The Sabres rolled past the Lightning 6-2 in a game the underlying numbers suggested was far closer. Tampa held 52% of shot attempts (38–35), a 4.71–4.51 expected-goals edge, and a 9–4 advantage in high-danger chances. Yet they lost by four. Goaltending determined this gap entirely.

U. Luukkonen stopped 36 of 38 shots with a 0.947 save percentage, outperforming his expected save total by 2.71 goals. A. Vasilevskiy, Tampa's backup, conceded 9 goals on 14 shots—a 0.643 save percentage that underperformed expected outcomes by 3.20 goals. J. Johansson, Tampa's starter, actually outperformed expectations by 1.71 goals, but one backup's elite performance and another's catastrophic collapse determined the final margin far more than play at even strength.

Buffalo was lethal off the rush, converting 6 of 34 chances. Tampa managed only 2 of 33, despite being the team that dominated possession and chance quality. The Sabres also scored 3 goals from low-danger areas—plays the opposing goalie typically stops. Luukkonen's performance masked what was, for most of the game, a team that controlled neither possession nor shot quality.

This was a game where the scoreboard and the underlying data diverged sharply—a 6-2 final that rewarded execution in one dimension while ignoring evidence across nearly every other metric.

The model favors Tampa Bay at 56% win probability, projecting a 2.4-2.7 goal margin. But the forecast hinges on a critical caveat: both teams are running on unsustainable PDO luck.

Start with underlying play. The Lightning command the game at 54.0% expected goals compared to Buffalo's 50.0%, translating to 53.0% Corsi dominance. That territorial advantage—Tampa will spend the night in Buffalo's end—underpins the 56% call. But it's modest, and the model knows it.

The real edge comes from the regression opportunity. The Lightning's PDO sits at 1.019; the Sabres' at 1.020. Both teams are scoring and preventing at rates beyond what their shot profile allows. That gap always closes.

On firepower, Jason Zucker leads Buffalo at 1.07 expected goals per 60 with 14 goals and 14 assists, but only 6 rank as high-danger—his 15.1% shooting rate runs ahead of his positioning. Josh Doan (1.01 xG/60, 15G 18A) carries a leaner 11.5% shooting rate, more volume-dependent than conversion-driven.

Brandon Hagel anchors Tampa's attack at 1.10 xG/60 with 23 goals and 26 assists but just 2 high-danger goals—his 15.8% shooting rate has done heavy lifting beyond shot quality. Anthony Cirelli is the cleaner producer: 0.95 xG/60, 15G 18A with 5 high-danger goals at 17.0% shooting. That proportion—high-danger tallies to total output—reads as legitimacy beneath Cirelli's production.

The model sees Tampa's advantage as production parity backed by superior volume: Hagel and Cirelli are outworking Zucker and Doan. But with PDO regression looming for both sides, the Lightning's 56% edge is less a prediction than a probabilistic nod to the team that's been the better underlying operation—tempered by the knowledge that luck corrections can happen anytime.