The Sports PageOne Month, Six Pieces, One Miss: The Astros Made History and Then We Watched Them Lose It
The question we ask every Sunday: what did we claim this week, and did reality agree?
The scoreboard. Six issues, one gradeable prediction: zero hit, one missed, zero still pending. Five pieces made no forward forecast at all — they were analysis, retrospective, or methods, and that is legitimate work. Saying so out loud is not an excuse; it is the accounting.
Here is the surprising part. Issue #185 said Cleveland was where players go to die. The adjusted data showed Cleveland is in the top third of career preservation. We were wrong in a direction that flattered a bad story, and the piece said so on the record. That is the miss we owe you.
Why we bother: a prediction nobody grades is just a story.
The Week in Review: Issues #184–189
- #184Sep 28 — Our New Rule Beat Our Old One. Random Numbers Beat Them Both. — Three weeks of pick’em data. Our old entry scored 202. Our clever new rule scored 213. A hat scored 224. The question to ask any winning idea: does it beat doing nothing in particular?
- #185Sep 29 — We Were Wrong About Cleveland. — We predicted Cleveland overworks and shortens careers. The adjusted survival data put Cleveland in the top third. The piece is built around the miss. Arizona is where careers actually end.
- #186Sep 30 — The Same Fifteen Picks. Fifty Points Apart. — Ordering your picks does not raise your expected score. It doubles your chance of winning. Those are different things. If you mostly agree with the crowd, which games you put your big numbers on is the only decision you have left.
- #187Oct 1 — Two Fans Spent the Same $4,000. One of Them Has $87,000. — What happens to a $4.55-per-week losing edge when you run it 40 years and ask what the money was doing instead. A losing edge does not drain you to nothing; it finds a level and holds you there.
- #188Oct 2 — The Mets Got One Day of Hope. The Cubs Got 126. — Hope-days: the area under the playoff probability curve. The Cubs banked 126 probability-days of belief from April to September. The Mets banked one. Both finished with nothing. The endings were identical. Everything that differed was in the middle.
- #189Oct 3 — We Found a Pattern in Forty Games. Then We Killed It. — Penn State returned 7.6% of last year’s offense and won 45–0. Clemson returned 41% and lost by 41. The correlation held across 49 week-one games (r = −0.276, t = −1.97). Then we gave it eleven seasons and it fell apart. This is what overfitting looks like while you are doing it.
The Prediction Scorecard
| # | Prediction | Outcome | Grade |
|---|---|---|---|
| #185 | Cleveland overuses players and shortens careers; “the stats are going to prove me right about this” | Cleveland is in the top third of career preservation across 1,848 players who changed clubs, 2015–2022. Arizona is where careers actually end. | MISS — wrong direction, wrong mechanism |
Issues #184, #186, #187, #188, #189 made no forward forecasts and are not graded. Five of six pieces this week were analysis, retrospective, or methods. That is a legitimate mix, not a dodge.
What We Got Right (Carried Forward)
The AL West Prediction: From 70% to 100% to History
Last Sunday: our Issue #182 prediction — 70% probability that the AL West champion would finish at .500 or worse — became 100% after Saturday’s results. The full resolution arrived this week.
Houston won their final game 9–0 over the Oakland Athletics. Texas lost. The Astros finished 81–81, winning the division by virtue of the tiebreaker. It is the first .500 division title in 57 years of baseball, across 270 division winners. Our model said this was 70% likely entering the final weekend. It was.
Then Houston was swept by the Chicago White Sox in the Wild Card, 2-0, losing 6–3 and 7–3. That is not a forecasting failure. We made no prediction about the Wild Card. It is new data: an 81-win club played a 90-win club and lost. One series tells you about nothing. Whether Houston deserved to be there was not what Issue #182 asked.
What We Got Wrong
The Cleveland Call
The original claim, made in a conversation that became notes for this paper: “Cleveland is where players go to die. And they die a horrible death, because they get overused. And I’ll tell you, the stats are going to prove me right about this.”
The adjusted survival analysis across 1,848 player-moves from 2015–2022 put Cleveland in the top third. When you hold experience, prior role, and year fixed, players who go to Cleveland are more likely to still be playing two years later than the league average, not less. Arizona is the bottom of the list, not Cleveland.
The mechanism we assumed has a selection problem: losing clubs sign players other clubs have finished with. Run the naive comparison and you confirm the theory by construction. You never actually measured anything. The piece explained that trap. We walked into it anyway.
This is the piece to read if you want to understand how a prediction can feel airtight and be completely wrong.
“We said the stats would prove us right. They proved us wrong. Committing to a direction before looking is the only way to say that honestly.”
— The Sports Page, Issue #185Over-Reactions and Under-Reactions
Over-Reactions
The hat benchmark in Issue #184 was the right test. But three weeks is a small sample: all three strategies landing left of center could still be noise. A hat at 224 over three weeks is evidence. Over a full season it becomes a verdict. We built the right frame. We have not run the experiment long enough to fill it.
Under-Reactions
Issue #188 established that the Mets banked exactly one hope-day this season, against Colorado’s two. That number — one probability-day of genuine playoff belief, total, from mid-April to late September — deserved a harder look at why. The piece correctly identified that the Mets were mathematically finished by late June. We did not chase the follow-up question: what did that look like from Opening Day?
What Saturday Showed
Week 5 of the college football season was quieter than its predecessor. Three top-25 teams lost Saturday, down sharply from eight the week before. Only one result qualified as a genuine upset.
