The Sports PageThe Turnaround We Wrote About Beat Michigan on Saturday
The question we ask every Sunday: what did we claim this week, and did reality agree?
The scoreboard. Four issues this week, zero gradeable predictions: zero hit, zero missed, zero still pending from this week. All four were analytical and retrospective pieces. The Bills and Notre Dame remain our two active forward picks from Issue #202; both still pending and reported below.
Here is the honest note: we owe you no new scorecard this week. The four pieces published — schedule strength, pitching changes, coaching turnarounds, and the Rays’ forecast miss — were all retrospective analyses of what already happened. None contained forward predictions. The Bills and Notre Dame are the standing claims, and they have not changed today.
Why we bother: a prediction nobody grades is just a story.
The Week in Review: Issues #205–208
- #205Oct 19 — Who Did They Even Play? — Notre Dame’s schedule strength, measured. The front half was light. The back half has three ranked opponents remaining. We said we’d count again when it’s over.
- #206Oct 20 — We Bet Runs Go Up After a Pitching Change. They Do — Before It. — Our editor’s hypothesis, tested against five years of play-by-play. Runs climb before the change, not after. The new arm is no worse than the old one, inning for inning.
- #207Oct 21 — Nearly 200 Bad Teams Changed Coaches. One Did This. — Indiana under Cignetti gained 28 SP+ points in year one. The average new-coach bump across 194 programs: 3 points. Then Indiana beat Michigan in Ann Arbor on Saturday.
- #208Oct 22 — The Rays Were Supposed to Win 79 Games. — Our napkin model forecast Tampa Bay at 79. They won 98 and opened the World Series at home on Friday.
The Prediction Scorecard: Active Picks
| # | Prediction (Issue #202, Oct 16) | Current State | Grade |
|---|---|---|---|
| 1 | Notre Dame wins CFB National Championship (~16% at issue time; updated to <5% after BYU loss) | BYE this week. Record stays 6–1. Next: Navy, Oct 31. CFP path still requires winning the ACC championship game. | PENDING — no update this week. Title odds remain well below 5%. Path narrow but open. |
| 2 | Buffalo Bills win Super Bowl (~15%) | Beat Raiders 33–30 last week. Now 5–1. BYE this week. Model ranked them No. 1 in the league at time of pick. | PENDING — season ongoing. Bills are one game better than when we wrote the piece. No reason to revise the probability downward. |
Issues #205–208 contained no forward forecasts. Issue #202’s Bills and Notre Dame picks remain the active claims. The Bills’ 5–1 record is consistent with a team the model liked; their bye week is a chance to reset heading into the second half. Notre Dame’s path requires things to go right in both their remaining games and the standings. The model updates on evidence; right now there is none new to update on.
What Yesterday’s College Football Showed
Notre Dame was on bye. The Irish need next week; they get Navy on October 31 before a gauntlet of three ranked opponents. Two of the teams we wrote about on Tuesday spent Saturday doing things that are hard to explain.
Indiana is 7–1. They beat Michigan 13–6 in overtime in Ann Arbor. Three days after we published the piece showing that of 194 bad teams that hired new coaches, the average improvement was 3 SP+ points, and Indiana’s +28 was the outlier of the entire dataset — three days after that, they won in Michigan Stadium. You can look at that result and call it confirmation. Or you can remember what Tuesday’s piece said: for every Indiana there are 194 teams that got three points better. One result, however good, does not relocate the distribution.
BYU was No. 2 in the country. UCF was not ranked. UCF won in double overtime, 39–31. BYU was a 6.5-point favorite entering Saturday; based on the pregame line and a standard deviation of 16.5 points, that translates to roughly a 35% win probability for UCF. One in three. UCF won. This is not shocking. A 1-in-3 result is expected to happen roughly once every three times the situation arises.
Texas Tech was No. 8. Cincinnati was unranked. Cincinnati won 37–24. Texas Tech was probably a double-digit favorite; based on published lines of roughly 10 points, Cincinnati was about a 27% shot. Two ranked teams lost to unranked opponents on the same afternoon.
One in Three, One in Four, Same Saturday
Here is the math the broadcast will not run. UCF had roughly a 35% chance to beat BYU. Cincinnati had roughly a 27% chance to beat Texas Tech. The probability of both upsets happening on the same Saturday, assuming independence:
There are roughly 13 Saturdays in a college football regular season. The probability that at least one of those Saturdays produces both upsets in a given year: about 1 − (1−0.094)^13 ≈ 72%. This was probably going to happen this year. We got it in Week 8.
You can look at this and say: two top-10 teams lost the same day, that’s historic. You can also say: the expected number of top-10 teams losing in any given week is between one and two, and this week was two. Both are true. One of them tells you something about sports; the other tells you something about math.
The lesson is the one Issue #198 built the whole Wisconsin-Penn State piece around: a low-probability outcome is not evidence that the forecast was wrong. UCF winning at 35% odds is not a surprise. It is exactly what a 35% probability means: it happens about one in every three times. Saturday was one of those times. If you are surprised, the problem is that you treated 65% as 100%.
