the same 17-rung ladder without being shown a single one of those numbers, and they came back with a different answer at every rung that matters.
What Kalshi Has Been Doing To This Number Since April
The Dallas win-total ladder is 17 separate yes-or-no contracts on a CFTC-regulated exchange, one per threshold. The "10+ wins" contract pays a dollar if Dallas finishes the regular season with 10 or more wins and nothing if it does not, which means its price in cents reads directly as the market's probability. Across all 17 rungs the ladder has traded 244,048 contracts lifetime with 186,161 still open, so this is a real book, not a listing nobody touches.
The drift has been upward without being smooth. On April 24 the 10-win contract closed at 43¢. Through May it swung between 43 and 58. It has closed at 52 or better every day since August 18. Those are daily closing prices; the board further down quotes the midpoint of the live bid and ask, which is why the same contract reads 53.5¢ there and 54¢ here. The 9-win contract has gone further, from 51¢ in late April to 66.5¢ now. The 11-win rung moved from 34¢ to 38.5¢.
Then Monday happened. Dallas coach Brian Schottenheimer confirmed that left guard Tyler Smith, a three-time Pro Bowl selection, needs surgery on a torn ligament in his left thumb and is headed to injured reserve, which costs him a minimum of four games and probably six weeks. T.J. Bass, who has started 10 games in three seasons, takes his place at left guard.
The 10-win contract moved one cent on that news, closing at 54¢ Tuesday against 55¢ Sunday, on $2,657 of trading Monday and more than $4,500 in the 24 hours since. The 12-win rung took the bigger hit, closing 24¢ against 27¢. Traders decided a guard missing a quarter of the season shaves the ceiling and barely touches the middle.
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 Team That Scored Like A Contender And Defended Like Nobody Else
Strip out the record and the 2025 Cowboys were two separate teams sharing a bus.
The offense was legitimately excellent. Dallas gained 6,663 net yards, second in the NFL and 46 behind the Rams. Dak Prescott played all 17 games and threw for 4,552 yards with 30 touchdowns against 10 interceptions, a 99.5 passer rating. George Pickens caught 93 balls for 1,429 yards and nine scores. CeeDee Lamb added 1,077 in 14 games. Javonte Williams ran for 1,201 and 11 touchdowns at 4.8 a carry. Brandon Aubrey made 36 of 42 field goals including one from 64.
The defense was the worst scoring unit in the league, and the underlying numbers say that was not bad luck. Dallas finished 23rd in sacks with 35. It took the ball away 12 times all season, 30th in the NFL, and finished at minus-nine in turnover margin, 29th. A defense that cannot pressure the quarterback and cannot take the ball away has no cheap way to get off the field, and it showed up in games: 44 allowed at Denver in October, 44 at Detroit in December, and 34 in each of the last three losses.
That is the arithmetic the market is arguing with. Dallas converted 41.1% of its third downs and scored touchdowns on 56.9% of its red-zone trips, both perfectly respectable, and still lost nine games, because the other team kept scoring too.
The Defense They Went Out And Bought
The offense returns essentially whole. Prescott, Pickens, Lamb, tight end Jake Ferguson (82 catches, eight touchdowns) and Williams are all back. Every player who led the 2025 team in passing, rushing or receiving yards is on the roster; the only name off that leaderboard is backup quarterback Joe Milton III, now on the practice squad.
The defense was gutted and restocked. Checking each current Cowboy against the team he actually played for last season, 12 arrived from somewhere else. Eight of them carry real 2025 defensive snaps:
- Rashan Gary, edge, from Green Bay: 7.5 sacks and 45 tackles last season.
- Von Miller, edge, from Washington: nine sacks.
- Jalen Thompson, safety, from Arizona: 95 tackles and six passes defended.
- Cobie Durant, corner, from the Rams: three interceptions and seven passes defended.
- Dee Winters, linebacker, from San Francisco: 101 tackles and an interception.
- P.J. Locke, safety, from Denver: three passes defended.
- Otito Ogbonnia, tackle, from the Chargers: half a sack in 10 games.
