The Model Fought Kansas City and San Francisco for Two Months. After Week 3 It Surrendered.

In July our model looked at Kansas City and saw an aging dynasty with a leaking defense, and priced it as a losing team. It looked at San Francisco and saw famous names in front of a defense graded 30th, and faded the 49ers harder than any contender in the league. Both calls argued with the market, and both have gone the market's way. Kansas City is 3-0 and projects for twelve wins with the best title odds in the AFC. San Francisco is 3-0 and projects for ten with a 73 percent playoff chance. Between them the re-run has added seven wins to the two teams we were most sure about in the wrong direction.

The model does not defend its preseason numbers. It reads every snap that has been played and re-prices the roster, and three weeks of Patrick Mahomes at a 109 passer rating and Brock Purdy at 133 without a sack taken were enough to overturn two months of grades. Elsewhere Miami has fallen to two wins, Buffalo has finished a five-win climb, and the Rams are 1-2 and still the favorite by a distance no other team is within twenty points of.

The forty-eight games that have been played are in the books. The remaining fourteen weeks and the postseason were played out two hundred times with every player's three weeks of real production seeded into his line and the injury list as it stands on Tuesday morning. What came back is below, and so is the honest accounting of how the model has done so far, because that part matters more the deeper into a season we get.

The scorecard

Our matchup projections have picked a favorite in all forty-eight games and gotten twenty-two of them right. Seven of sixteen in Week 1, seven of sixteen in Week 2, eight of sixteen in Week 3. That is worse than a coin flip, and the misses have not been close ones. We had Seattle at 90 percent in Washington and Marcus Mariota threw four touchdowns. We had Atlanta at 82 percent at home against Carolina and lost by 31. We had the Chargers at 80 percent twice and they lost both by twelve. We had the Rams at 90, 96 and 80 percent and they won once.

Three weeks is the stretch of the season when the league looks least like the season before it, and the model's inputs are the season before it. That is an explanation, not an excuse. The projections below have absorbed all forty-eight results, and the teams that have made the model look worst are the ones whose projections have moved the most.

Kansas City turned the model around in three Sundays

In July our model stepped off the Chiefs bandwagon at eight wins. After Week 1, before Kansas City had played, it stepped further off to seven and a half. Then Patrick Mahomes beat Denver 31-10, beat Indianapolis 33-30, and went to Miami and won 24-10, and he did it with a 109 passer rating, seven touchdowns and two interceptions while Kenneth Walker III ran for 360 yards, the most in the league. When we re-ran the season, Kansas City came out at 12-5 with a 92 percent playoff chance and a ten percent shot at the title, the best championship number in the AFC.

That is a four-and-a-half win move in three weeks, and it is the largest swing our model has produced for any team in either direction this season. Part of it is three banked wins, one over the division rival we had as the AFC's best. Part of it is an offense that graded in the bottom third in September and now projects for 431 points, third in the league, because the players have done things the summer grades did not expect. The pass rush still projects for 33 sacks, 28th in football, and that is the one line the model has not moved. Everything else has.

San Francisco went from a hard fade to a division contender

The July preview called the 49ers a team the market was overrating on reputation and projected six and a half wins. After Melbourne we moved them to nine and said one game did not make a defense. Then San Francisco beat Miami 35-13 and Arizona 36-30, Purdy finished three games 60 of 83 for 789 yards, nine touchdowns, one interception and zero sacks taken, and the defense our model graded 30th in the summer held the Rams to seven and the Dolphins to 13. When we re-ran the season, the 49ers came out at 10-6 with a 73 percent playoff chance, a three-and-a-half win climb since July that is the largest upward move in the league.

The fine print is a calf muscle. Nick Bosa sat out the Arizona game and is expected to miss multiple weeks, Mike Evans has a rib strain, and Trent Williams left with a stinger. The defense projects ninth in points allowed with Bosa healthy, and Arizona's 30 points on Sunday were a look at it without him. But the model no longer calls this team a coin flip for the playoffs. It calls it the division leader, two games clear of the Rams with the head-to-head banked, and the Week 5 game in Seattle is where the division starts to be decided.

The Rams are 1-2 and still the favorite

Los Angeles lost by twenty in Melbourne, beat the Giants, and lost 30-26 in Denver on a night Matthew Stafford threw for 403 yards. Myles Garrett is on injured reserve until at least Week 6. The Rams are two games behind San Francisco in their own division with the head-to-head already lost. And when we ran the rest of the season, Los Angeles came out at 13-4 with a 99 percent playoff chance and a 38.5 percent shot at the title. Detroit is next at 19.

The model is not being stubborn. Davante Adams has 358 receiving yards, the most in the league, and projects for 1,574. The offense projects for 482 points, second in football. The defense, even without Garrett, projects to allow the second-fewest points and yards in the league because the coverage grades are elite. What has changed is the standings, and the standings are why the record projection has slid from fifteen wins in July to thirteen now. The second San Francisco meeting is the most consequential game left on the schedule for either team.

Buffalo finished the climb

Our boldest July call was Buffalo at five wins. After Week 1 we moved the Bills to eight and wrote that one shootout win in Houston did not make a secondary. Then Buffalo beat Detroit 41-31 and the Chargers 24-16, Greg Rousseau piled up six sacks in three games, and when we re-ran the season the Bills came out at 10-7 with a 65 percent playoff chance and the AFC East lead. That is a five-win climb since the summer, the largest for any team, and the fine print is the same as it was: the defense has gone from projected dead last to projected 27th, and Buffalo has won three games by scoring 34 a game. It is enough against most of the schedule. It will be tested against the Rams in Week 5, where our model has the Bills at six percent, the most lopsided line on the board.

