New York's 4–3 overtime win was built against the run of play: the Islanders won despite a 5.73–3.92 expected-goals deficit, a 59–41 split decisively favoring Vegas. This was a heist.

I. Sorokin stole it. The Islanders goalie saved 2.73 goals above expected on 26 of 29 shots (0.897 SV%), turning what should have been a Vegas blowout into a fight to the wire. Vegas underperformed its own xG by 2.73, squandering the dominant quality it generated.

The Golden Knights controlled the game. They out-shot New York 29–24 and out-chanced them 14–7 on high-danger opportunities. Vegas created 9 rebound chances and converted one—they were crashing the net effectively. A. Schmid made 20 of 24 saves (0.833 SV%), but the problem wasn't his save percentage; it was Vegas's inability to convert its chances at a normal rate.

New York's path to victory defied logic. The Islanders scored 4 goals from low-danger spots, a mark that doesn't sustain. They went 0 for 7 on high-danger chances and won anyway—only because I. Sorokin was exceptional (93% save rate on 14 high-danger shots). Both teams thrived off the rush: Vegas scored 2 on 30 attempts; New York scored 3 on 31.

Jean-Gabriel Pageau ended it at 03:02 of overtime with a snap shot from the top of the circles, the kind of opportune finish that summarized New York's night: right place, right moment, and the goaltender beaten.

The model projects a 2.3–2.7 goal Golden Knights victory with Vegas at 57% win probability. The edge stems from underlying dominance and regression tailwinds.

Vegas owns the underlying play. The Golden Knights generate 54.0% expected goals at 5v5 versus the Islanders' 48.0%, and control territorial momentum at 52.0% Corsi. This is not a marginal advantage—Vegas will spend the night pinning New York in their defensive zone, and the Islanders' inability to establish offensive rhythm compounds the problem.

The deeper advantage lies in PDO. Vegas runs 0.983, well below league neutral, while New York sits at 0.993. Translation: the Golden Knights are underperforming their chances through a combination of finishing variance and bounces. Neutral PDO would sharpen their expected-goals edge into something decisive. The Islanders have no such buffer; they're already trending toward their underlying luck.

Anders Lee is the Islanders' offensive engine, generating 1.20 expected goals per 60 minutes with 9 goals and 2 high-danger chances. His 6.4% shooting rate is sustainable for a high-volume threat but leaves room for downward regression. Bo Horvat wields a more dangerous shooting hand: 12.8% on 0.87 xG/60. That rate rarely holds; his 18 goals reflect finishing outperformance on modest volume. Neutralizing Lee is the priority; Horvat's output self-corrects if territorial control tilts toward Vegas, which the underlying numbers predict it will.

For Vegas, Brandon Saad is marked inactive, removing a 0.91 xG/60 weapon. Tomas Hertl takes on expanded opportunity at 0.88 xG/60 but is running cold at 4.3% shooting on 5 goals—another regression tailwind for the Golden Knights.

New York's defensive architecture features Alexander Romanov as the structural weakness, while Vegas identifies Rasmus Andersson in that role. Neither team generates meaningful scoring off high-danger chances or rebounds, making territorial control and volume the decisive factors. This is a possession game, and Vegas projects to own it. The model's lean on the Golden Knights stands.