The Ducks were not as bad as 92 points
They controlled play at five-on-five all year. Then the goaltending and the finishing both went the wrong way at once.
Ninety-two points and no playoffs reads like a bad team. The underlying numbers say something narrower and more fixable: a club that won the run of play and lost almost everything that sits on top of it.
They had the puck
Anaheim finished at 52.2% of shot attempts and 50.4% of expected goals at five-on-five — not a powerhouse, but the right side of even across eighty-two games. Per sixty minutes they generated 3.41 expected goals and allowed 3.35. That is a team playing roughly level hockey and, on the balance of chances, a little better than its opponents.
Teams that do this finish above the line more often than not.
Two things went wrong at once
The goaltending. Anaheim's netminders finished a combined −16.1 goals saved above expected. Lukas Dostál went −4.26 across 56 games at an 88.8% save percentage; Ville Husso −3.79 in 20 at 88.4%. On high-danger shots Husso stopped 76.3%. A team playing even hockey cannot absorb that.
The finishing. The other end failed by a similar margin — the group finished 20 goals below its own expected total. Shooting and saving both running cold in the same season is what a 0.981 PDO describes, and PDO is the least sticky number in hockey.
Why that matters for this year
Three of the model's regression flags fired on this roster, and all three point the same way: the PDO, the finishing gap, and a save percentage below replacement level. None of them is a statement about how good the Ducks are. They are statements about how unlikely it is that the same two things go wrong twice.
The penalty kill is the exception, and the real problem. 76.4% is not variance — it is a structural hole, and it allowed 8.66 goals against per sixty shorthanded. That one does not fix itself.
The part that is genuinely good
Leo Carlsson is already a driver rather than a prospect: on the model's component profile he grades in the 95th percentile for chance creation and the 88th for finishing, on 14 minutes a night. Cutter Gauthier is the other name the model keeps returning. Top-end talent is the hardest thing on this list to acquire, and Anaheim has it.
Five-on-five rates and expected goals are computed from NHL play-by-play with Backcheck's own xG model, calibrated so league expected goals match league goals. Impact components are model-estimated percentiles, not official stats. Figures are 2025-26 final.