Jacksonville has the best defense in football

Thirty-six points allowed in three games. The Jaguars held Cleveland to ten, held Denver to 20 in the one game they lost, and held New England to six in a 35-6 rout. Our model now has Jacksonville at 11-6 with a 90 percent playoff chance and a 9.5 percent title shot, second in the AFC behind Kansas City. The defense projects to allow 269 points, the fewest in the league, and the division is already Jacksonville's: Houston is 0-3 with the best pass rush in football and three touchdown passes from C.J. Stroud, and has fallen from eleven wins in July to nine. The Week 4 game in Cincinnati, between two eleven-win projections, is the best matchup on the early schedule.

Where the arguments moved

Miami is the collapse. The Dolphins are 0-3, have scored 36 points, and lost De'Von Achane to a torn ACL on Sunday. Malik Willis has a 74 passer rating. Our model has them at 2-14, down from six wins after Week 1 and seven in July, the lowest projection it has produced for any team this season. In the summer we said the market had Miami right at five. It turns out the market was generous.

Denver is the projection that will start arguments. The Broncos are 2-1 with wins over Jacksonville and the Rams, and our model dropped them from eleven and a half wins to 9-8. The reason is Bo Nix, who has an 80 passer rating with four touchdowns and three interceptions, and an offense that has scored 60 points in three games. The wins are banked and the defense earned them. The model prices the quarterback's numbers, and they have not been the numbers the summer projection expected.

Carolina and Minnesota are the climbs the model resisted in September and stopped resisting. Bryce Young has thrown for 939 yards and seven touchdowns, the Panthers beat Atlanta 34-3 on the road, and Carolina projects for 9-8, up from six. The Vikings are 3-0 having allowed 41 points, Dallas Turner has four sacks, and Minnesota projects for 9-8, up from six. Both are three-win moves since the summer. Both are built on a unit the model graded in the bottom third in July playing like a top-ten one.

Cleveland has won two straight with Deshaun Watson at a 104 passer rating and climbed from three wins to six. Las Vegas is 3-0 with Kirk Cousins throwing nine touchdowns and projects for nine wins, up from six, with the caveat that the three teams it beat project for 16 wins between them.

Going the other way, the Chargers are 0-3 with Justin Herbert throwing four interceptions and three touchdowns, and have fallen from nine wins in July to six. Tampa Bay is 0-3, Baker Mayfield dislocated his thumb and is out at least three weeks, and the Buccaneers project for three wins. Philadelphia scored seven points in Chicago against a backup quarterback, slid from 12-5 to 11-6, and saw its title odds cut from seven percent to under two. Green Bay, still without Parsons and Jacobs, gave up 35 at home to Atlanta and fell to seven wins.

The quarterback carousel

Six teams have started more than one quarterback in three weeks, and the model has priced all of it. Jaxson Dart is out for the season with a knee injury and the Giants traded a fifth-round pick for J.J. McCarthy on Monday; New York is 2-1 anyway and projects for six wins with Jameis Winston. Jayden Daniels dislocated his elbow and Marcus Mariota beat Seattle; Washington projects for five. Caleb Williams has a Grade 2 hamstring and Case Keenum beat Philadelphia by twenty; Chicago projects for ten. Michael Penix Jr. returned from his knee and beat Green Bay by 21; Atlanta holds at ten. Sam Darnold is back for Seattle. Kyler Murray was concussed, returned, and has a 54 rating; Minnesota is 3-0 regardless.

What did not move

Detroit is still 13-4 with the highest-scoring offense in football at a projected 491 points, and its title odds climbed from eleven and a half percent to 19 as Philadelphia and the Rams stumbled. Baltimore is still 11-6, and its title odds tripled to six and a half with Lamar Jackson carrying a 113 rating and Derrick Henry on pace for 16 touchdowns. Cincinnati slid a win to 11-6 after losing in Pittsburgh and the AFC North is now a coin flip between the Bengals and Ravens with the Steelers at ten wins a step behind. Tennessee is still three wins with the lowest-scoring offense in football. Atlanta and New Orleans are where they were, at ten and eight.

Go argue with it again

The philosophy has not changed and neither has the method. The model prices the players on the field, what they have actually done, who is actually available, and the schedule ahead. Three weeks ago that produced a league in which the Rams were untouchable, Kansas City was a losing team and San Francisco was a fade. Today it produces a league in which the Rams are still the favorite, Kansas City has the best title odds in the AFC and the 49ers lead the NFC West, and the difference is three Sundays of Mahomes and Purdy playing at a level their summer grades did not allow for. We got both wrong in July. The numbers below are what the model thinks now that it has watched them.

Every one of the 32 team previews has been rewritten for this run: the projected record against the July and September ones, the three-game body of work, how our matchup projections did on those three games, the players the model has climbing and falling, and what the next month looks like by our numbers. Start with the full 2026 league projection, standings and odds, then go find out whether three weeks was enough to change your mind about your team, or whether the machine still owes you an apology.

All 32 team previews

AFC East
Buffalo Bills
Miami Dolphins
New England Patriots
New York Jets

AFC North
Baltimore Ravens
Cincinnati Bengals
Cleveland Browns
Pittsburgh Steelers

AFC South
Houston Texans
Indianapolis Colts
Jacksonville Jaguars
Tennessee Titans

AFC West
Denver Broncos
Kansas City Chiefs
Las Vegas Raiders
Los Angeles Chargers

NFC East
Dallas Cowboys
New York Giants
Philadelphia Eagles
Washington Commanders

NFC North
Chicago Bears
Detroit Lions
Green Bay Packers
Minnesota Vikings

NFC South
Atlanta Falcons
Carolina Panthers
New Orleans Saints
Tampa Bay Buccaneers

NFC West
Arizona Cardinals
Los Angeles Rams
San Francisco 49ers
Seattle Seahawks