Rico Dowdle is out. Cleveland's starting center is out. Pittsburgh traded a starting cornerback to Dallas on the eve of the game.
So we did the obvious thing that most simulations skip: we rebuilt Steelers–Browns on the players who are actually suiting up tonight, handed out the football the way these two teams have actually handed it out this season, and played the game 10,000 times.
Then we lined every catch and carry contract up three ways — the Kalshi ladder, our own model, and the roster-true simulation — and went looking for the places where they don't agree.
The Calls
Here is what the simulation says to do, after checking every read against outside projections and the books (next section). These are simulation reads, not published positions — the sim is a new harness and has not yet been graded the way our published board is. Every price is a Kalshi two-sided mid.
| Contract | Side | Kalshi | Our model | Roster-true sim | Outside check | Read |
|---|---|---|---|---|---|---|
| Jerry Jeudy 2+ catches | NO | 42.5¢ YES | 57.0% | 31.9% | Books lean under; usage agrees | Strongest read |
| Roman Wilson 3+ catches | YES | 38.5¢ | 42.4% | 62.3% | Targets agree; role at risk | Lean — only if he keeps the job |
| Michael Pittman Jr. 4+ catches | — | 43.5¢ | 60.9% | 38.5% | Books lean under | Pass. Our model's YES does not survive. |
| Jaylen Warren 3+ catches | — | 68¢ | 49.9% | 66.5% | Outside 3.1–3.8 catches | Pass. The market has his role right. |
| Warren and Quinshon Judkins carries | — | 63.5¢ / 47.5¢ | 19.4% / 19.8% | 51.9% / 23.1% | Outside 16–19 carries each | Withdrawn. Our number is the outlier. |
Why Jeudy is the cleanest read. Our model projects Jeudy for 4.7 targets because of what he has done in past seasons. In 2026 he has 5 targets total — four in Week 1, one in Week 2, none in Week 3 — while his snap share slid from 63% to 51%. Denzel Boston, KC Concepcion and Harold Fannin Jr. have taken the passing game, and Cleveland's beat coverage says he has fallen behind the rookies. The books agree the usage is real: the over on 1.5 catches is plus money at both BetMGM and DraftKings, roughly 43% once the margin is stripped — right where Kalshi sits. The sim goes further, to 32%.
Why Wilson is a lean, not a lock. He has been targeted six times in every game this season, with Pittman in the lineup and without him — the books' own analysts cite the same number. Our model projects 3.2 targets because it still sees him as a third receiver; the sim gives him 6.6. But the outside check turned up the risk the targets hide: Wilson played 24 snaps to Germie Bernard's 17 in Week 3, and after his turnover Bernard took 15 snaps to Wilson's 5, per 93.7 The Fan. FantasyPros' consensus has him at 2.4 catches. If he keeps the job, 38.5¢ is cheap. If Pittsburgh is done waiting, it isn't — and the same read runs the other way on Bernard, whose 2+ catches sits at 35.5¢.
Why we withdrew the carries. In the first draft of this piece both starting running backs were watch-list unders, because our model and the sim agreed against the market. The outside check killed it. FantasyPros and Covers project Warren for 17.8 to 18.8 carries and Judkins for 16.3 to 18.2, against book lines of 17.5 and 16.5. When everyone outside our building agrees with the market and only we disagree, the likeliest explanation is us — here, a carry distribution we have never fitted and no adjustment for game script.
Second Opinion: The Outside Check
We ran every read past outside projections, the books and other public models before publishing. Where they agree with the sim, it stayed. Where they didn't, it moved.
| Player | Our sim | Our model | Outside | Who they back |
|---|---|---|---|---|
| Jerry Jeudy | 1.7 targets, 0.8 catches | 4.7 targets, 2.2 catches | 5 targets in 3 games; books lean under 1.5 catches; FantasyPros 2.0 catches | Sim on usage and price |
| Michael Pittman Jr. | 4.4 targets, 3.1 catches | 7.4 targets, 5.2 catches | 8 targets in 2 games; 10.7% target share (FantasyPros); books lean under 3.5 catches (Covers) | Sim |
| Roman Wilson | 6.6 targets, 3.8 catches | 3.2 targets, 1.9 catches | 6 targets every game; FantasyPros 2.4 catches; book 2.5 catches, over +140 | Split — sim on targets, outside lower on catches |
| Germie Bernard | 0.5 targets | — | FantasyPros 1.2 catches; 8 targets in Week 2, 1 in Week 3 | Sim is too low if Wilson's role shrinks |
| Jaylen Warren | 15.9 carries, 4.3 catches | 12.5 carries, 3.0 catches | 17.8–18.8 carries, 3.1–3.8 catches; book 17.5 carries | Market — sim closer than the model, still light |
| Quinshon Judkins | 13.9 carries | 13.4 carries | 16.3–18.2 carries; book 16.5 | Market — we are the outlier |
| Harold Fannin Jr. | 4.2 catches | 4.1 catches | 4.4–5.1 catches; book 4.5 | Market, roughly |
On the game, outside models land around ours. Dimers has Pittsburgh at 60.0%, numberFire at 53.4% and BetMGM's model at 54.8%, against our 58.8%. SportsLine's own 10,000-game run projects 37 total points, per CBS Sports, a point and a half under the market — we still don't publish totals.
