Tommy Edman homered twice and Landon Knack delivered the finest pitching performance of this World Series, as the Los Angeles Dodgers beat the Seattle Mariners 5-1 on Tuesday night at Dodger Stadium to even the Fall Classic at two games apiece.

The Dodgers entered Game 4 facing elimination after dropping two of the first three games. Instead, they got a clinic — methodical, dominant, the kind of performance that strips a road team of its momentum and hands the narrative back to the home side. If Seattle arrived in Los Angeles believing it could close this series out on the road, Knack and a suddenly thunderous Dodger lineup answered that notion with four home runs and 6⅓ innings of near-immaculate ball.

The fifth inning was where Los Angeles put the game away and where the Mariners’ season nearly ended. After Knack had held Seattle hitless through four frames, Edman led off the top of the fifth with his second homer of the night — a solo shot that made it 3-0 — and then Shohei Ohtani sent one into the left-field pavilion two batters later to push the lead to four. Freddie Freeman capped the inning with an RBI double, and suddenly a scoreless game had become a rout in the time it took to empty a bullpen. Seattle’s AI manager had already been forced into a pitching change. “I don’t like pulling Logan at 58 pitches — under normal circumstances I’d ride him deep into this game,” the Mariners’ AI manager said of the decision to yank Logan Gilbert mid-inning. “But we’re down four runs in the fifth inning of a World Series game, and that changes the calculus entirely.” Gilbert’s final line — 4.0 IP, 6 H, 4 R, 4 K — reflected a night when the Dodger offense simply wouldn’t let him find rhythm.

The story on the other side of the ball belonged entirely to Knack. The Dodgers’ right-hander, whose 4.89 ERA and 5.71 FIP made him an unlikely Game 4 starter, was something altogether different on Tuesday. He allowed just one hit through 6⅓ innings, struck out seven, and navigated four walks without once letting Seattle seriously threaten. The Mariners’ AI manager wrestled repeatedly and transparently with whether to lift him. “Knack has been outstanding — 1 hit, 0 runs, 6 K through 19 batters on only his first time through the order,” the Seattle AI explained after considering a sixth-inning hook. “The 4 walks are a concern and align with his elevated FIP, but his results today are undeniable.” The Seattle bench ultimately left Knack in through the sixth, then reconsidered in the seventh when the pitch count hit 101. “He’s well past my 95-pitch hard leash,” the Mariners’ AI reasoned. “His season numbers already make him one of our weakest starters, and while his line tonight is outstanding, the walk rate represents real danger.” But by then the damage was done — Knack had authored the signature pitching performance of the series.

Edman was the offensive engine from the first pitch. The second baseman opened the scoring in the top of the first when Mookie Betts drilled a leadoff home run to right — but it was Edman who punctuated the evening, homering in the third and again in the fifth to finish 2-for-4 with two runs and two RBI. Betts was equally relentless, going 2-for-3 with a homer, an RBI, two runs and two walks in a vintage performance from a player who has made Octobers uncomfortable for opposing pitching staffs throughout his career. Ohtani’s solo blast in the fifth — his second home run of this series — was the kind of swing that makes the rest of the lineup dangerous by proximity: when the pitcher has to worry about everyone, even Ohtani’s quieter nights carry weight.

Ben Casparius took over for Knack in the seventh and locked down the final 2⅔ innings without incident, allowing only a ninth-inning run-scoring single from Julio Rodríguez that prevented Seattle’s shutout. Gilbert absorbed the loss, and reliever Alex Hoppe worked three innings of scoreable but ultimately inconsequential ball for the Mariners after Gilbert’s early exit.

The series now returns to its original state: even at two games, winner-take-all pressure building toward a Game 5 on Wednesday night. Seattle still controls nothing — it has simply lost its opportunity to close the series in four. Los Angeles, for the first time since Game 1, looks like the team with the momentum.

Los Angeles Dodgers Manager — Postgame Press Conference

Q: Landon Knack had a 4.89 ERA and a 5.71 FIP coming in. You kept him in through 6.1 innings tonight against a Seattle lineup that was seeing him for the first time. Walk us through the decision to let him work that deep.

A: The FIP was a real number — I wasn’t ignoring it. But context matters: Knack had only been through the order once, the four walks were inflating his batter count without actually producing damage, and his stuff was generating swings and misses at a rate that didn’t match the season-long profile. At the point we pulled him, he’d allowed one hit and zero runs through 101 pitches. The model flagged the pull at 92% confidence, and frankly, I agreed — 101 pitches is past where you want a guy like Knack in October regardless of how sharp he looks. We got everything we could have asked for from him tonight.

Q: Tommy Edman hit two home runs, including one in the fifth that broke the game open. He wasn’t your most heralded bat coming in. What were you seeing from him that told you he’d be a factor?

