Montreal controlled the ice—55% of shot attempts—yet lost 3-2 to Boston in a game where expected goals split 4.80–4.82. That split said dead-even. The scoreboard begged to differ, and the answer sat between the pipes.

The Canadiens' evening vanished at the highest-leverage moments. They went 0/8 on high-danger chances, the kind of drought that hardens quickly into regret. Montreal's 55% shot advantage never translated downstairs; they scored twice off the rush on 29 chances but found nothing in the core. They underperformed their expected goals by 2.80, pouring possession into a game that wouldn't repay it.

Boston was equally silent on grade-A looks, going 0/9 on high-danger chances. Yet they broke through. Twice on the rush (26 chances), three times total—and that margin is J. Swayman. He stopped at 0.929 save percentage, 26 saves on 28 shots, in a performance that ran 2.80 goals better than expected. When both teams miss this badly—17 combined high-danger chances, zero conversions for Montreal, zero for Boston—the goaltender who turns that into a win becomes the differentiator.

S. Montembeault made 20 saves on 23 shots (0.870 SV%), yet it wasn't enough. Montreal's underperformance of their xG by 2.80 cost them. Boston, despite underperforming their 4.82 xG by 1.82, still found three. The asymmetry was Swayman's work.

Neither team crashed the crease with conviction: both created 5 rebound opportunities, both converted zero. Neither dominated the rush—Montreal 2/29, Boston 2/26. The game was an evenly distributed mess, which should have meant an evenly distributed result. Instead, one goaltender's save percentage wrote 3-2 across the scoreboard.

D. Pastrnak's 2 assists in 20:15 framed the attack. N. Zadorov added 1 in 21:05. For Montreal, N. Suzuki had 1 assist in 24:35.

The model can't decide. It projects a Boston win by 2.5 to 2.7 goals but gives Montreal 53% win probability—the hockey equivalent of flipping a coin. There's no edge here.

The Bruins are hotter by record, sitting 8-2-0 in their last ten. The Canadiens are underwater: -6 in goal differential over that same span. But dig deeper and the contradiction dissolves. Both teams' underlying metrics are nearly identical. Boston holds 47.0% expected goals share at 5v5; Montreal sits at 48.0%. Their Corsi percentages are identical at 49.0% apiece. This is not a gap between a dominant and a struggling team. It's a gap between two evenly matched ones, with Boston getting hot at exactly the right time.

The complication: both teams are running lucky, and luck expires. Boston's PDO sits at 1.023—high enough to regress hard. Montreal's is 1.018. Both of these numbers tell you the same story: unsustainable results propping up records that don't quite match the underlying play. Add in the goaltending situation—J. Dobes and J. Korpisalo are both on the back-to-back—and you have two tired opponents leaning on performance that won't hold.

Where this game tips depends on the edge players. Cole Caufield leads Montreal's scoring with 33 goals, but his expected production (0.97 xG/60) doesn't quite match that output. His 20.0% shooting rate explains the gap; regression is waiting. Oliver Kapanen's 0.83 xG/60 is cleaner, and his 4 hard-chance goals suggest a forward who converts opportunities rather than just generating them.

Boston counters with Viktor Arvidsson, who generates at 1.04 xG/60—the most dangerous rate on the ice. He's cashed 18 goals with 13.0% shooting, which is sustainable talent, not luck. Tanner Jeannot runs just 0.75 xG/60 but has found 6 goals anyway, leaning on a 2-hard-chance advantage to overcome a modest 8.3% shooting rate.

The defensive anchors matter too. Elias Lindholm is Boston's weak link; Mike Matheson is Montreal's. In a game this tight, how well each team's defense can shelter their goaltenders and limit the other's best shooters will decide it.

This is as close to a true toss-up as you'll find. Boston has momentum. Montreal has more even 5v5 play. Both will regress. The team that avoids the biggest collapse wins.

Data: MoneyPuck 5v5 models, NHL API