Missouri, unranked, beat No. 8 Florida 45–17 in Gainesville. The Tigers put up 560 yards of offense on a team that entered the game as a two-score favorite. Florida had been 4–0 and climbing the rankings. Missouri had been getting there too, and on Saturday the distance between them dissolved inside one quarter.
Notre Dame stayed unbeaten, beating North Carolina 37–26 to move to 5–0. The Irish have won 15 straight, a run that now covers the full length of the current coaching staff’s tenure. Miami rolled past Clemson 41–13. Ohio State beat No. 15 Iowa 31–14; Jeremiah Smith had a 57-yard touchdown and set an Ohio State receiving record in the process.
The upset arithmetic I cannot give you. Each Sunday this edition is supposed to sum every underdog’s pre-game win probability and compare it against the actual upset count. A 1-in-13 shot is startling in its own game and near-inevitable across fifty. The pre-game rating snapshot for week 5 was not taken before Saturday’s kickoff. Without it, any arithmetic I print grades games on ratings that have already absorbed the results — which is how you convince yourself you saw everything coming. I will not do that. The lesson holds even when I cannot show the math.
What to Watch Today — and Why Those Games
The NFL card today ranks by nearness to a coin flip, not by marquee matchup. The spread is the market’s best estimate of the true gap between teams. A game at pick’em or −1.5 is two clubs the market genuinely cannot separate. Those are the games where a single bounce matters most.
This is not a prediction. We hold no NFL model. Four weeks is not a model. What you see below is the published spread as of this morning; a 50-50 game is labeled for what it is.
| Matchup | Spread | Why watch |
|---|---|---|
| Rams at Eagles (1pm ET) | Rams −1.5 | Eagles missing AJ Brown, Lane Johnson, DeVonta Smith. Rams a half-point from pick’em. Closest thing to a true coin flip on the board today. |
| Cardinals at Giants (1pm ET) | Giants −1.5 | Giants lost Brian Burns for the season (torn ACL). Two teams the market values identically. One of them will cover by five. |
| Jaguars at Bengals (1pm ET) | Bengals −2.5 | A field goal away from a coin flip. Both teams have struggled to be what they were supposed to be this year. |
| Cowboys at Texans (1pm ET) | Texans −2.5 | Houston just closed the AL West. Now the other Houston tries to hold off Dallas. Joe Mixon is inactive. |
| Broncos at 49ers (4pm ET) | 49ers −3 | San Francisco at home. Denver has won three of four. A three-point line means the favorite covers with a field goal. |
| Chiefs at Raiders (4pm ET) | Chiefs −4.5 | The clearest not-a-coin-flip on this row. Listed because the family has stakes in the Raiders and the Chiefs are a reasonable 4.5-point favorite, not a prohibitive one. |
| Jets at Bears (1pm ET) | Bears −3 | Caleb Williams is out. Breece Hall is out. Two teams playing without their most interesting player. |
| Lions at Panthers (SNF, 8:20pm ET) | Lions −3 | Sunday Night Football. Lions −3 means the game could go either way on a late field goal. That’s a proper prime-time game. |
Two games not on this table: Ravens −11.5 over Tennessee, Vikings −10 over Miami. Both injured on offense. The market has already decided. Watch them for the teams. Not coin flips.
October Arrived. The Mets Did Not.
The 2026 MLB postseason is underway. The Division Series opened yesterday.
In the American League: the Chicago White Sox beat the Houston Astros in the Wild Card 2–0. The Yankees swept the Red Sox 2–0. The White Sox now face Cleveland; the Yankees face Tampa Bay. Yesterday’s ALDS Game 1: White Sox 3, Guardians 0. Rays 1, Yankees 0.
In the National League: the Braves beat the Phillies in three games in the Wild Card. The Padres beat the Cubs 2–0. Yesterday’s NLDS Game 1: Dodgers 5, Braves 3. Brewers 3, Padres 2. Both Game 2s are today.
The Mets finished out of the postseason. Issue #188 told you how to measure their season: one hope-day. October confirmed it.
What to take home
Three things worth keeping from this week. First: commit to a direction before you look. The Cleveland piece only exists because the prediction was stated in advance. A wrong call discovered after the fact is data; a wrong call documented before is accountability. You can use the first to update. You can use the second to learn.
Second: the hat is the baseline, not the ceiling. Before you call any strategy good, ask what random assignment would score. Issue #184 showed that three weeks of confidence-pool data produced a clever rule, a cleverer rule, and a hat — in that order, worst to best. Most winning ideas have not been compared to nothing in particular.
Third: a prediction confirmed is not a prediction about what happens next. Houston went 81–81, won the division, and lost in the Wild Card in two games. The model called the first event. The second event was a new question with new evidence. Knowing one thing does not carry you forward; it tells you the cost of the next bet.
Road Ahead
The queue holds seven pieces. The most perishable are the pick’em analysis runs — those will eventually need refreshed data as the season accumulates weeks. The most durable: a lottery pricing piece and a calibration methods piece, both of which can publish any week of the year.
Football dominates the schedule from here through January. Every queue piece faces the variety rule, so we will not run back-to-back CFB or NFL. The next few weeks will likely alternate between baseball postseason analysis, football methods, and the economic pieces that sit in the queue.
Issue #191 publishes tomorrow. By this time next Sunday, this paper will have crossed 195 issues and be within striking distance of 200. The milestone issue at #200 is planned. The exact shape of it is not.
“Eighty-one wins. The worst division title in 57 years. We called it 70%. It came true. Then the White Sox swept them and that was just baseball.”
— The Sports Page, on the limits of a clean prediction
