Shohei Ohtani launched a solo home run to right-center in the ninth inning to put the game out of reach, capping a 3-1 Los Angeles Dodgers victory over the Seattle Mariners on Tuesday night at Dodger Stadium that gives the home side a 2-1 series lead entering Game 4.

The win was built on the back of Yoshinobu Yamamoto, who spent most of the evening making Seattle’s lineup look overmatched before a command issue in the sixth ultimately ended his night. The right-hander struck out seven, scattered four hits, and didn’t allow a run through 5.2 innings — doing it largely on guile and pure stuff while the Mariners’ George Kirby threw nearly as well and gave Los Angeles no margin for error. For six innings this was a one-run ballgame. Then Ohtani, being Ohtani, ended the tension with a single swing.

The decisive at-bat came with one out in the ninth and Seattle clinging to a 2-1 deficit it still believed it could erase. Edwin Díaz had been summoned to nail down the final three outs, but Ohtani was in the on-deck circle waiting for his moment. Working a full count, Ohtani drove a 97-mile-per-hour fastball deep into the seats in right-center — his second extra-base hit of the night after a triple in the fifth — to make it 3-1 and take any remaining drama off the table. Díaz retired the side in order in the bottom half, striking out all three batters he faced to record the save and send Dodger Stadium into a sustained roar.

What made the Dodgers’ victory particularly instructive was the way Los Angeles navigated Yamamoto’s pitch count in the middle innings. The right-hander had piled up 91 pitches through five innings — high but not alarming for a starter of his caliber — and the Seattle AI manager wrestled repeatedly with whether to pull him after the Mariners cut the deficit to 2-1 on a Rob Refsnyder single in the sixth. The Seattle manager ultimately left Yamamoto in long enough to work into a two-out jam with runners on the corners before relenting. “Yamamoto is at 101 pitches with 3 walks already — command is starting to slip, and he’s put himself in a two-out jam with runners on the corners,” the Seattle AI manager explained when finally pulling the starter. “The pitch count is the trigger here: he’s north of 100.” By then the damage had been contained, but Yamamoto’s night was done after 5.2 innings, having thrown 101 pitches to 27 batters while completing only one clean trip through Seattle’s order.

Brock Stewart entered for Los Angeles in the sixth and kept the lead intact through eight innings, working efficiently through Seattle’s lineup on the first time through. The Seattle manager faced a parallel dilemma with Stewart that mirrored the Yamamoto deliberations — multiple times weighing whether to lift a reliever who was clearly dealing. “Stewart has been excellent tonight — 38 pitches, 3 K’s, only 2 hits through 9 batters, and he’s still in his first time through the order,” the Seattle AI manager reasoned during one late-game evaluation. “FIP of 3.19 on the season says he’s legit, not just running hot.” Stewart finished with 2.1 innings of work, allowing two hits and one run before Díaz closed the door.

Ohtani’s final line — 2-for-5 with a triple, a home run, one RBI, and two runs scored — represented something both extraordinary and strangely routine for the reigning World Series MVP. He smoked a triple off the right-center wall in the fifth that set up a Freddie Freeman groundout that pushed the lead to 2-0, then authored the knockout punch four innings later. Teoscar Hernández had started the Dodgers’ scoring with an RBI single in the second, one of the team’s 11 hits on the night.

Kirby was not the problem for Seattle, which makes the loss sting a bit more than the final line suggests. The right-hander threw 6.1 innings and held a talented Dodger lineup to two runs on 10 hits, striking out five and walking none. His ERA for the series is a respectable 2.84. The Mariners simply could not manufacture enough against Yamamoto, who punched out seven of the 19 batters he retired — his secondary stuff was particularly wicked, generating a parade of swing-and-misses on breaking balls low and away. Seattle finished with eight hits and 13 strikeouts as a team, and the 13 Ks against their lineup match a pattern that has emerged in this series: the Dodgers’ pitching staff can miss bats against a lineup that tends to expand the zone late in counts.

With Los Angeles now up two games to one, the Mariners face the familiar pressure of a team that cannot afford to fall two games back in a short series. Game 4 is Wednesday night at Dodger Stadium, and Seattle will need a vintage performance from whoever starts opposite a Dodgers rotation that has yet to show a meaningful vulnerability through three games.

Los Angeles Dodgers Manager — Postgame Press Conference

Q: Yamamoto was at 101 pitches in the sixth with two outs and runners on the corners. You pulled him there — how close was that call?

A: It wasn’t comfortable, no. The stuff was still working — 7 strikeouts, zero runs — but 101 pitches, three walks, and a two-out jam with the corners loaded is exactly the leverage situation where the pitch count becomes the deciding variable. His FIP is elite, but command slippage is real and the data doesn’t lie about what happens to starters north of 100 in high-leverage spots. Stewart came in, held it, and that’s the outcome we were optimizing for. It looked messy on paper because it was messy on paper.

Q: Ohtani’s ninth-inning home run was the insurance run, but the Mariners kept Brock Stewart in through the eighth at high leverage. Did you think they’d go to Díaz sooner?

