What a hot finish is worth in October

I wanted to know whether the Padres’ strong finish should change what I expect in October. I keep hearing they’re the hottest team in baseball, and I want that to matter. In 2022, they beat the Dodgers, then lost to the Phillies. I was there for the Phillies series. This year, I looked at the history to see how much confidence a finish like this should give me.

On Oct. 2, the day before the Division Series, Brewers manager Pat Murphy described the Padres this way:

Writers had been saying it for weeks:

• “Winners of seven straight, the Padres are the hottest team in baseball.”D.J. Short, NBC Sports, Sept. 14. Source
• “The top of the NL is tough this year, but no one’s hotter than the Padres.”Will Leitch, MLB.com, Sept. 20. Source
• “Fernando Tatis Jr. is the engine driving the Padres, baseball's hottest team, into NLDS showdown vs. Brewers”Headline, CBS Sports, Oct. 2. Source

A strong finish can tell us a team is good, or that it’s improved. Momentum is a separate claim: that recent wins improve its October chances beyond what its current ability would predict. That’s the part I wanted to test.

I built forecasts for postseason matchups from 1995 to 2025, then tested whether adding each team’s recent record improved them. I set the definitions and test before scoring the results.

The 2022 run is a useful reminder that winning in October and arriving hot are different things. The team with the weaker finish won both of those series. The Padres finished 16-14 over their last 30 games, beat a Dodgers team that finished 20-10, then lost to a Phillies team that finished 14-16.

The Padres became a better team after the break

San Diego reached the All-Star break at .500, having allowed more runs than it scored. It fell to 50-53 on July 23, then went 41-18.

Exhibit 1 follows the season through a sliding 30-game window, the same length I use to define a hot finish. Each point is the record or run differential over the 30 games ending that day, so neighboring points mostly overlap. The vertical lines mark the break and two roster moves. They show timing, not cause.

San Diego was winning more and outscoring opponents by more. The improvement wasn’t just a better record in close games: the Padres went 13-10 in one-run games after the break, close to its 13-12 before it, against opponents of about the same strength.

The Padres’ run differential per game improved by 1.7 runs from before the break to after it. Of the 275 playoff teams from 1995 to 2026, leaving out 2020, only the 2001 Athletics and the 2017 Twins improved more.

To see where the improvement came from, Exhibit 2 compares the offense, rotation, and bullpen before and after the break. It pairs actual results with what the underlying performance would predict: expected runs from hits, walks, and outs for the offense; FIP, an ERA-scale measure based on strikeouts, walks, and home runs, for pitchers. Agreement between the two gives me more confidence in the change.

The offense drove most of the improvement. San Diego scored 3.9 runs a game before the break and 5.2 after, and the runs its underlying production would predict rose by the same amount. The lineup went from below the league average to well above it.

The bullpen was good all season and better after the break. Its FIP fell to 3.26, against 4.11 for the league’s relievers. That’s the part of Murphy’s description the numbers support most clearly.

I have less confidence in the rotation’s improvement. Starters’ ERA fell by almost a run, but their FIP rose slightly and remained worse than the league average. The lower ERA depended on factors FIP leaves out, including hits on balls in play and the order of events. Robbie Ray, acquired on Aug. 3, had a 5.02 ERA in nine starts for San Diego.

There’s enough here to rate the Padres more highly than in July. The next question is whether the winning streak deserves any additional credit.

The hotter team has won half the time

I compared playoff teams from 1995 through 2025, excluding 2020’s expanded bracket: 30 postseasons and 244 matchups. Each matchup counts once, whether it was a single wild-card game or a seven-game series.

In 219 of those matchups, one team had the better record over its final 30 games. The hotter team won 109, almost exactly half.

Team strength was a better predictor. I rated each team by its run differential, adjusted for the opponents it played and weighted toward recent games. The team with the better rating won 57% of matchups. Playoff teams are close to one another, and a short series is noisy.

That comparison alone can’t isolate momentum. Teams differ in strength, opponents, home-field advantage, and series length. I needed to account for those differences before testing whether a strong finish added anything.

A hot finish added no clear predictive value

The baseline forecast uses each team’s rating, home-field advantage, and the series format to estimate its chance of winning. I then added the final-30 record and compared the forecasts.

To score a forecast, I used the Brier score: the squared gap between the probability and what happened, averaged over matchups. A confident miss costs much more than a cautious one. Lower is better, and saying 50% every time scores 0.25.

I fit the models on 1995 to 2019 and scored them on 2021 to 2025, seasons excluded from fitting. The top panel of Exhibit 3 shows that test. The bottom panel uses all 30 postseasons to estimate how much a hot finish was worth, in points of series probability.