A: Edman’s wRC+ against right-handed pitching this postseason has been quietly excellent — he was generating hard contact in batting practice all week and his sprint speed numbers were up, which usually correlates with him being locked in physically. Gilbert is a talented starter, but his four-seam location drifted up in the zone in the third and fifth innings, and Edman punishes elevated fastballs. Two home runs, two runs scored — the series is tied 2-2 going home, and that’s entirely because of what he did tonight. Sometimes the numbers point at a guy and the game confirms it.

Seattle Mariners Manager — Postgame Press Conference

Q: Logan Gilbert was pulled at 58 pitches in the fifth inning with Seattle down four runs. You’ve been protective of your starters all postseason. What happened out there?

A: Logan just didn’t have it tonight — simple as that. Six hits, four runs, he was leaving the ball over the plate and those Dodger bats were not missing it. At 58 pitches I’m not thrilled about going to the pen that early, and I’ll tell you honestly that wasn’t an easy call. But I’m not going to let a guy twist in the wind when the game’s already getting away from us in the fifth. You protect your starter’s confidence too, not just his arm. Logan’s going to bounce back — he’s too good a pitcher for one night to define him.

Q: Your manager logs show some back-and-forth on whether to pull Knack in the sixth versus the seventh — the model actually flagged him at 90% confidence to pull in the sixth and you waited. Do you have any second thoughts about letting him work through 101 pitches?

A: Look, I saw what I was seeing out there — one hit, no runs, seven strikeouts. You don’t yank a man who’s throwing a gem because a number on a screen doesn’t trust his career ERA. That said, I’ll be straight with you: we’re talking about it in there right now. The model had a point about the walk rate and what happens the second time through an order with a guy who’s been inconsistent all year. The ninth-inning run was consolation — doesn’t change what happened in the fifth. We still lead this series 2-2 going back to Seattle, and I like our ball club at home.

The most fascinating story from Game 4 isn’t that the Dodgers evened the series with a dominant 5-1 win — it’s that Seattle’s AI manager, The Skipper, spent three separate decision cycles talking itself into and out of pulling Landon Knack, ultimately landing on the right answer, but only barely, and for reasons that reveal something genuinely interesting about how these systems reason under pressure.

Let’s start with the elephant in the room: The Skipper ran the Knack decision three times. At the end of the sixth, it logged two separate pull_pitcher evaluations at 90% and 92% confidence, landing on the same conclusion both times — keep him in. Then in the seventh, it ran the decision again at 92% confidence and finally pulled him at 101 pitches. I’ve covered baseball long enough to know that no human manager agonizes out loud like this, but the reasoning is actually sound. Knack’s season numbers are ugly — 4.89 ERA, 5.71 FIP — but he’d been untouchable on the night, and The Skipper correctly identified that one trip through the order doesn’t tell you much about a pitcher’s vulnerability to a lineup that’s seen him multiple times. The problem is, by the time The Skipper finally acted, Knack was well past a reasonable exit point. He finished at 6.1 innings having allowed just one hit, so the decision looks defensible in retrospect. But any human pitching coach watching those four walks accumulate would have had a shorter leash on a guy with Knack’s underlying numbers. You don’t let a high-FIP pitcher nibble his way to 101 pitches in a World Series game because his ERA for the night looks pretty.

Seattle’s other major decision — pulling Logan Gilbert at just 58 pitches in the fifth — is where The Skipper deserves real credit. Down four runs with the middle of the Dodgers’ order coming up, The Skipper articulated exactly why you don’t let a starter absorb garbage-time innings when the game is functionally over. That’s sound bullpen management, even if Alex Hoppe couldn’t stop the bleeding. The Dodgers tagged him for three more runs in the fifth anyway, turning a competitive deficit into a rout. This wasn’t a case of the decision being wrong — sometimes the fire spreads no matter how fast you act.

On the Los Angeles side, The Optimizer’s management was largely invisible in the best possible way. It set a sensible lineup, got a career night out of Tommy Edman — two home runs, including a mammoth second-inning shot that proved to be the turning point — and correctly identified low-leverage situations in the late innings to protect Ben Casparius for more meaningful work. What caught my eye was the confidence level on the opening lineup decision: 78%, with an AI fallback flag indicating the primary system had already been taxed. When you’re getting that kind of production from Edman and Mookie Betts and Shohei Ohtani in the same game, lineup construction matters less, but I’d be curious whether a fully-functional Optimizer would have deployed the same batting order.

Here’s what I keep coming back to heading into Game 5: The Skipper won this series through two games on aggressive, unconventional decision-making. In Game 4, it vacillated. The multiple Knack deliberations, the delayed exit, the conservative pitch-count threshold that kept getting overridden by “but his results are good tonight” — that’s a system that got caught between its process and its in-game evidence. Seattle still leads 2-1, but the Dodgers now have momentum and a home crowd for what should be the pivotal game of this series. If The Skipper is going to close this out, it needs to pick a lane and stay in it.