“65% and 73% are not certainties. They are the two-thirds and three-quarters that keep losing every year.”
— The Sports Page, on BYU and Texas Tech, October 25, 2026What Readers Read · October 15 – October 21
Counts are unique pageviews per GoatCounter, with author visits excluded. Pulled 2026-10-22.
What to Watch Today — and Why Those Games
The Bills are on bye, so you can root abstractly today. The NFL card is ranked below by nearness to a coin flip: the spread is the market’s best estimate of how far apart these clubs actually are. A game near pick’em is one where a single play, bounce, or conversion holds the most information about which club is better. Those are the games worth watching, even when neither team is yours.
| Matchup | Spread | Win prob. | Why watch |
|---|---|---|---|
| Browns vs. Titans (1pm ET) | CLE −1.5 to −2.5 | 54–57% Cleveland | The board’s closest thing to a coin flip. The market can barely separate these clubs. One play decides what both teams think of themselves heading into Week 8. |
| Chiefs at Seahawks (SNF, 8:20pm ET) | SEA −1.5 to −3 | 54–57% Seattle | Kansas City as an underdog at Seattle. The market is giving Mahomes almost nothing. One of those games where you watch to find out which number was right. |
| Bengals at Ravens (1pm ET) | BAL −2.5 to −3 | 57–59% Baltimore | A 3-point game in the AFC North. Division games shrink in one direction or open in another; this one separates a contender from a pretender. |
| Lions vs. Packers (4:25pm ET) | DET −3 to −3.5 | 59% Detroit | The highest total on the board (53.5 points). A 3-point spread on a game expected to score 53 means almost any outcome is on the table. |
| Jets vs. Dolphins (1pm ET) | NYJ −4.5 to −5.5 | 62–66% Jets | If you follow the family teams, the Jets are favored at home. The spread moved — check the line before you decide anything. |
Spreads from market sources as of Sunday morning. Win probabilities use a 13.5-point NFL standard deviation: a 1.5-point favorite wins about 54% of the time, a 3-point favorite about 59%, a 5-point favorite about 64%. We hold no NFL model. Six weeks of data is not a model. What you see is the market’s estimate, not ours.
The World Series: Two Games In
The Brewers lead the Tampa Bay Rays 2–0. Game 1 in Milwaukee: Brewers 3, Rays 2. Game 2: Brewers 5, Rays 4. Game 3 is Monday at Tropicana Field, 8pm ET.
Issue #208 on Thursday showed what made the Rays worth watching: they won 98 games against a napkin forecast of 79. Of 750 team-seasons since 1999, only 14 beat the baseline by more. And yet they are down two games to a Brewers team that also came out of nowhere — 103 wins against a forecast of 90. Both of this year’s finalists are at the far right of the distribution we built on Thursday. The piece said that champions almost always beat their forecast. It turns out challengers for the title do too.
Being down 2–0 in a best-of-seven is not a death sentence. Teams down 2–0 have won 25% of World Series matchups historically. But the Rays need to win Monday in their own park. Game 3 is the most important game on the calendar this week.
Over-Reactions and Under-Reactions
Over-Reactions
The broadcast will spend the next week burying BYU. They were undefeated and No. 2 and they lost in double overtime to an unranked opponent in what everyone will call a collapse. What it actually was: a 35% event. BYU is still a good football team. Their AP ranking will drop. Their actual quality has not moved by 35% overnight. One game moves the number, not the team.
Under-Reactions
Indiana is 7–1. They beat Michigan in Ann Arbor in overtime after Issue #207 named them the statistical outlier across 194 coaching changes. The piece said it could not confirm that Cignetti was the cause — “we cannot see what Indiana would have done without him.” We still cannot. But the dataset has one more data point now, and it points the same direction it has all season.
What to Take Home, and the Road Ahead
Three things worth holding from this week. First: the pitching change piece on Tuesday showed that an editor’s intuition — stated publicly, tested honestly, with real data — can come back wrong. The right reaction to that is not embarrassment. The right reaction is the piece we ran, which printed the result plainly without softening it.
Second: the schedule strength piece on Monday said the Irish back half has three ranked opponents remaining. Notre Dame has a bye this week, then Navy, then those three. The back-half schedule we wrote about on Monday becomes the real story starting November 1.
Third: the Rays will win or lose the World Series in Tampa the next few days. The piece we published Thursday will hold regardless of outcome. Teams that beat the forecast by 19 games are lucky or good or both. The Series result will not tell us which. The full season’s evidence already did that work.
The queue this week holds the payroll piece, which is durable and can wait. It also holds Sorsby, which is held until the one-year follow-up date in 2027. The queue is thin; a piece is needed before the week is out.
“The piece said Indiana was the exception. Then they won in Ann Arbor. The exception keeps excepting.”
— The Sports Page, on what happens when you publish a finding on Tuesday and the subject plays on Saturday
