3-2: Final (NY). 29-26: Shots (NY-TOR). 2.14-1.83: xG (NY-TOR). -0.86: Kirk GSAx Toronto owned 60.7% of Corsi in this game, the kind of possession share that usually signals territorial control. But the xG ledger told a different story: New York generated 2.14 xG to Toronto's 1.83 . The gap between shot volume and shot quality is the analytical signature of this game. Toronto out-attempted New York by a wide margin but produced lower-value looks per attempt, while New York concentrated fewer chances into higher-quality opportunities. When all five goals arrived in the second half of the game (one in the second period, four in the third), the quality edge caught up. The scoreless first period The first twenty minutes produced no goals and a relatively even shot split (New York 8, Toronto 9). Both goaltenders were comfortable. The Corsi imbalance was already building in Toronto's favour, but neither team converted a look worth writing about. For the model, the first period was noise: low event volume, low combined xG, nothing to separate the two sides. Three goals in 6:33 Toronto entered the third period leading 1-0 after Daryl Watts scored a power-play goal at 14:43 of the second (0.06 xG, assisted by Renata Fast ). Ella Shelton extended it to 2-0 at 8:51 of the third (0.04 xG, assisted by Claire Dalton ). Both Toronto goals came on low-probability shots, a combined 0.10 xG. The model would have expected those two shots to produce zero goals roughly nine times out of ten. Then New York scored three times in 6 minutes and 33 seconds. Denisa Krizova cut the deficit at 9:30 of the third (0.10 xG, 5-on-5), assisted by Casey O'Brien setup. Maja Nylen Persson tied the game at 14:53 on the power play (0.17 xG), with assists from Casey O'Brien and Sarah Fillier . Casey O'Brien then scored the winner at 16:03 (0.19 xG, 5-on-5), assisted by Nicole Vallario. New York's three goals totalled 0.46 xG. Like Toronto's pair, these were not high-probability chances. The difference was that three of them beat the goaltender in quick succession. Goaltending, decomposed Raygan Kirk faced 29 shots worth 2.14 xG and conceded three goals. That is -0.86 goals saved above expectation . A GSAx of -0.72 in a one-goal loss means the goaltending line carried most of the deficit. The model expected roughly 2.3 goals against on that workload; Kirk allowed three. In a game that ended 3-2, that gap is the approximate margin. Kayle Osborne faced 26 shots worth 1.83 xG and allowed two. Her GSAx for the night was -0.17 , essentially even with expectation. She was not a factor in either direction. The comeback was driven by finishing, not by a goaltending collapse at one end or a brick wall at the other. The game in cumulative xG The scoring timeline P2 14:43 · Daryl Watts (SH) · xG 0.05 · NY 0, TOR 1. P3 8:51 · Ella Shelton · xG 0.04 · NY 0, TOR 2. P3 9:30 · Denisa Křížová · xG 0.10 · NY 1, TOR 2. P3 14:53 · Maja Nylén Persson (PP) · xG 0.16 · NY 2, TOR 2. P3 16:03 · Casey O'Brien · xG 0.19 · NY 3, TOR 2 New York's power play The Sirens had 5 power-play opportunities and generated 0.53 xG across all of them. They converted once: Maja Nylen Persson's goal at 14:53 of the third, the tying marker at 0.17 xG. That single conversion on the man advantage was the pivot point. Without it, New York needs to score three at even strength in under seven minutes, a much harder ask. Toronto's lone power play produced 0.00 xG. Toronto's best non-goals The chances the model graded as Toronto's best looks at Osborne. Every entry here is a shot the average PWHL shooter converts at least 7% of the time. P2 4:21 · Clara Van Wieren · xG 0.21. P2 3:24 · Natalie Spooner · xG 0.14. P1 15:32 · Emma Woods · xG 0.12. P2 0:49 · Claire Dalton · xG 0.12. P3 4:19 · Renata Fast · xG 0.11. P3 1:00 · Renata Fast · xG 0.10 What this read isn't Not a claim Toronto deserved to win on process. They dominated Corsi but trailed in xG. Possession volume and expected-goal quality pointed in opposite directions, and the quality side won.. Not a goaltending-only explanation. Raygan Kirk 's -0.72 GSAx was the largest single-player swing, but Toronto also scored twice on a combined 0.10 xG. Both teams were finishing above expectation; New York simply had one more finish.. Not a penalty narrative. Toronto drew five power plays and scored on one of them. New York drew one and did not. The special-teams imbalance was real, but the xG gap on the power play (0.59 to 0.00) tells the sharper story.. Not a model failure. Five goals on roughly 4.2 combined xG is within normal variance. The comeback was dramatic for viewers, but the xG process was close to even all night. 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), empty net, penalty shot, power play, six shot-type dummies.. 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/108 .