Colorado stole this one. The Penguins controlled 63% of expected goals at 5v5, held 52% of shot attempts, and won the high-danger chances 7-4. They still lost 6-2.

This is what happens when goaltending variance breaks badly. S. Wedgewood was 2.16 goals above expected; A. Silovs was 2.51 goals below. The Penguins had the chances. The Avalanche had the execution.

Colorado generated 3.51 goals above their xG — elite finishing or bounces breaking right. Pittsburgh underperformed their xG by 2.16, wasting high-danger looks that should have buried the Avalanche. The Penguins created openings. The Avalanche converted them.

One shift tipped the balance: Ross Colton's wrist shot from the wing at 16:20 of the third period, on a night when Colorado was lethal in transition. The Avalanche scored 3 goals on 15 rush chances — a conversion rate that rendered Pittsburgh's territorial advantage pointless.

Colorado also dominated the net-front, generating 3 rebound opportunities and converting 1. Two Avalanche goals came from low-danger areas, the kind a goalie should stop. S. Wedgewood faced 32 shots and stopped 30; A. Silovs faced 29 and stopped 24. A 0.938 save percentage beats a 0.828 save percentage.

For Pittsburgh, the math is cruel: a 63-37 expected-goals advantage means nothing when the other team finishes and your goalie doesn't. The Penguins had possession and chances. They lacked execution.

The model projects a 2.8–2.6 Avalanche win with Colorado at 53% win probability. This is about as close to a pick'em as you'll find.

Colorado owns the underlying play. The Avalanche shoot 57.0% xG% and control 57.0% of Corsi at 5v5—that territorial advantage should feel comfortable in a low-score game. Pittsburgh generates just 51.0% xG% and 50.0% Corsi%, a significant disadvantage when the puck spends two-thirds of the game in your defensive zone.

The Penguins come in wounded. They're -7 on goal differential over their last 10 games, the kind of sink that doesn't recover in a single night. For Pittsburgh to steal this, they need to tighten defensively and avoid giving Colorado the possession advantage they've built.

Here's the wrinkle: Colorado's PDO sits at 1.021—they're running hot—which means some regression is baked into any projection favoring them. Pittsburgh's 1.012 is closer to sustainable, a minor edge in a coin flip that cuts both ways.

Artturi Lehkonen drives Colorado's offense. He's at 1.11 xG/60 with 18 goals and 22 assists; six of those goals have come from high-danger chances, and he's shooting 14.9%. He's the player Pittsburgh must neutralize. Ross Colton supports him at 1.10 xG/60, but his shooting line tells you where Pittsburgh's defense has the best chance to limit damage—Colton has just 7 goals on 15 assists with a 4.9% shooting rate, well below league average.

Pittsburgh's offense hinges on Rickard Rakell, who carries 1.01 xG/60 and has converted 13 goals and 13 assists. Three of his tallies are high-danger; he's shooting 12.1%, a rate that suggests he's overperforming his chances. Filip Hallander, who leads the team at 1.28 xG/60, is likely out on IR and unavailable, removing Pittsburgh's most dangerous creator from the matchup.

Nazem Kadri is Colorado's defensive liability; Caleb Jones is Pittsburgh's. Both teams sit at 0% high-danger and rebound goal rates—an anomaly neither can sustain. This is a possession game with Colorado favored by underlying numbers but punished by heat-adjusted regression. Pittsburgh's only path runs through tighter defensive structure and Rakell's conversion. The 53% call is right: either team can win, and the model knows it.