- Jonathan Bullard, tackle, from New Orleans: 26 tackles.
The other four are corner Ameer Speed (two games in Houston), linebacker Tyrus Wheat (15 games in Detroit), running back Emari Demercado (312 rushing yards in Arizona) and tackle Broderick Jones, who played 11 games for Pittsburgh.
Add a full season of Quinnen Williams, who came over from the Jets midway through 2025 and played only seven games in a Dallas uniform, plus rookie safety Caleb Downs. Out the door go the team's sack leader Jadeveon Clowney (8.5), Osa Odighizuwa (3.5), leading tackler Kenneth Murray Jr. (81) and Donovan Wilson, whose two interceptions led the team.
One name that is not part of this offseason at all: Micah Parsons. He played 14 games for Green Bay last season and had 12.5 sacks doing it. His absence is already inside the 7-9-1, which turns out to matter enormously once the models start arguing. Dallas visits Green Bay in Week 6.
New to this kind of market? Our guide to betting NFL futures on Kalshi walks through how a season-long contract actually pays, and our Dallas Cowboys win total over/under page carries the sportsbook side of the same number.
The 15 Teams That Have Been Here Before
Since 2010, 15 teams have finished a season allowing more points than anyone else in football. Three of them reached a 10-win pace the following year. That is 20%, against a league-wide baseline of 35.6% across all 480 team-seasons in the same window. Widen it to every bottom-three scoring defense and the rate is 12 of 45, or 26.7%.
Those numbers argue hard against 53 cents. They are also the wrong reference class, and the reason is visible in the list itself. The 2020 Lions allowed 519 and went 5-11, then went 3-13-1. The 2016 49ers allowed 480 and went 2-14, then went 6-10. The 2021 Jets allowed 504 and went 4-13, then went 7-10. Those teams were not good at anything. They had no functioning quarterback and no offense to build on, and their next seasons went exactly where you would expect.
Condition on Dallas's actual shape instead. Take every team since 2010 that finished top-10 in scoring and bottom-three in points allowed in the same season, and six of 10 hit a 10-win pace the next year. Narrow it further to the ones that also finished with a losing record, as Dallas did, and it is three of five:
- 2012 New Orleans went 7-9 scoring 461 and allowing 454, then went 11-5.
- 2015 New York Giants went 6-10 scoring 420 and allowing 442, then went 11-5.
- 2016 New Orleans went 7-9 scoring 469 and allowing 454, then went 11-5.
- 2015 New Orleans went 7-9 with the same profile and went 7-9 again.
- 2010 Dallas went 6-10 and improved only to 8-8.
Five cases is not a law, and anyone quoting 60% off a sample that size is selling. But the direction is real and it is the direction the market has been trading all summer: a good offense stapled to a terrible defense is a fixable problem in a way that a bad football team is not.
The Board
We handed the whole ladder to nine AI models and let none of them see a price. Each one built a single distribution over Dallas's final win total and read all 17 rungs off it, then read the other eight models' reasoning anonymously and revised or held. The standing panel is eight seats and a ninth was added for this board. Eight of the nine came back with a second-round ladder; the ninth dropped out between rounds, so its first-round numbers stand. What follows is where they landed.
| Wins | Kalshi price | Panel | Gap |
|---|---|---|---|
| 5+ wins | 96¢ | 94.8% | -1.2 |
| 6+ wins | 93.5¢ | 90% | -3.5 |
| 7+ wins | 89¢ | 82% | -7.0 |
| 8+ wins | 75¢ | 71.5% | -3.5 |
| 9+ wins | 66.5¢ | 58.5% | -8.0 |
| 10+ wins | 53.5¢ | 41.6% | -11.9 |
| 11+ wins | 38.5¢ | 26% | -12.5 |
| 12+ wins | 25¢ | 15% | -10.0 |
| 13+ wins | 18¢ | 7.5% | -10.5 |
| 14+ wins | 13.5¢ | 3.2% | -10.3 |
| 15+ wins | 6.5¢ | 1% | -5.5 |
Prices are the midpoint of the live bid and ask on Kalshi as of September 8, 2026, 11:19 a.m. ET. The panel column carries model estimates, not predictions of fact and not financial advice.
