5-2: Final (VAN). 2.22-1.73: xG ex-EN (VAN-NY). 10-9: HD (NY-VAN). -1.78: Osborne GSAx This was a high-scoring game where the chance-quality split ran one way on the model and the other way on the feed-provided high-danger count. New York carried more shot-attempt share at 5-on-5 (Corsi 57% ) and produced 10 high-danger chances to Vancouver's 9. The model graded Vancouver's in-play offence higher anyway: 2.22 xG to 1.73 once the late empty-netter is stripped. The first-period scoring flurry put New York behind early, and Vancouver finished four of their 4 regulation goals on shots the model graded under 0.12 xG. Three goals in 2:20 Anna Shokhina opened the scoring at P1 4:55 on a 0.11 xG 5-on-4 wrister. Anna Segedi added a 5-on-5 wrister 1:53 of game-clock later at P1 6:48 (0.10 xG). Sarah Fillier answered for New York 27 seconds after that at P1 7:15 with a 0.10 xG shot. Three goals in 2:20 of game-clock, all on shots the model graded between 0.10 and 0.11 xG. Three mid-quality looks all finding the net inside a two-and-a-half-minute window is the kind of compressed sequence that decides games before either team has time to settle into a structured pattern. Vancouver's finishing run The Goldeneyes scored four regulation goals on shots combining for 1.98 xG (the four scoring attempts: 0.11, 0.10, 0.02, 0.05). That works out to 1.80 times expected on a per-shot basis. The 0.02 xG slap shot from Sophie Jaques at P1 16:00 and the 0.05 xG wrister from Tereza Vanišová at P3 5:27 were the lowest-quality scoring shots of the night. Across a season that finishing rate sits much closer to 1.0; in a single game on four shots it can land anywhere. Goaltending, decomposed Kayle Osborne faced 28 live shots worth 2.22 xG and conceded four. Her GSAx came out at -1.78 . Three of the four goals against her were under 0.12 xG, one under 0.03. The save process underneath the GSAx number was much closer to expected than the headline suggests; the empty-net goal is excluded from her line. Kristen Campbell stopped 22 of 24 live shots worth 1.73 xG, finishing at -0.27 GSAx . Essentially the model calling her workload almost exactly. Both Sirens goals were on shots in the 0.10 to 0.12 xG range; nothing about her save process was outlier-shaped. The game in cumulative xG The scoring timeline P1 4:55 · Anna Shokhina (PP) · xG 0.10 · VAN 1, NY 0. P1 6:48 · Anna Segedi · xG 0.09 · VAN 2, NY 0. P1 7:15 · Sarah Fillier · xG 0.12 · VAN 2, NY 1. P1 16:00 · Sophie Jaques · xG 0.02 · VAN 3, NY 1. P2 13:19 · Anna Bargman · xG 0.12 · VAN 3, NY 2. P3 5:27 · Tereza Vanišová · xG 0.05 · VAN 4, NY 2. P3 19:42 · Mannon McMahon (EN) · xG 0.00 · VAN 5, NY 2 New York's looks above 0.07 xG The Sirens' chances the model graded above the 7% scoring-probability threshold. The Fillier and Bargman scoring attempts sit in the middle of the list; the rest are the chances Campbell turned away. Plenty of in-tight looks, none of the breakaway-grade variety. P2 13:31 · Kristin O'Neill · xG 0.23. P1 9:35 · Kayla Vespa · xG 0.14. P3 8:08 · Paetyn Levis · xG 0.14. P2 13:19 · Anna Bargman (goal) · xG 0.12. P1 7:05 · Anna Bargman · xG 0.12. P1 7:15 · Sarah Fillier (goal) · xG 0.12 What this read isn't Not a claim New York deserved to win. The Corsi and high-danger numbers favoured the Sirens, but Vancouver's chance-quality on the model was higher and the in-play xG split tilted home. The result reflected the underlying read on chance quality even if the territorial and HD-count metrics disagreed.. Not a claim Kayle Osborne played a structurally bad game. Four regulation goals on shots all under 0.12 xG is mostly variance. The save process underneath the GSAx number was clean.. Not a claim Vancouver's finishing rate is repeatable. Four regulation goals on a team xG below two runs above the game-level expected finish by roughly a factor of two. Single-game finishing variance is enormous; the chance-creation profile is the keeper here, not the conversion rate.. Not a model failure. A near-even underlying chance-quality game ending 5-2 with a finishing run on one side and a small Corsi lean on the other is well within the variance the model's probability outputs allow on small samples. Methodology xG model : PWHL-calibrated logistic regression on 16,709 non-EN shots. Same fit as the methodology post . Features: distance, angle, rebound (≤3 s), penalty shot, power play, overtime, scorer-tagged quality chance, seven shot-type dummies. Empty-net shots are priced outside the main model (MoneyPuck / NST convention).. GSAx : summed xG on shots faced minus goals allowed, computed per-goalie from the PBP feed. Empty-net shots and goals are excluded from both sides of the ledger.. Numbers regenerate on every build. A PBP re-sync or xG recalibration reshapes every figure in this post on the next static export. Full play-by-play, shift overlays, and the per-player game log for this game are at /games/81 .