The lesson from the check is the useful part. Real-life usage fixed our model's pass-catcher numbers, and the outside world agrees. It did not fix our carries, and the outside world says so too.
The Roster Sheet
This is what "tonight's real rosters" means, player by player. Statuses are from the final injury report filed Wednesday; inactives post about 90 minutes before kickoff.
| Player | Team | Status | What the simulation does |
|---|---|---|---|
| Rico Dowdle (RB) | PIT | Out (toe), second straight game | Removed. His carries go to the running backs who are left — overwhelmingly Warren, as in Week 3 (17 of 19 running-back carries). His 7 targets are reassigned across the receiving room. |
| Joey Porter Jr. (CB) | PIT | Traded to Dallas | No change — he had not played a snap in 2026, so nothing in this season's numbers included him. |
| Jalen Ramsey (S/CB) | PIT | Questionable — broken wrist, playing in a soft cast | Kept in. See "What the simulation doesn't know." |
| Brandin Echols (CB) | PIT | Questionable (concussion) | Kept in. |
| Elgton Jenkins (C) | CLE | Out — did not clear concussion protocol | Out. Luke Wypler starts at center. |
| Teven Jenkins (G) | CLE | Out (back), fourth straight game | Out — and already absent from every Cleveland snap this season. |
| Tylan Wallace (WR/KR) | CLE | Out (knee) | Removed. He had no targets in 2026; his work was on returns. |
| Tyson Campbell, Grant Delpit, Tytus Howard | CLE | Full practice, no designation | Playing. |
Both quarterbacks are confirmed. Aaron Rodgers is not on the injury report. Deshaun Watson remains Cleveland's starter, with Shedeur Sanders the backup.
Why Our Model and the Simulation Disagree
The two harnesses use the exact same distribution for how a player's night varies around his expected number. What they disagree about is the expected number — specifically, how much of the ball each player gets.
| Player | Model targets | Sim targets | 2026 so far |
|---|---|---|---|
| DK Metcalf | 7.4 | 8.2 | 24 targets in 3 games |
| Roman Wilson | 3.2 | 6.6 | 6, 6, 6 |
| Jaylen Warren | 3.8 targets, 12.5 carries | 5.5 targets, 15.9 carries | 17 carries in the game Dowdle missed |
| Pat Freiermuth | 4.4 | 4.9 | 14 targets |
| Michael Pittman Jr. | 7.4 | 4.4 | 3 and 5 targets in his two games |
| KC Concepcion | 7.4 | 6.6 | 20 targets |
| Harold Fannin Jr. | 5.9 | 6.0 | 3, 6, 9 |
| Denzel Boston | 6.5 | 5.3 | 16 targets |
| Jerry Jeudy | 4.7 | 1.7 | 4, 1, 0 |
| Isaiah Bond + Malachi Corley | 4.8 combined | 0.3 combined | 1 target combined |
That last row is the whole story in one line. Our model is still feeding almost five targets a game to two Browns receivers who have played under 15% of snaps this season. Those targets have to come from somewhere, and in real life they have gone to Concepcion, Fannin and Boston.
The Warren lesson
The single biggest number on this page is a mistake — ours.
When Dowdle was ruled out, our projection file set his carries to zero and did not hand them to anyone. That left Pittsburgh projected for 13 total rushing attempts, which is not a real NFL game plan, and left Warren at 12.5 carries. Against a Kalshi price of 63.5¢ for 16+ carries, our model said 19.4%. That looks like a 44-point "edge." It is a bug.
Give Warren back the work he actually did in Week 3 and the sim lands at 15.9 carries and 51.9% for 16+. That is still about 12 points under the market — and outside projections of 17.8 to 18.8 carries say even the sim is light, most likely because it doesn't know Pittsburgh will run the clock with a lead. But the 44-point gap is gone. His catches tell the cleaner story: 3+ receptions is 68¢, the sim says 66.5%, and our model's 49.9% disappears. On Warren's role, the market was right and our model was wrong, and the reason is a roster that changed and a projection that didn't.
