I used to work at Robinhood. Now I build NFL models for fun and test them in public. This week's most interesting number: Houston by 3.1. Most expect a comfortable Texans win. My model says Tennessee keeps it close. All 13 Sunday forecasts in the reply. Results Monday. Last week I said the model I've been working on was coming to life. I like football. I like data. Apparently watching the games wasn't enough, so I gave myself homework. Three games I'm watching: ★ TEXANS vs TITANS Projected margin: Houston by 3.1 I have Houston winning, but this one's a lot tighter than the narrative around it. Tennessee hangs around. ★ PATRIOTS vs RAIDERS Projected margin: New England by 6.8 The model likes the Patriots more than most. The depleted secondary is the biggest reason I'm cautious. ★ BROWNS vs JETS Projected margin: Cleveland by 0.85 Basically a coin flip that leans Cleveland. Interesting enough to track, not enough to pretend it's certain. These forecasts predate the latest injury updates. What I'm testing is how close the projected margins land. I'm not claiming the model is right. Monday, I'll show what held up, what missed, and what needs more work. Everything stays visible. Follow along for Sunday forecasts and Monday results.

If Aaron Rodgers passes Drew Brees next year, he's a better quarterback than Peyton Manning. Not close. He's 42. He just passed Peyton today, in Paris. I used to work at Robinhood, so naturally I ran the math on the rest of the list. Brady: 649 Brees: 571 Rodgers: 540 Manning: 539 The record breaker was a 4 yard fade to Darnell Washington in the first NFL game ever played in France. 19 of 30, 212 yards, no picks. Steelers win 16 to 10. (My model had Pittsburgh this morning.) THE MATH Rodgers has 13 TDs through 7 games. About 1.9 per game. Brees is 31 away. At this pace he finishes the year around 559. He'd need one more season and about 13 more TDs. Brady is 109 away. That's roughly three more full seasons. He'd have to play until he's 46. Brady retired at 45. Brady's record is safe. Brees's isn't. And the moment Rodgers passes him, the Peyton debate is over. Tell me why I'm wrong. I post the numbers every week and grade myself every Monday. Follow if you want to see who was right.

Baker Mayfield got sacked 6 times and hit 12 times against Minnesota. He dislocated his throwing thumb on the final drive. He's back Sunday. His starting right tackle isn't. Baker was never Tampa's problem. Here's why. 1. The offense was broken before he got hurt Tampa went 0 and 3 with him. 2 TDs, 2 picks. Easy to blame the quarterback. But look at the Minnesota game. 2 for 17 on third down. An 85 yard Chris Godwin touchdown wiped out by holding. The hold was on Luke Goedeke. The same right tackle who's out Sunday. 2. He's walking into the same mess First game back from a dislocated throwing thumb. Behind a line that let him get hit 12 times the last time he played. No Goedeke. No Emeka Egbuka either, who's out with a concussion and a foot injury. That's not a comeback. That's a stress test. 3. The backup numbers tell you everything Jalon Daniels deserves credit. He beat Dallas, still Tampa's only win this year. But he threw 3 TDs and 5 picks while Baker was out. The Bucs went 1 and 2 in his starts. Different quarterback. Same problems. My call Tampa's issue is the line, not the quarterback. If they lose Sunday, people will blame Baker. They'll be wrong. Tell me I'm wrong.

I used to work at Robinhood. Now I build NFL models for fun and test them in public. Caleb vs. Maye will get all the attention. My model picks Chicago. New England’s injury report makes that call interesting. Two linemen out. Another questionable. Here’s where Thursday night could turn. I build NFL models for fun. Then I put the predictions somewhere I can’t conveniently forget them. This week: Bears to beat the Patriots. What makes me agree with the pick? New England has Ben Brown and Alijah Vera-Tucker ruled out. Morgan Moses is questionable. Chicago is allowing opponents to convert just 25.8% of third downs. A patched-up offensive line trying to sort out disguised pressure is a matchup I’d want to attack. But there’s a very real way this prediction goes wrong. Chicago ranks 29th in EPA allowed on first and second down. Maye can punish this defense before it ever gets to those third-down situations. That’s what I’m watching: Can New England create big plays early, or does Chicago keep forcing Maye into obvious passing downs? Caleb’s 86.1% adjusted completion rate leads PFF’s qualifying quarterbacks. That’s encouraging. It also doesn’t mean every throw carries the same difficulty. I’m taking Chicago. I’m watching the protection and early downs to see whether the reasoning holds up. Who’s your winner, and what am I missing? Follow me for the forecasts and postgame breakdowns. Friday, we check this one. Saturday, the Sunday Film Room returns.