The held-out test can’t separate the models. Adding the final-30 record changed the score by less than a hundredth, and the interval runs from a small gain to a small loss. The sample is too small to detect a modest benefit reliably: when I added an effect of 3 points to simulated seasons, the same test detected it only 7% of the time.

Using all 30 postseasons gives a more useful estimate of the effect’s size. A team that won about three more of its last 30 games than an otherwise equal opponent had an estimated five-game series win probability about 2 percentage points lower than its strength predicted.

The 95% interval, the range of effects consistent with the data, runs from 6 points worse to 2 points better. Before scoring anything, I’d set 3 points as the smallest effect worth caring about, about what two extra wins of season-long quality are worth in a five-game series. The interval excludes it. Measured by the last 15 games, by run differential, or by the record after the break, the estimate also came out at or below zero.

I wouldn’t read the negative estimate as evidence that winning hurts. The interval includes zero and a small positive effect. It does exclude the benefit I’d defined as meaningful before running the test.

Earlier work points the same way. Jay Jaffe found “a whole lot of nothing, an essentially random relationship between recent performance and first-round success” for 1995 to 2008, and a 2014 FiveThirtyEight analysis that accounted for full-season records found late-season heat “was not statistically significant when it came to forecasting playoff proficiency.”

From 2005 to 2025, I also forecast each postseason using only earlier seasons. The hottest third of playoff teams, roughly those that won 20 or more of their last 30 games, won 51 matchups. The forecast that ignored how they finished expected 51.

The result held up in two further checks. In the last eighth of the 1995 to 2014 regular seasons, 6,106 games, a team’s last 30 games didn’t improve the rating’s forecast of what came next. And from 2009 on, when complete playoff rosters are available, the final-30 record added no clear value once I accounted for the players on each roster.

The rating already gives September more weight than April. Recent performance matters through that estimate of team strength; these tests give me no clear reason to add a separate momentum adjustment.

The closest historical matches were good teams that fell short

I also looked for historical playoff teams with a similar profile to these Padres. I matched on seven regular-season traits: overall strength, final-30 record, improvement after the break, bullpen, rotation, offense, and how much of the offense came from the top three hitters. Outcomes didn’t affect which teams were picked or their order.

None of the eight won the World Series. The 2007 Rockies, who went 22-8 over their final 30 games, reached it and lost. Three lost in the wild-card round, including last year’s Padres, and three lost in the LCS.

That’s too small a sample to estimate San Diego’s title odds. Across all 274 playoff teams, 30 won the title, about one in nine, so eight ordinary playoff teams would produce about one champion. These produced none, which is within what chance allows. The selection also depends heavily on improvement after the break. Leave that trait out and five of the eight change.

San Diego has about one chance in three against Milwaukee

Milwaukee is the stronger team entering the series on Oct. 3. The Brewers won 103 games and outscored opponents by 214 runs, both the best in baseball.

They were also at least as hot. San Diego had the better record over its last 20 games, 17-3 to 15-5, but Milwaukee matched or beat it over the last 15, 18, and 30, and after the All-Star break. “Hottest team in baseball” depends on where you start counting, which is why I fixed the window before looking.

ESPN’s Bradford Doolittle made the same comparison: “As hot as San Diego was down the stretch, in terms of team temperature, the Brewers were the hottest team in baseball when the postseason opened.”

I froze the forecasts at 4:53 p.m. Pacific on Oct. 2, before either team announced its Division Series roster. Exhibit 5 shows the same matchup under each model. They’re alternative forecasts, not pieces to add together.

Every version lands between 31% and 36%. Adding the final-30 record raises San Diego’s chance by about a point, only because the fitted effect of heat is slightly negative and Milwaukee was hotter. I wouldn’t read anything into that. The roster version is a little lower because Milwaukee’s regulars had better seasons by these measures.

The pitching matchups help explain the challenge. Robbie Ray starts Game 1 against Jacob Misiorowski, who went 16-5 with a 1.80 ERA. Michael King faces Logan Henderson in Game 2. San Diego’s best argument is the one Murphy made: a bullpen that was better than the league all year. Its weakness is a rotation whose underlying numbers trailed the league.

I also have some reason to question how strongly the models favor Milwaukee. From 2005 to 2025, forecasting each season with only earlier seasons, favorites given 60% to 70% won 31 of 54, less often than predicted. I’ve kept the frozen forecast, but that history limits how much confidence I put in the exact probability.

The Padres are a better team than they were in July, and that gives me a reason to be optimistic. The offense and bullpen improved. The rotation is still the concern. If Ray and King can give that bullpen leads to protect, I like their chances.

Milwaukee should be favored. But about one chance in three leaves plenty of room for the Padres to win. In 2022, they beat a Dodgers team that had won 111 games. I’m hoping they can do it again.