The 0.64 Run Gap Hiding Inside A Team Total

Measuring The Gap Instead Of Assuming It

It would be easy to wave at a pitcher's ERA and call it irrelevant to his own team's bat total, since he does not face his own lineup. That reasoning is correct as far as it goes, but it skips a real question: does a team actually score at a different rate on the days its ace starts, for whatever bundle of reasons, be it lineup construction, game states that end innings early, or simple variance? The only way to answer that is to pull the specific dates and check, not assume the answer either way.

We pulled Arizona's full hitting game log for the 2026 season and matched it against Soroka's 15 start dates individually.

Fifteen games is not a large sample, and a 0.64 run gap sits inside the range noise alone can produce over that many games. But it is not nothing either, and it runs in one direction consistently enough to note: Arizona's run-scoring range in Soroka's starts includes an 0-run and a 1-run game, both from the individual game log, alongside a pair of 9-run outbursts that pull the average back up. The distribution is wide, not just low.

The Case For The Over Anyway

The number that argues past the gap is on the other side of the matchup. Colorado's team ERA is 5.48, one of the worst full-staff marks in the league, and its starter tonight, Gabriel Hughes, has made only five starts this season for a 4.25 ERA over 29.2 innings. That sample is too thin to treat as a stable estimate of true talent, and his last outing, on August 4, went 5.1 innings with five earned runs allowed.

Recent form, though, complicates the over more than the season table suggests. Arizona has scored 10, 1, 4, 1 and 4 runs in its last five games, clearing 4.5 in exactly one of those five. A team can be a good full-season offense and still be running cold at the moment the bet is placed, and the last five games say Arizona is running cold.

A Small Sample Pitcher Line, Decomposed

Freddy Peralta's season ERA is 5.37, and treating that as a single number obscures more than it reveals. It is the blend of two separate pitching samples that do not belong in the same average.

Twenty-two starts is a real sample and 4.99 is a fair estimate of what that version of Peralta produced. One start is not a sample at all, it is a single data point, and a seven earned run outing in 3.2 innings is exactly the kind of outlier that a one-game estimate cannot be trusted to represent. The honest read is that Peralta's true talent level tonight is closer to his 22-start Mets number than to either extreme, and treating the blended 5.37 as predictive would be a mistake in the other direction from ignoring it entirely.

One additional split deserves a mention before trusting that gap fully. The Athletics hit far better at Sutter Health Park, where tonight's game is played, than they do on the road: 5.00 runs per game and a .790 OPS at home, against 3.69 runs and .644 on the road. The full-season run differential describes an away-heavy sample that will not be on the field tonight.

Where This Card Contradicts Our Own Projection

Honesty requires leading with the disagreement rather than burying it. The Dodgers run line at -1, priced -202, is on the card. Our Atlas v1 model projects Los Angeles at 3.93 runs tonight, and the only signal it produced at a qualifying edge was the Dodgers under 4.5. That is a direct disagreement, generated by our own system, with laying a number that requires the Dodgers to win by two runs or more.

The recent game log deepens rather than resolves the tension. Los Angeles has scored 2.8 runs per game across its last five outings and is 2-8 over its last ten, a meaningfully colder stretch than its 70-48 season record implies. A projection of 3.93 runs is not a projection of a blowout, and a run line that requires a multi-run margin is betting against the version of this offense that has shown up over the last week and a half.

The pitching matchup is the argument that exists outside the model. Tarik Skubal posted a 2.79 ERA across 16 Detroit starts, with 116 strikeouts against a 0.91 WHIP over 96.2 innings, one of the sport's best pitching lines, and his single Dodgers start on August 4 held: six innings, two earned runs, six strikeouts. Kansas City's Noah Cameron sits at 4.37, and the Royals' road offense is thin at 3.58 runs per game and a .671 OPS. The case for -1 is built on Skubal shutting the door, not on the Dodgers' bats erupting, and a shutdown start plus a modest Dodgers output is a two-run game, not a blowout, which is exactly the scenario the alternate line is designed to price.

What The Decomposition Does And Does Not Buy

Every regression in this piece is small in the statistical sense. Fifteen starts, five games, twenty-two starts against one. None of these samples are large enough to treat their point estimates as certainties, and confounders, quality of opponent, ballpark, lineup health, are not controlled for in any of the tables above. What decomposition buys is not certainty. It buys the ability to see when a season aggregate is quietly averaging together two different situations, which is the single most common way a headline number misleads a bettor working from box scores alone.

Every rate, split and game log figure in this piece was pulled live from official statistics during this session, cross-referenced by date where a conditional claim required it. The Atlas v1 projection is reported exactly as the system produced it, including where it disagrees with the card.