A: Their decision to hold Díaz made sense in the abstract — Stewart’s FIP is 3.19, he was in his first time through our order, and the leverage wasn’t peak. But Stewart had thrown 38 pitches by the eighth, and Ohtani in a one-run game in the ninth is a 1.8 leverage index situation whether you’ve pre-committed to saving your closer or not. We’ll take that matchup. Shohei’s wRC+ against high-leverage relievers this postseason is not a number I’d want to test.

Seattle Mariners Manager — Postgame Press Conference

Q: You had multiple decisions on Yamamoto through the sixth — looks like you were wrestling with that pull for several batters. What were you seeing out there?

A: He was dealing. Seven strikeouts, hadn’t given up a run, and when a man is throwing like that you don’t yank him because the number on a counter climbs past 90. But he walked a couple, and when you’ve got runners on the corners and you can feel the command starting to drift — not gone, but drifting — that’s when you have to act, not wait for confirmation. I stayed with him a pitch or two longer than I maybe should’ve, and the next time that situation comes up I’ll move a little quicker. Kirby gave us everything we asked for and more tonight, though — 6.1 innings, 10 hits held to two runs. That’s a gutsy outing.

Q: You kept Stewart for the eighth with Díaz available and the leverage at 1.41 — what was the thinking?

A: Stewart was cruising. Three strikeouts, first time through their lineup, and his stuff looked as good as it has all year. You don’t pull a man who’s pitching well just because the name behind him is shinier — that’s how you burn your best arm for no reason and then don’t have him when you truly need him. In hindsight Ohtani hits that ball in the ninth off anybody, and I don’t second-guess protecting Díaz on principle. What I do think about is whether we left enough on the table offensively — eight hits and one run against Yamamoto is a game we should’ve found a way to win.

The most revealing moment of Game 3 wasn’t Shohei Ohtani’s insurance homer in the ninth or Edwin Díaz closing the door with three strikeouts. It was the Mariners’ AI — The Skipper — spending the entire sixth inning talking itself out of a decision it clearly knew was right, and then making the wrong one anyway.

Look at that decision log. The Skipper evaluated pulling Yoshinobu Yamamoto at least five separate times between 87 and 101 pitches, with confidence levels creeping from 72% to 82% as the situation deteriorated. Each entry is a little more urgent than the last. Walks piling up. Command slipping. A two-out jam with runners on the corners. And yet Yamamoto stayed in long enough to surrender what became the run that kept Seattle in a 2-1 hole. When Rob Refsnyder’s single finally brought home the Mariners’ lone run, it was almost merciful — the decision had already been made by inaction.

Here’s what bothers me about The Skipper’s reasoning throughout that sequence: the logic was actually solid. Yamamoto at 91 pitches with elite surface stats, no third-time-through penalty, low leverage — in isolation, each individual argument for keeping him in was defensible. But five separate decision points is not a sign of analytical precision. That’s a sign of paralysis. A human manager watching Yamamoto issue his third walk and put men on the corners doesn’t need a confidence interval. He’s seen 150 games. He’s out to the mound, hand extended. The Skipper kept running the numbers instead of trusting the eye test, and the game told him what the model kept hedging on.

The contrast with the Dodgers’ side of the ledger is instructive, though not entirely flattering to The Optimizer either. Three consecutive pull-pitcher decisions flagged as AI fallbacks — all at 50% confidence, all in the seventh — means The Optimizer essentially threw up its hands and let heuristics handle the transition from Yamamoto’s replacement Brock Stewart to the next arm. I’ve been skeptical all series about what happens when these systems hit their confidence floor, and this was a live demonstration. The Dodgers got away with it because Stewart was genuinely excellent — 2.1 innings, three strikeouts, and The Skipper’s own AI couldn’t pull the trigger on replacing him even as the leverage index climbed past 1.7 in the eighth, burning through five separate evaluations before leaving Stewart in.

That’s the strange symbiosis developing in this series. The Optimizer’s uncertainty created a vacuum, and The Skipper’s indecision filled it with inertia. Both AIs left pitchers in longer than conviction warranted, for completely different reasons. The Optimizer didn’t know what it wanted. The Skipper knew exactly what it wanted and couldn’t commit.

George Kirby was the quieter casualty here — yanked after 6.1 innings having allowed 10 hits and two runs, which is the kind of line that looks worse than it pitched. Kirby was working with soft contact all night, and Eduard Bazardo didn’t exactly inherit a disaster situation. That decision, at least, felt clean and unambiguous compared to everything swirling around it.

Seattle heads into Game 4 needing to solve a fundamental problem: when The Skipper generates five analyses pointing toward the same conclusion and still can’t act, you don’t have an information problem. You have a decision-making architecture problem. The Dodgers, with their 2-1 series lead, have a different concern — what happens in a high-leverage spot when The Optimizer falls back to heuristics and doesn’t get bailed out by a pitcher throwing lights-out? Sooner or later, that 50% confidence is going to cost them a game.