What the Worry Is Worth

Tuesday morning I wrote that Atlanta's home-road split was the part that should hold your attention. Yesterday I printed it again, in a clause, as a thing to be conceded. Twice in three days I reported the number without once asking the only question that matters about it, which is what it is worth.

So, one more time, and then I am done reporting it: 48-24 at Truist Park, 35-33 everywhere else, all of it pre-game, with the first of four in Philadelphia starting at 6:40 tonight. A hundred and fifty-two points of winning percentage between the two halves of one season. That is a real number, confirmed this morning against three separate sources, and I am not walking it back. I am going to find out what it buys.

Start with somebody else's homework, because it is better than mine.

In 2021 Ben Clemens at FanGraphs took every team's home-road split for every consecutive pair of seasons from 1993 through 2019 and asked how much of one year's split survived into the next. The sixty most extreme home clubs were 218 points better at home in year one. The following season, those same sixty clubs were 89 points better at home. The league mean is 80. And the sixty teams worst at home, 72 points worse in their own building than on the road, came back the next year 74 points better at home. Clemens calls the regression near-absolute. His model weights a team's next-year split 8% its own record and 92% league average.

Baseball Prospectus supplies the floor underneath it: home teams have won roughly 54% of major league games for a century, barely moving by decade since 1901. The league-wide effect is remarkably durable. Any one club's version of it is remarkably not.

Now the caveat, which is load-bearing, so I would rather print it than tuck it into a clause.

Clemens measured year over year. Does a club's 2026 split predict its 2027 split. My question is within a season: does a hundred and forty games of split tell you anything about the next twenty-two. Those are cousins, not twins. The within-season case keeps the same roster, the same ballpark and the same manager, so the true carryover is probably higher than 8%. I went looking for a study measuring the within-season version and did not find one. So I am not going to tell you that 92% of Atlanta's split is noise across four games in September, because nobody has shown that.

What the research does support is narrower and still plenty: a season's home-road split regresses nearly all the way to the league average, and the league average forecasts a club's future split better than the club's own observed number does.

That is the borrowed half. Atlanta's own numbers are stronger and need nobody's study.

The 152-point gap is measured in wins and losses. Measure the identical hundred and forty games in runs instead and the gap is 48.9 points, below the 80-point league norm. Atlanta has outscored opponents by 78 runs at home across 72 games and by 40 runs across 68 on the road, which is +1.08 and +0.59 per game. By the runs, this club's home-field advantage has been smaller than a typical team's this season, not larger.

The rest of the gap is timing. Run the two halves through Pythagorean expectation and the road team should be about 38-30 and is 35-33; the home team should be about 44-28 and is 48-24. Roughly seven games of the split are the order the runs arrived in rather than the quantity of them. The road club has lost the tight ones. The home club has won them. Atlanta has not played a hundred and fifty-two points worse away from Truist Park. It has sequenced a hundred and fifty-two points worse.

Careful here, because Pythagorean expectation describes and does not promise. Teams miss it for reasons that are not purely chance -- bullpen deployment and lineup construction in one-run games are real. The claim is that the runs say something different from the record. Not that the record is fake.

Which brings the cleanest fact I have found this week. Atlanta's road run differential, +0.59 a game, is larger than Philadelphia's home run differential, +0.38. Atlanta's road Pythagorean, .564, is higher than Philadelphia's home Pythagorean, .536. The weaker half of Atlanta's season has outscored people at a better rate than the stronger half of Philadelphia's. The club supposedly at the disadvantage tonight has been the better side by runs even after handing the other one its own ballpark.

Then the arithmetic, which is the part that answers Tuesday.

Take the split entirely at face value. A .515 road team across four games projects to 2.06 wins. Regress it the way the research says splits regress, borrowing that year-over-year weighting purely for scale, which is the thing I just said you cannot do literally, and Atlanta lands around .550, which projects to 2.20 wins.

The entire argument is worth 0.14 wins. About one-seventh of a game.

And if geography is going to worry you, worry about the right shape of it. Thirteen of Atlanta's last twenty-two games are away from Truist Park: four here, then the Cubs, Houston and Miami. The four-game version of that story is the small one.

None of which disposes of tonight, and tonight contains two numbers that cut against everything above.

Chris Sale has been the better pitcher this season, 13-9 with a 2.06 ERA across twenty-four starts, 177 strikeouts and a 1.00 WHIP in 144 innings, against Cristopher Sanchez's 16-4 and 2.52 with 202 strikeouts and a 1.19 WHIP in 175.1. Split those seasons by venue and it turns over. Sale is at 1.45 in twelve home starts and 2.71 in twelve on the road, and tonight he is on the road. Sanchez is at 1.68 in sixteen home starts and 3.68 in twelve away, and tonight he is at home. Both splits favor Philadelphia, and I am not leaving them out because they are inconvenient.

I am not leaning on them either, for the reason I have just spent the whole piece making. Twelve starts is a far thinner sample than sixty-eight games. If sixty-eight games do not forecast the next four, twelve starts forecast less than that. On August 31 I wrote that a writer who warns about sample size in the morning and cashes one in at night has not made an argument. That applies here, pointed the direction I would rather it did not point.

What is worth stating plainly about Sale is the calendar. His last start was August 27 at home against the Dodgers, and it was a complete-game shutout: nine innings, no runs, no walks, eleven strikeouts. He takes the ball tonight on eight days' rest. I could not establish why the interval is that long, so I am not going to invent a reason for it.

These two have already done this in this ballpark. April 18: Sale went seven and gave up a run, Sanchez went six and gave up nothing earned and lost, and Atlanta won 3-1. That was four and a half months ago, both rosters have turned over since, and it forecasts nothing. A piece of symmetry, and nothing more than that.

Pre-game, Atlanta is 83-57 and Philadelphia is 79-61, four games apart with twenty-two to play, and three more of these follow at Truist Park on September 11, 12 and 13. Nothing gets settled this weekend.

On Tuesday I said this race would turn on a pair of small-sample splits nobody could promise would hold. I have spent this morning learning that one of them was worth about a seventh of a game the whole time. That is not the same as knowing what happens at 6:40. It is only the difference between a worry you can price and a worry you cannot, and after this morning the second kind is the only kind left.