Drake Maye’s first 3 games: 6 interceptions. His next 3: 10 touchdown passes. ONE interception. If you’re still grading him off September, catch up. Over those last three weeks, his 119.3 passer rating leads qualified quarterbacks. That’s a turnaround worth discussing before tonight becomes another Caleb vs. Maye argument decided entirely by the scoreboard. I’m picking Chicago. I also think Maye’s recent play deserves respect. Those opinions can coexist. New England has Ben Brown and Alijah Vera-Tucker ruled out. What I’m watching is how Maye handles pressure behind that reshuffled line. Does he find the quick answer? Does he recognize when the big play isn’t there? Does he protect the ball when the first read disappears? Those answers tell us more than “his team lost.” Who are you taking tonight: Maye or Caleb? Give me the football reason. Follow for the forecasts before kickoff and the breakdowns afterward. Friday, we check what actually happened.

I used to work at Robinhood. Now I build NFL models for fun and test them in public. Washington is 1–4–1. Jayden Daniels just threw for 300 yards, 2 TDs and 0 picks. Against the unbeaten 49ers. Blaming the quarterback is easy. Explaining what he actually did wrong takes more work. I picked San Francisco. They won 23–17. Daniels still completed 29 of 40 passes and helped turn a 23–3 deficit into a six-point game. That doesn’t erase his mistakes. It also doesn’t make Washington’s record an individual quarterback grade. If the argument is that he’s part of the problem, name the decision or the play. “Look at the record” isn’t enough. Washington fans: what’s the strongest criticism of Daniels’ performance last night?

My Monday call: San Francisco beats Washington. San Francisco’s passing offense: 100/100. Washington’s pass defense: 6/100. Those are baseline model ratings. Quite the introduction. But the reason I lean San Francisco starts with Purdy. The matchup ratings put San Francisco’s passing offense at the top of the league—in both efficiency and consistency. Washington’s pass defense sits near the bottom. That’s the matchup I’m watching. I expect Brock Purdy to keep moving the chains and make Washington work to get off the field. The strongest argument against my call? San Francisco’s injury list. Nick Bosa and Deebo Samuel are out. Mike Evans is questionable. Those baseline ratings aren’t a fresh rerun accounting for every absence. Missing receivers matters, and a defense without Bosa gives Jayden Daniels an opportunity to make this uncomfortable. Washington is also without its leading rusher, Jacory Croskey-Merritt. I expect more responsibility to fall on Daniels. My prediction: San Francisco gets to 6–0. The spreadsheet picked the unbeaten team. Groundbreaking. But the passing matchup is why I agree with it. Tuesday: we grade the call. Saturday: the full Sunday Film Room. Follow @ArmChairLogan for the forecasts and the breakdown afterward. Wrong calls stay up. Washington fans: is Daniels attacking a defense without Bosa your best case—or am I overlooking a better matchup?

My NFL predictions went 5–7 on Sunday. Caleb Williams went 24-for-24 from a clean pocket. One quarterback input still had Tyson Bagent starting. That’s a process problem. Here’s what I’m changing before the next Film Room. MONDAY RECEIPTS — WEEK 6 5 correct. 7 wrong. The three highlighted calls went 1–2. The spreadsheet has lost its right to be smug. WHAT HELD UP Carolina beat Philadelphia 25–14. New England, Kansas City, Buffalo and Green Bay also won. Those count. So do all seven misses. WHERE THE READ BROKE DOWN ★ TAMPA BAY I backed the defense. Tampa took a 16–3 lead into the fourth quarter, then lost 26–19 in overtime. Jalon Daniels threw two fourth-quarter interceptions. A defense keeping the game winnable doesn’t mean the offense can finish it. Review: how the forecast handles inexperienced quarterbacks against pressure—and how uncertain that estimate should be. ★ JACKSONVILLE I highlighted Houston’s missing players. C.J. Stroud still threw for 312 yards and three touchdowns. Jaylin Noel had 107 receiving yards. The question isn’t just who’s missing. It’s what the available players can do. Review: replacement roles, receiving depth and explosive passing. CHICAGO AND BALTIMORE One input assumed Bagent. Williams played. Another still listed Lamar Jackson. Tyler Huntley started. I flagged those mismatches and kept the baseline calls. Footnotes don’t fix inputs. THE UPDATE PLAN • Verify the quarterback and key lineup assumptions before publishing. • Test how missing starters affect the specific matchup, including pass protection and replacements. • Compare the proposed changes against the old version on past games, using only information available before kickoff. I’m not adding a “never trust Tampa” button because one lead disappeared. And I’m not calling the model improved until the testing supports it. Full scorecard below. The original stays up. Follow @ArmChairLogan before Saturday’s Film Room. I’ll show what changed, what’s still being tested, and the next forecasts. Which deserves the first deep dive: quarterback assumptions, protection, or replacement players?

