2-0: Final (OTT). 18-42: Shots (OTT-TOR). 69%: Toronto Corsi. +2.88: GSAx, Philips If the Seattle game was a lesson in how a team can win the possession battle and still lose the scoreboard, this was the reverse. At Coca-Cola Coliseum on Friday night, the Toronto Sceptres outshot the Ottawa Charge 42 to 18 , owned 69% of all shot attempts , and lost 2-0. The Charge took seven faceoffs in their own end every period, killed five power plays, and let Toronto hit the slot 19 times in the first period alone. Then they went home with a regulation win and all three points because Gwyneth Philips stopped every single shot Toronto took. What Philips actually saved On a neutral night, a goalie who faces 42 shots worth 2.88 xG gives up about three goals. Gwyneth Philips gave up zero. The difference between the model's expectation and what actually happened is +2.88 goals saved above expectation in a single game. Elite goaltending across a full PWHL season runs somewhere around +0.5 GSAx per game; a performance in the +3 range is the kind that ends up on a season-summary list. Framed differently: Gwyneth Philips faced nine shots the model graded as high-danger chances and did not concede on any of them. Toronto's shot distribution across the first two periods was the shape of a team that was pinning its opponent in the defensive zone, and the shot locations matched the territorial pressure. What stopped being true around 11:20 of the third is that any of those chances were finding space. The game in cumulative xG The scoring timeline P3 11:20 · Brianne Jenner · xG 0.13 · OTT 1, TOR 0. P3 19:37 · Sarah Wozniewicz (EN) · xG 0.00 · OTT 2, TOR 0 Toronto's six best non-goals The highest-xG shots in the game that did not find the net. Every entry here is a chance the average PWHL shooter converts between 12% and 22% of the time. On a normal night, at least one goes in and we are writing a different post. P2 1:00 · Blayre Turnbull · xG 0.21. P2 0:32 · Jesse Compher · xG 0.19. P1 17:32 · Clara Van Wieren · xG 0.17. P2 0:30 · Jesse Compher · xG 0.16. P2 5:26 · Daryl Watts · xG 0.13. P2 4:31 · Emma Woods · xG 0.13 The other side of the ledger Ottawa scored two goals on 1.22 xG. Strip Sarah Wozniewicz 's late empty-netter (0.00 xG on a shot from their own zone into an unattended net) and the Charge produced 1.22 xG of actual in-play offence. That is a modest single-game output against a good defensive team, not a dominant one. Brianne Jenner 's 11:20 goal was a 0.11-xG shot: not a broken-coverage high-danger look, but a mid-range attempt that found a hole. Under the reverse read, it is the symmetric version of Danielle Serdachny 's 0.04-xG goal for Seattle three nights before. On any given shift, finishing is the coin flip that makes the final look inevitable in retrospect. Reading these two games together Ottawa's two games this week produced the same Charge roster and two opposite shapes in the possession-versus-result plane: Apr 8 vs. Seattle (home). Ottawa 57% Corsi, 3.24 xG for, 3 goals for, lost 3-5 .. Apr 11 at Toronto. Ottawa 31% Corsi, 1.22 xG for, 2 goals for, won 2-0 . Neither game was representative of how Ottawa is likely to play across the remaining schedule, and neither finishing ratio will repeat. Together they describe the honest width of the distribution around any single-game read. A team can produce the same underlying process for two consecutive games and end up on opposite sides of the standings because the scoreboard is compressed into one or two finite outcomes while the model is summing hundreds of probabilities. That is the argument for looking at underlying chance quality over many games rather than any single result, and these two games landing within three days of each other is as clean an example of the argument as the season is likely to produce. What this read isn't Not a claim Ottawa was lucky. Goaltending is a repeatable skill. Gwyneth Philips 's performance is an outlier in magnitude, not in kind. She was the first goaltender selected in the 2024 PWHL draft and was the playoff MVP when Ottawa reached last year's Final. A full-season GSAx profile is what tells us whether games like this are extreme tail or an emerging norm.. Not a claim Toronto played poorly. 42 shots, 3.02 xG, nine high-danger chances is the process you want. The model priced their output fairly; the goalie priced it at zero.. Not a single-game indictment of the model. xG predicts process, not outcomes. A 3.02-xG team shutting out happens maybe 3-4% of games across a large corpus. Tails are not bugs.. Not an argument against possession metrics. Corsi and xG both pointed the same direction in this game. The argument is always about reading one game through a process lens and many games through an outcome lens, not about which number is 'right'. 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, and seven shot-type dummies. Empty-net shots are priced outside the main model (MoneyPuck / NST convention).. GSAx : goals-saved-above-expected, defined as summed xG on shots faced minus goals allowed. Computed per-goalie from the same PBP feed. Shots that resulted in empty-net goals are not counted against the defending goalie.. Corsi / Fenwick : Corsi counts all shot attempts (including blocks); Fenwick strips blocked attempts. Both are symmetric around 50% by construction.. Every figure regenerates on each build. A PBP re-sync or xG recalibration reshapes all numbers in this post automatically on the next static export. Full play-by-play, shift overlays, and the per-player game log for this game are at /games/104 .