We Ran the Season Again After Week 2. San Francisco Is Real. Kansas City Made Us Apologize.
Two weeks ago we re-ran the 2026 season after the opening Sunday and told you Buffalo was back and the Rams never left. Week 2 had notes for us. San Francisco followed the Melbourne massacre by hanging 35 on Miami without punting a meaningful drive. Patrick Mahomes threw for 382 in an overtime game his team refused to lose. Atlanta scored three points at home behind a third-string quarterback and an undrafted rookie. So we fed every snap of it back into the machine, locked all thirty-one real results into the standings, and played the rest of the season 200 times over. Here is what moved, and what would not budge
First, what changed under the hood
A housekeeping note, because we believe in showing our work. Every game already played is no longer simulated at all: the real score goes into the record and the season is played out from there. Player form now moves faster too. We retuned the blend so that a snap played in 2026 counts for two and a half snaps of 2025 evidence, and a team's own results reach full weight at ten games instead of twelve. In English: the model still leans on a full season of last year's film this early, about three quarters of the vote, but what it watched these two weeks now speaks sooner and louder. Two weeks of Purdy to Kittle, two weeks of Malik Willis behind that line, two weeks of everything, all priced in.
San Francisco is real, and the model has stopped arguing
We faded the 49ers hard in July. Under seven wins, an eight percent playoff chance, a defense graded near the bottom of the league. They are 2-0 with a 62-20 aggregate, they beat the Super Bowl favorite by twenty on a neutral field, and Brock Purdy just walked through Miami's secondary going 20 of 22 for 287 with two scores and a rushing touchdown. The re-run has San Francisco at 10-7 with a 60 percent playoff chance, the biggest riser in football from where we had them six weeks ago, and the projection now hands Purdy 4,048 yards and 27 touchdowns. The number to watch is on the other side of the ball: our summer sim had this defense allowing 424 points, and the current one says 339. Raheem Morris has four sacks of a helpless quarterback in the home opener, a rookie in Romello Height getting his first, and Christian McCaffrey celebrating his 100th career touchdown on a workload the staff is finally managing. Kaelon Black led the team in carries in Melbourne. The plan the roster critics doubted is simply working.
Kansas City made us apologize
In August the model stepped off the bandwagon and we printed it: a losing record, a thinned roster, a division that had passed them. Kansas City's response was 31-10 over Denver and then 33-30 in overtime over Indianapolis on Sunday night, Mahomes 32 of 47 for 382 and three scores without a pick, now 8-0 in regular-season overtimes and looking every inch recovered from the knee. The re-run puts the Chiefs at 11-6 with a 93 percent playoff chance, and the projection hands Mahomes 4,136 yards and 28 touchdowns with nearly 500 more on the ground. It is the largest self-correction this model has made in two seasons of running it, and it is entirely deserved. We priced the roster and ignored the two constants that have decided this franchise's Januaries for a decade: Mahomes in one-score games, and a staff that wins the margins. Both showed up inside two weeks.
The Rams never left, part two
Los Angeles is 0-1 with a twenty-point humiliation on its record, its shiny new edge rusher was blanked on 27 pass-rush snaps in Melbourne, and Matthew Stafford posted the second-worst first half of his career. The model's reaction: 14-3, a 46 percent share of the Lombardi, and a playoff berth in every single simulation. That is not stubbornness, it is arithmetic. One locked loss costs a fourteen-win roster exactly one win, and nothing about the talent changed in one bad afternoon. The Rams still grade first in the league on offense and first on defense in our runs. What did change is the neighborhood: the team that embarrassed them shares their division and no longer projects as a pushover, and Seattle sits at 12-5 with the second-best title odds in football. The NFC West is now the deepest division in our sim, and the road to SoFi in February runs through it.
The rest of the risers
Cincinnati is 2-0 and up to 12-5 with 90 percent playoff odds, the bruising 20-6 handling of Houston the kind of complete win the model loves. Philadelphia quietly sits at 12-5 and 93 percent. Detroit holds at 12-5, still the most complete roster in the NFC by our grades. Jacksonville lost in Denver, watched a nine-game winning streak die, and still projects at 11-6 with the AFC South lead, which tells you what the model thinks of the teams chasing them. And Denver deserves its own line: down 13-3 at halftime to that Jaguars team, Bo Nix went 22 of 31 for 288 in a 17-0 second half, and the Broncos re-ran at 9-10 wins as the strongest team in the wild-card scrum.
The fallers, in order of pain
Atlanta first, because the collapse is total. The bold call was Michael Penix Jr. dragging a young team to eleven wins; he has yet to take a snap, Tua Tagovailoa has yet to be healthy enough to bridge to him, and Cooper Rush plus undrafted rookie Jack Strand produced three points at home in a 34-3 loss to Carolina, Strand's first NFL pass going back the other way for six. The Falcons re-ran at 8-9 with a 36 percent playoff chance and falling, and the injury report, Terrell, Ebukam, Jawaan Taylor, is getting longer on a short week. Bijan Robinson passing 4,000 career rushing yards before turning 25 deserved a better stage.
Houston is 0-2 for the second straight September, and this time the model has taken the division away: 10-7 on the median run, 56 percent, chasing Jacksonville. C.J. Stroud threw a career-high 55 passes in Cincinnati with Nico Collins inactive, Jayden Higgins on season-ending IR and Tank Dell still working back, and led the team in rushing with 29 yards. The talent argument survives; the depth argument is over. Indianapolis, the projection we flagged in July and trimmed in September, has finished the trip: 0-2 despite pushing Kansas City to overtime, and now 7.7 wins on the average run with a 14 percent playoff shot, seventy-four points allowed doing exactly what the summer grades warned. Miami has crashed to 4-13 and one percent, Malik Willis absorbing sacks at a rate our sim projects to 62 on the season, which would chase modern history. Tampa Bay's Mayfield-regression collapse is arriving right on schedule at under five wins. Tennessee is braced at 3-14.
Buffalo, still the argument
And then there is the call we will be judged on. The Bills are 2-0, Josh Allen looks elite, and the market has them among the favorites. Our re-run barely moved: 9-8 with a 46 percent playoff chance. The model watched the same two games you did and still does not believe a defense that lost most of its best players in one offseason can survive seventeen weeks of this conference. Two weeks ago we called Buffalo our boldest fade. After two Bills wins, it is now our boldest call by a distance, and we are not backing off it. If we are wrong anywhere, it is here, and we will run the season again next week and tell you.
What week 3 can move
Atlanta gets Green Bay on Thursday night with a quarterback room held together by tape. Houston needs a win before the hole becomes the season, and the model prices their next three as must-have. San Francisco's number climbs again with any road proof, and the Rams get their first chance to answer Melbourne. The games are real now, and the model is listening faster than it did in July. Every result goes into the bank, every snap moves a grade, and we will keep publishing what comes back, especially when it makes us apologize.