My model picked Seattle. Seattle gained 353 yards. Denver gained 168. JSN caught 12 passes for 134 yards. Denver won 17–14 in overtime. Apparently they award wins for points. Rude. The model got the winner wrong. Here’s what I’m reviewing. Seattle limited Bo Nix to 107 passing yards. JSN produced. Those parts of the matchup supported the original reasoning. But Denver’s defense forced three turnovers: a Shaheed fumble and two Darnold interceptions. The overtime interception set up the winning field goal. Seattle’s yardage advantage produced just 14 points. That’s the lesson: moving the ball and finishing possessions are different problems. For the next model review, I’m testing whether accounting differently for quarterback pressure, turnover risk and scoring efficiency improves predictions across past games. One loss gives me questions. It doesn’t justify rewriting the model until it agrees with yesterday’s score. Denver fans: what did the pregame analysis underrate most—your pass rush or Seattle’s turnover risk? Tomorrow: the full Sunday forecasts. Monday: every result. The homework continues.

My Thursday call: Seattle beats Denver. JSN has 583 receiving yards through five games. Denver’s entire wide receiver room has 529. One receiver. More yards than the whole room. But the reason I lean Seattle starts with Bo Nix. Seattle has generated pressure on 44.3% of opposing dropbacks this season. Nix has completed just 35.4% of his passes when pressured. That’s the matchup I’m watching. I expect Seattle to make Denver work for its scoring drives. Denver’s secondary has injury concerns, too. Tuesday’s estimated report had Surtain and Hufanga as nonparticipants and Moss limited. JSN was also limited with an ankle injury. Final availability matters on both sides. The strongest argument against my call? Seattle has EIGHT giveaways in its last three games. Denver’s defense can punish that. The Seahawks have enough to win this game. They also have an impressive ability to donate possessions. I’m picking Seattle to clean up enough of that mess. Friday: we grade the call. Saturday: the full Sunday forecasts. Denver fans: is your best case the defense forcing turnovers, or something I’m overlooking in the matchup?

I picked Houston and New England to win. Both won. I expected wins by about 3 and 7. They won by 23 and 28. The spreadsheet is trying to claim full credit. Exact forecasts: Houston by 3.1 → won by 23. New England by 6.8 → won by 28. Getting the winner right is one thing. Predicting how close the game will be is another. I’m tracking both. The model doesn’t get an A for answering half the question. Saturday: the next forecasts and what I learned from these misses. How would you grade these two calls: credit for the winners, or a poor forecast because neither game was close?

Monday results are in: 8 correct winners, 4 wrong, 1 tie. The winner calls and margin forecasts are graded here, including the misses: xitter.cf/ArmChairLogan/status/2… Next forecasts Saturday.

THE MODEL'S SUNDAY FORECASTS (OCT. 11) ★ = the three games highlighted above PHI vs JAX: Jacksonville CHI vs GB: Chicago CIN vs MIA: Cincinnati CLE vs NYJ: Cleveland ★ HOU vs TEN: Houston ★ IND vs PIT: Pittsburgh LV vs NE: New England ★ MIN vs NO: Minnesota NYG vs WAS: Washington DEN vs LAC: Denver DET vs ARI: Detroit SF vs SEA: Seattle BAL vs ATL: Atlanta† The model's least confident calls: Chicago is projected to win by less than 2. Minnesota's projected win is barely above a tie. † Atlanta is my judgment call with Lamar ruled out. The model still assumes he starts, so I'll grade that override separately. Any changes get a timestamped reply. The original stays visible. Monday's scorecard: - All 13 winner results - How far each projected margin missed - The three starred games separately - Whether my Atlanta override helped or hurt The misses stay up. Which call do you disagree with, and what am I missing?