That is also why we built this. A projection that is one injury behind can manufacture a "strong" disagreement out of nothing.
The Game
The game simulation does not carry a new opinion on who wins. It is centred on the market — Steelers by 2.5 across a nine-book consensus — and drawn on our fitted key-number lattice, the part of the model that knows football scores land on 3 and 7 far more often than on 2 and 5.
| Result | |
|---|---|
| Steelers win | 58.8% |
| Browns win | 40.5% |
| Tie | 0.7% |
| Steelers −2.5 covers | 53.6% |
| Steelers −3.5 covers | 45.7% |
| Steelers win by exactly 3 | 7.9% |
| Browns win by exactly 3 | 8.2% |
| Average score | PIT 20.6, CLE 18.1 |
The single most likely final margin is the Browns by exactly 3. That sounds strange for a game Pittsburgh is favored in, but it is how home field shows up in real football — home teams win a lot of games by a late field goal. Kalshi agrees: read straight off its spread ladder, Browns-by-exactly-3 is priced at 8.0%, against our 8.2%.
Where the ladder and our simulation part ways is the other field goal. Kalshi's ladder puts 10.0% on Pittsburgh winning by exactly 3 — Browns +3.5 at 57.5¢ against Browns +2.5 at 47.5¢. Our simulation says 7.9%. That makes Browns +3.5 look about three cents rich to us. It is inside the five-point band we call priced, so it is a curiosity, not a call.
Our ratings-based game lean is switched off this week. Our power ratings have been leaning toward underdogs more than they should, and we are fixing that before we publish another one. Until then, the game numbers above are the market's centre with our shape on it — nothing more.
The total is a pass. The market sits at 38.5, and the sim, centred there, puts 49.4% of games under it. The only totals signal we price is an outdoor-wind adjustment, and it starts above 10 mph; the National Weather Service has about 10 mph from the south at kickoff, with gusts forecast anywhere from 17 to 22 mph and rain chances rising late, so it adds nothing. Our totals model was retired from publication on September 16 because it does not beat the market.
The Full Three-Way Board
Every catch and carry contract where the market sits between 30¢ and 70¢ — the band where our published numbers are measured to be calibrated. A gap counts as a disagreement at 5 points or more.
| Player | Contract | Kalshi | Model | Sim | Who disagrees |
|---|---|---|---|---|---|
| Roman Wilson | 2+ catches | 63.5¢ | 54.8% | 79.7% | All three |
| Roman Wilson | 3+ catches | 38.5¢ | 42.4% | 62.3% | Sim alone |
| Jerry Jeudy | 2+ catches | 42.5¢ | 57.0% | 31.9% | All three |
| Germie Bernard | 2+ catches | 35.5¢ | — | 12.0% | Sim alone |
| Michael Pittman Jr. | 3+ catches | 64.5¢ | 73.9% | 51.1% | All three |
| Michael Pittman Jr. | 4+ catches | 43.5¢ | 60.9% | 38.5% | Model alone |
| Jaylen Warren | 3+ catches | 68.0¢ | 49.9% | 66.5% | Model alone |
| Jaylen Warren | 4+ catches | 49.5¢ | 38.3% | 53.5% | Model alone |
| Jaylen Warren | 5+ catches | 30.5¢ | 29.5% | 43.3% | Sim alone |
| Jaylen Warren | 16+ carries | 63.5¢ | 19.4% | 51.9% | Model and sim under |
| Jaylen Warren | 19+ carries | 40.5¢ | 5.1% | 24.6% | Model and sim under |
| Quinshon Judkins | 2+ catches | 62.5¢ | 63.1% | 62.3% | Agree |
| Quinshon Judkins | 3+ catches | 38.5¢ | 47.0% | 45.5% | Model and sim over |
| Quinshon Judkins | 14+ carries | 69.5¢ | 46.4% | 52.1% | Model and sim under |
| Quinshon Judkins | 17+ carries | 47.5¢ | 19.8% | 23.1% | Model and sim under |
| Raheim Sanders | 2+ catches | 56.5¢ | 48.3% | 59.1% | Model alone |
| Raheim Sanders | 3+ catches | 34.5¢ | 34.7% | 41.1% | Sim alone |
| Denzel Boston | 3+ catches | 56.5¢ | 60.6% | 49.7% | Sim alone |
| Denzel Boston | 4+ catches | 35.5¢ | 46.9% | 37.4% | Model alone |
| Harold Fannin Jr. | 4+ catches | 64.5¢ | 51.9% | 52.7% | Model and sim under |
| Harold Fannin Jr. | 5+ catches | 46.5¢ | 41.5% | 42.5% | Agree |
| Harold Fannin Jr. | 6+ catches | 30.5¢ | 33.2% | 34.5% | Agree |
| DK Metcalf | 4+ catches | 53.5¢ | 50.9% | 55.0% | Agree |
| DK Metcalf | 5+ catches | 35.5¢ | 40.8% | 44.7% | Model and sim over |
| Pat Freiermuth | 3+ catches | 62.5¢ | 57.7% | 61.5% | Agree |
| Pat Freiermuth | 4+ catches | 39.5¢ | 43.8% | 47.5% | Sim alone |