Read down the gap column and the shape of the disagreement is obvious. It is not one rung. Every threshold from six wins up is priced above the panel's median, and the gap widens as the number climbs. Only two individual seats clear a market price anywhere on the ladder, both at the 8-win rung. Convert both ladders into a single expected win total and the market is paying for 9.67 wins while the panel's blended distribution centers on 8.88.
The widest gap in plain cents is the 11-win rung, 38.5¢ against 26%. The widest in proportion sits at the top of the ladder: 15+ wins costs 6.5¢ where the panel says 1%, and 14+ costs 13.5¢ where the panel says 3.2%. Not one of the nine seats put 14 or more wins above 4%.
Every Seat's Number
| Model | 8+ | 9+ | 10+ | 11+ | 13+ | Median wins | Wins from the rebuild |
|---|---|---|---|---|---|---|---|
| Fable | 70% | 56% | 40% | 26% | 7.5% | 9 | +1.5 |
| Opus | 71.5% | 58.5% | 43% | 29% | 9% | 9 | +1.3 |
| Sonnet | 66% | 51% | 35% | 21.5% | 5.3% | 9 | +1.3 |
| ChatGPT | 75.3% | 59.8% | 41.6% | 25.1% | 6.3% | 9 | +1.3 |
| Gemini | 76% | 64% | 48% | 32% | 9% | 9 | +1.3 |
| Grok | 73% | 58.5% | 45% | 30% | 8% | 9 | +1.8 |
| Kimi | 59% | 48% | 38% | 25% | 7.5% | 9 | +1.5 |
| GLM | 63% | 50% | 37% | 22% | 4.5% | 9 | +2 |
| DeepSeek | 74% | 60% | 45% | 30% | 9% | 9 | +1.5 |
| Panel median | 71.5% | 58.5% | 41.6% | 26% | 7.5% | 9 | — |
| Kalshi price | 75¢ | 66.5¢ | 53.5¢ | 38.5¢ | 18¢ | — | — |
Every seat on this panel is graded against real market settlements — records to date: Gemini 87% on 7,680 graded calls · Grok 85% on 6,396 graded calls · Kimi 86% on 3,836 graded calls · GLM 84% on 3,557 graded calls · DeepSeek 84% on 3,630 graded calls. Recomputed daily; the full scoreboard is public.
These are model estimates, not predictions of fact and not financial advice. Every number in this piece gets graded in public once the season settles; the running tally lives on the full graded scoreboard.
All nine seats put Dallas's median season at exactly nine wins. That unanimity is the finding, because it means the panel and the market broadly agree on where the middle of this team sits and disagree entirely about how much room is above it. The market's ladder implies a fat upper tail. The panel's does not.
The one seat closest to the market, Gemini at 48% on the 10-win rung, still sits five and a half cents under the price.
More live boards from the same panel: Giants win total 2026, Dallas's Week 1 and Week 17 opponent, where 8+ wins trades at 49¢. Patriots win total 2026: 10+ wins at 58.5¢, five cents clear of Dallas on the same threshold. Cowboys vs Giants odds for Sunday night's opener, with Dallas at 59.5¢ to win it. Prices fetched September 8, 2026.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $213M traded
- 2027 Pro Football Champion · $84M traded
- Pro Baseball Champion · $68M traded
- 2028 U.S. Presidential Election winner? · $65M traded
- 2028 Republican presidential nominee · $64M traded
- Fed decision in Sep 2026? · $38M 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
The second round produced one collective correction, and it was about Micah Parsons. Each model saw the others under its own shuffled set of letters, so the same forecaster is "F" to one seat and "E" to another; the quotes below are verbatim and all of them are pointing at the same model.
Two models built their first-round numbers on the idea that Dallas had just lost its best pass rusher. Gemini priced the defensive rebuild at negative half a win on that basis. Once the panel read each other's reasoning, five separate seats pointed at the same error.
