Vancouver controlled this game—56% of expected goals (4.26–3.34), 56% of shot attempts (31–24)—but Montreal won 4–3. A heist.

J. Dobes was the difference. The Montreal goaltender stopped 28 of 31 shots with a 0.903 save percentage, 1.26 goals above expected. K. Lankinen's 0.833 SV% came in 0.66 goals short of expected. That two-goal swing in goaltending was the game.

Both teams scored in transition. Vancouver made 3 goals off 34 rush chances; Montreal made 4 off 28 rush chances—Montreal more efficient, Vancouver less so. On the power play, Vancouver buried 2 of 11 shots; Montreal buried 2 of 9. Both dangerous, neither dominant.

High-danger chances finished 7–7. Neither team converted a rebound opportunity. The game turned on goaltending and efficiency in lower-leverage shots.

The turning point came late in the third: Conor Garland's tip-in from the high slot at 16:13 put Montreal ahead and held.

For Vancouver, this was the frustrating loss. The Canucks owned the underlying metrics and exited empty. E. Pettersson registered 1G, 2A in 21:47. Q. Hughes added 2A in 26:25. I. Demidov contributed 1G, 2A in 13:47 for Montreal.

Montreal's 51% win probability isn't a prediction—it's a shrug. The model projects 2.6 to 2.5 goals in Montreal's favor; the difference amounts to randomness.

Montreal owns the underlying play. At 5v5, the Canadiens generate 48.0% of expected goals to Vancouver's 44.0%, and 49.0% of shot attempts to the Canucks' 46.0%. That underlying edge matters. It doesn't guarantee a win.

What might cancel it out: regression. Montreal's PDO of 1.018 means they're converting at an unsustainable clip—variance will pull them downward. Vancouver's PDO of 0.974 is the inverse: the Canucks are running cold and due for positive regression. In a single game, neither reversal is guaranteed, but both are directionally real.

Both teams face goaltending uncertainty. T. Demko is on a back-to-back for Vancouver; whether he starts or a backup takes the net affects everything. S. Montembeault faces the same situation for Montreal. A tired starter plays differently from a fresh one—if either goalie is gassed, it erases preparation advantages.

Three Stars

Cole Caufield (Montreal) — 33 goals, 0.97 expected goals per 60 minutes, and a 20.0% shooting rate. He's clicking at the rate that wins games individually. The Canucks' defensive design around neutralizing him will likely determine whether Montreal's xG advantage translates.

Filip Chytil (Vancouver) — 1.03 expected goals per 60 minutes on just 3 goals. Chytil is high-variance—he's generating chances at an elite rate but hasn't converted. Vancouver needs variance to reverse in his direction; one game could do it.

Drew O'Connor (Vancouver) — 13 goals, 0.85 xG/60, shooting at 11.8%. O'Connor is a secondary threat who punches above his expected output, useful ballast if Chytil's regression cuts one way or another.

Oliver Kapanen (Montreal) carries 0.83 xG/60 with 19 goals and a 17.8% shooting rate—solid secondary scoring that adds depth but won't move the needle alone.

The Pick

Take Montreal at pick'em or better. The underlying play is real enough to trust, even if the model won't commit beyond 51%. Vancouver's PDO suggests positive regression is coming, but "coming eventually" and "tonight" are different propositions. Goaltending noise cuts both ways; that's the one true coin flip. When you have edge and uncertainty canceling out, edge wins.