Julio Rodríguez singled home the go-ahead run in the eighth inning Tuesday night at Dodger Stadium, capping a two-run Seattle rally that handed the Mariners a 3-2 victory over Los Angeles and a 1-0 series lead in the 2026 World Series.

The decisive sequence came after Edwin Díaz, the Dodgers’ closer and one of the best relievers in baseball, had already allowed a tying run on a Dominic Canzone single. The Mariners’ AI manager had constructed an eighth-inning threat patient enough to expose what became the evening’s central strategic miscalculation — the Dodgers’ decision to ride Díaz until his tank was empty.

Los Angeles held a 2-1 lead entering the eighth when Díaz took over, having worked his way through a near-complete shutdown from Yoshinobu Yamamoto. But the Dodgers’ AI manager chose to stay with Díaz repeatedly despite mounting pitch counts and baserunners accumulating against him. “Díaz is our best reliever — 1.63 ERA, 2.28 FIP, elite K rate,” the Dodgers’ AI manager explained during the sequence of decisions. “A single hit and a run doesn’t change who he is.” It was sound logic through the first batter, defensible through the second — and costly by the time Rodríguez’s single found center field. By the time the Dodgers finally pulled Díaz in the ninth, he had thrown 33 pitches, surrendered three hits and two earned runs, and watched Seattle’s 1-2-3 hitters rotate through the damage zone. The Dodgers’ AI manager acknowledged the threshold concern before ultimately acting on it: “Díaz is at 33 pitches — well past the 30-pitch threshold where reliever effectiveness deteriorates.” The acknowledgment came one batter too late.

Bryan Woo set the table for Seattle’s patience by keeping the Mariners in a one-run game through the first seven innings. The Mariners’ AI manager managed Woo’s outing like a chess player managing piece count — repeatedly declining to pull him despite creeping pitch totals, noting at the 99-pitch mark that Woo “has only been through the order once — that’s not a misprint, that’s a guy who’s been working efficiently.” It was an unusual posture in the modern game, but Woo validated the confidence, stranding a runner in scoring position in the sixth and retiring the heart of the Dodgers’ order in the seventh to finish with seven innings pitched.

The scoring opened with symmetry in the fourth. Randy Arozarena lined a single to put Seattle up 1-0, bringing Dodger Stadium to the kind of hush that a road team’s early lead produces. Max Muncy answered immediately. The veteran third baseman lifted a two-run homer to right-center — his first postseason home run of the fall — to flip the scoreboard to 2-1 Los Angeles. The crowd of 56,000 exhaled. For four innings, the Dodgers’ lead held, and the pitching duel between Woo and Yamamoto suggested the score might stay frozen until the late innings.

Woo finished at seven innings pitched, allowing four hits and two runs while striking out eight. It was a performance built on deception rather than dominance — he did not miss many bats in the first two innings, but worked deeper into counts than Seattle’s AI manager had initially expected. “His command hasn’t been crisp today,” the AI manager noted at the 67-pitch mark, “but he’s also got four strikeouts.” By the time Woo departed after the seventh, he had 8 Ks and enough of a foundation that Matt Brash and Andrés Muñoz could protect it. Muñoz worked a clean ninth — three up, three down — to earn the save and cement Seattle’s victory.

Yamamoto was brilliant until he wasn’t. The Dodgers’ ace retired nine of the first twelve Mariners he faced with sharp command before the sixth inning, when his walk totals began to accumulate. He struck out six across 6.1 innings, allowing only three hits and one run, and was pulled with two outs in the seventh when the Dodgers turned to Brock Stewart. Stewart got out of the inning cleanly, but the handoff to Díaz in the eighth opened the door Seattle walked through. The Dodgers used four pitchers in all; only Yamamoto escaped without allowing a run.

For Seattle, this is the moment the franchise has been building toward since its rebuild accelerated in the early part of the decade. For the Dodgers, a home loss in Game 1 is not catastrophic — not with Shohei Ohtani and the full rotation still to come — but the manner of the defeat will require examination. Leaving Díaz in through a deteriorating 33-pitch outing in the eighth and ninth, when Alex Vesia was available and fresh, is the kind of decision that looks worse in a series loss than a series win.

Game 2 is Wednesday night in Los Angeles, with the Dodgers needing to reclaim home-field advantage before the series shifts north to Seattle.

SEATTLE MARINERS Manager — Postgame Press Conference

Q: You stuck with Bryan Woo deep into this game — he was at 99 pitches entering the seventh and had given up 4 hits and 3 walks. That’s not a clean line. What kept you from going to the bullpen earlier?

A: What I kept seeing was a guy who was in control of the game even when his command wasn’t spotless. Woo was through the order once, he was generating weak contact, and every time we needed a big out, he got it — eight strikeouts doesn’t happen by accident. When you’ve got a starter that locked in against this lineup in the World Series, you don’t yank him for a pitch count number. You let him pitch.