"F double-counts Parsons, whose departure is already baked into the 2025 baseline stats in section 5, not a fresh 2026 subtraction." — Sonnet
"Forecaster E's −0.5 defense case was the strongest opposing argument but double-counts Micah Parsons, who was already absent in the 2025 7-9-1 baseline; netting him again is the wrong counterfactual." — Grok
Gemini took the correction and moved further than anyone:
"I revised heavily based on Forecaster F, who correctly noted Parsons' departure was already baked into the 2025 baseline. My round-1 defense penalty (−0.5) double-counted his absence." — Gemini
DeepSeek made the same repair on itself and raised its rebuild estimate from one win to one and a half. The panel's estimates of what the new defenders are worth converged from a spread of negative 0.5 to plus 2.5 into a band of plus 1.3 to plus 2.
Opus moved the other way, and its reason is the sharpest argument on the page against the market:
"I had shaded the 6/10 class from 60% to 46% but never charged Dallas for being rank 32 rather than merely bottom-three. That worst-defense list lands on a 7.4 pace again and again, and its only offense-rich member, 2015 New Orleans, repeated at 7-9. So 10+ goes 46 to 43." — Opus
Fable held its ladder unchanged and said why:
"My upside was never luck-based; it is Quinnen for 17 games instead of 7, Gary, and a takeaway floor of 12 that regresses regardless of personnel." — Fable
What Would Change The Panel's Mind
Four things, all checkable, each with the direction it pushes the number.
- Takeaways in the first quarter of the season. A total of 12 takeaways in 17 games is close to a statistical floor, and most of the panel's improvement comes from expecting that to correct. If Dallas is still under one takeaway a game through the Week 5 Tampa Bay game, the case for 10 wins goes with it. Pushes down.
- Tyler Smith's return date. He is out a minimum of four games. If he comes back on schedule around the Green Bay trip in Week 6, the offense's floor is intact. If the thumb lingers into November, the panel's 8-and-9-win mass thickens. Pushes down if delayed.
- The October road stretch. Houston in Week 4, Green Bay in Week 6, Philadelphia in Week 7. Those three teams went a combined 32-18-1 last season. A 2-1 record there and the panel's mass on 11 wins moves up toward the market's. An 0-3 and the ceiling is gone by Halloween. Pushes both ways, hardest.
- Whether the pass rush shows up against real offenses. Gary and Miller combined for 16.5 sacks last season on other teams. Gary is 28. Miller is 37. If that pair plus Quinnen Williams pushes Dallas from 35 sacks toward 45, the panel is too low. Pushes up.
Settlement Timeline
| Settles | January 18, 2027, at the end of the 17-game regular season |
| First data point | Sunday, September 13, 8:20 p.m. ET at the New York Giants |
| Tyler Smith's return window | Week 5 vs Tampa Bay (Oct 8) at the absolute earliest on his four-game minimum; Week 6 at Green Bay (Oct 18) is the realistic one on a four-to-six-week surgical timeline |
| The stretch that decides it | Three road games in four weeks: Houston (Oct 4), Green Bay (Oct 18), Philadelphia (Oct 26) |
| How it settles | Regular-season wins only. Ties do not count as wins, and Dallas played one last season. Postseason games are excluded. |
This page is re-scored when the story moves. Prices last read September 8, 2026, 11:19 a.m. ET.
The Bottom Line
Traders have spent five months pricing the Dallas Cowboys as a team that fixed itself. The offense certainly came back intact, the defense certainly got new names, and the turnover margin that helped bury 2025 is the kind of number that rarely stays that bad two years running. All of that is real, and it is why nine models that never saw a price still put this team's median season at nine wins rather than seven.
What none of them will pay for is the top of the ladder. The market says there is a 13.5% chance this team wins 14 games or more. The panel says 3.2%, and no seat on it goes above 4%. Somewhere between those two numbers is a question worth watching for 17 weeks: whether a defense that finished last in football can be rebuilt in one offseason, or only improved.
The first evidence arrives Sunday night in New Jersey.
Kalshi is a CFTC-regulated event exchange. Trading is 18+ and availability varies by state, as of September 2026. Nothing here is betting advice.