| Darnell Washington | 2+ catches | 59.5¢ | 59.1% | 58.2% | Agree |
| Darnell Washington | 3+ catches | 36.5¢ | 41.9% | 40.5% | Model alone |
Two patterns are worth more than any single row. When our model disagrees with the market alone, the roster usually explains it — Pittman 4+, Warren's catches, Boston 4+ and Raheim Sanders 2+ all fall back toward the market once the ball is handed out the way it has been. When the sim breaks away on its own, it is because of something that has happened this season that a prior-heavy projection hasn't caught up to — that is Roman Wilson. And when model and sim agree against the market, it is carries — which is exactly where outside projections side with the market instead.
What We Refused To Publish
- Every yardage contract. Across this week's full slate our model's raw numbers lean under on about 70% of yardage strikes — rushing yards a median of 14 percentage points below the Kalshi price — which is a model-level bias, not hundreds of independent edges. Rushing, receiving and passing yards are already held from our ladder for that reason. The sim inherits the model's yards-per-catch and yards-per-carry, so it inherits whatever is wrong there. No yardage reads here.
- Anytime touchdowns. Not what this model is built to price.
- A game lean or a totals position, for the reasons above.
- Raheim Sanders' rushing yards. Our model projected him for 26 rushing yards on 4.9 carries; he has 4 carries all season. That produced a 40-point "edge" on 15+ yards against a 20.5¢ price — a projection error, not a market error. On his real usage the sim projects him for about 7 rushing yards.
What The Simulation Doesn't Know
- Three games is three games. Shares and volume come from Weeks 1 through 3. For Pittsburgh's passing game we used only the two games Pittman played, because his presence is the biggest thing that moves who gets targeted. That is a small sample, stated plainly.
- The defensive injuries aren't priced. Jalen Ramsey's wrist, Echols' concussion and Cleveland's missing center and guard change the game, but our player projections are built from the offense's side, and our team-level injury adjustment only covers quarterbacks and skill players. The game simulation sidesteps this by centring on the market, which does price them.
- No game script. A team that falls behind throws more. The sim does not shift anyone's volume to fit the score it is simulating.
- No fitted carry distribution. Carries use a plain Poisson spread, which is tighter than real life.
- Rodgers and Watson pass attempts, completions and interceptions are not modelled and are not in the tables.
- Inactives aren't out yet. If Ramsey sits, Echols moves into the slot. If any receiver above is a surprise inactive, the targets move and so does every number on this page.
The distribution is what the engine can tell you. The rest is why it might be wrong.
Game: Pittsburgh at Cleveland · Thursday, October 1, 8:15 p.m. ET · Huntington Bank Field, Cleveland · matchup page
Board: Kalshi two-sided mids snapshotted around 12:10 p.m. ET, October 1, 2026. Game anchor: nine-book consensus, Steelers −2.5, total 38.5.
Method: 10,000 simulations, seed 20261001. Model column = engine run 2026w04-20261001T120150Z-fc8698, priced through the same zero-inflated lognormal and per-stat calibration the published board uses; our reproduction matches the shipped closed form to a median of 0.5 points (maximum 3.2) on receptions. Sim column = same process, volume and shares from 2026 play-by-play with tonight's inactive players removed and their work reassigned. Monte Carlo standard error at a coin flip: 0.5 points.
The live board moves. Check the ladder before you do anything with a number on this page.
Keep Going
- Steelers vs Browns Week 4 — the matchup page, every ladder and the trenches
- Every Kalshi prop on this week's slate, priced against the projection
- The Ladder — where our number and the market disagree across every strike
- Build a Steelers–Browns combo — same-game legs priced on their true joint odds
- Lions–Bills, simulated 10,000 times through Madden 27 and our model
- Patriots–Seahawks, simulated 100,000 times — the series precedent
- NFL win probability calculator · Week 4 player projections