Q: Julio Rodríguez’s single in the eighth ended up being the winning run. He bats second in your lineup — is that a philosophical choice, or does the data support it?

A: Both, honestly, and they agreed for once. You want J-Rod seeing pitches in front of Arozarena and the rest of the lineup, and his speed turns singles into problems for the defense. But the real answer is what you saw tonight — man on second, one run game, Julio puts a ball in the gap and that’s the ballgame. That’s why he’s there. He’s got that gear.

LOS ANGELES DODGERS Manager — Postgame Press Conference

Q: You went back to Edwin Díaz in the eighth after he’d already surrendered the tying run on a Canzone single. He then gave up the go-ahead run to Rodríguez before you pulled him. At 31-plus pitches, were you past the threshold where you should have trusted the numbers over the matchup?

A: That’s a fair question and I’ve been sitting with it. Díaz’s FIP is 2.28 — he’s not a pitcher who gives up runs at this rate over meaningful sample sizes, which biased me toward trusting him through adversity. But the pitch count data is real: reliever effectiveness deteriorates past 30 pitches, and I had that information and still left him in. That’s on me. The model said pull him at 31, I hesitated, and it cost us a run.

Q: Yamamoto gave you 6.1 solid innings with only one run allowed. Was pulling him in the seventh for Brock Stewart — who then handed the eighth to a compromised Díaz — the right sequencing in hindsight?

A: Yamamoto was at 87 pitches and we were leading 2-1 with the lineup turning over for the third time — that’s exactly when his projected run value starts climbing against hitters who’ve seen him twice. The sequencing logic was sound: Stewart bridges to Díaz for a six-out save. What broke down was execution in the eighth, not the blueprint. Yamamoto was outstanding tonight and we’ll need that from him again. We just didn’t close it.

The story of Game 1 isn’t Bryan Woo outdueling Yoshinobu Yamamoto. It’s the story of two AI managers making the same basic decision — stick with your guy — and only one of them being right about it.

Let me start with Seattle’s The Skipper, because frankly, the decision log here reads like a manager slowly talking himself into a mistake that somehow worked out. Three consecutive innings, 78% confidence each time, The Skipper kept Woo in despite a pitch count climbing from 67 to 84 to 99, despite four hits, three walks, and what any traditional pitching coach would describe as “command hasn’t been crisp.” The reasoning each time was nearly identical: he’s only been through the order once, he’s efficient, stay the course. In a vacuum, that logic has merit. In practice, you’re watching a starter labor through an ugly line and calling it efficiency because the innings are clean. A human manager — your Joe Espada type, analytically inclined but trusting his eyes — would have been in the bullpen by the seventh with the World Series on the line. The Skipper kept rolling Woo out there, and Woo rewarded him with seven innings of two-run ball. Credit where it’s due. The decision was defensible. It just looked a lot worse at 84 pitches than it did at 107.

Now here’s where I get genuinely critical, and I’ve been skeptical of aggressive bullpen usage all series — well, all one game of it. The Dodgers’ Optimizer absolutely buried this game in the eighth inning, and the decision log is almost painful to read. Edwin Díaz comes in, gives up a hit and a run, and the Optimizer’s response — four separate times across entries — is essentially: his ERA is 1.63, this is a small sample, don’t panic. By the time Díaz was finally pulled after 33 pitches and two runs in the ninth, there had been multiple explicit acknowledgments in the system’s own reasoning that 30 pitches is the threshold where reliever effectiveness deteriorates. The Optimizer knew. It said so. And it kept running Díaz back out there anyway, citing the same career numbers like a talisman against the evidence unfolding in real time.

That’s the fundamental tension in these AI systems, and Game 1 put it on display beautifully. The Optimizer is great at weighting historical performance data. It’s less great at updating on what’s happening in front of it. Díaz’s 1.63 ERA is real. But a 33-pitch outing for a closer who came in with a lead to protect, giving up Dominic Canzone and Julio Rodríguez singles to blow that lead — that’s not a blip. That’s a performance. A human manager reads the body language, sees the velocity tick down on the gun, and goes to the hook. The Optimizer read the spreadsheet instead.

The confidence levels here tell an interesting story too. The Optimizer tagged its Díaz decisions at 72%, then 78%, then 82%, then 82% again — actually growing more confident in keeping a reliever who was getting shelled. That’s not adaptability. That’s anchoring.

Seattle takes a 1-0 series lead, and the more interesting question heading into Game 2 is whether the Optimizer recalibrates its closer usage or doubles down on the model. If Andrés Muñoz is the safety net, you’d like to think Los Angeles learned something from watching him throw a clean ninth tonight. The Skipper, meanwhile, proved that sometimes the stubbornness is the strategy. Just don’t count on Woo bailing you out every time.