4-1: Final (MIN). 55%: NY Corsi. 8-3: HD chances (NY-MIN). -1.23: Osborne GSAx Final scores are summary statistics. They tell you who won and by how much, and almost nothing about how the game was actually played. Minnesota's 4-1 win over New York reads like a comfortable home performance. The underlying shot-by-shot data tells a different story. The Sirens controlled play at 5-on-5 (Corsi 55%), out-chanced the Frost in high-danger looks 8 to 3 , and produced 1.86 xG to Minnesota's 1.77 in regulation play (stripping the empty-netter). They lost by three because Minnesota finished mid-quality chances at a rate the model flatly says is not repeatable, and Kayle Osborne had her worst single-game GSAx of the season. Three regulation goals on 0.23 xG The most striking number in this game lives on the Minnesota side of the goal column. Their three regulation-time goals came on shots the model priced at 0.014, 0.07, and 0.14 xG . Combined: 0.26 xG of total goal expectation, three actual goals . Each individual shot was the kind that scores roughly one time in seven, one time in fourteen, and one time in seventy. All three went in. There is no skill explanation in the data for that sequence; it is the kind of three-shot stretch that makes a mid-quality shooter look like a sniper for a single afternoon. The other side: eight HD chances, one goal New York's eight high-danger chances are the highest single-game count for either team in this match-up. They produced one goal: Micah Zandee-Hart 's 5-on-5 strike at 7:46 of the first, on a shot the model priced at 0.016 xG. So the Sirens generated more in-tight looks than Minnesota and converted at a lower rate per chance. Both of those things can be true at the same time, and both reduce to the same coin-flip mechanic that produced Minnesota's scoring run from the other direction. Goaltending, decomposed Maddie Rooney faced 24 shots worth 1.86 xG and conceded one. That is +0.86 goals saved above expectation . A clean, above-replacement night without being an outlier. Kayle Osborne stopped 22 of 25 live shots worth 1.77 xG, finishing at -1.23 GSAx with Minnesota's late empty-netter excluded. That GSAx number is among the worst single-game lines a starter can post, and the bulk of it is the gap between the model's expectation on those three regulation-time shots and the goals they became. A goalie isn't responsible for whether a 0.014 xG shot finds the net. She is responsible only in the sense that "it went in" is what GSAx counts. The game in cumulative xG The scoring timeline P1 7:05 · Kendall Coyne Schofield · xG 0.02 · MIN 1, NY 0. P1 7:46 · Micah Zandee-Hart · xG 0.02 · MIN 1, NY 1. P3 5:30 · Abby Hustler (PP) · xG 0.17 · MIN 2, NY 1. P3 8:34 · Taylor Heise · xG 0.07 · MIN 3, NY 1. P3 18:39 · Britta Curl-Salemme (EN) · xG 0.00 · MIN 4, NY 1 Discipline New York spent 19 minutes in the box to Minnesota's 6. Three of those came on a third-period major and game misconduct to Micah Zandee-Hart , who had scored the Sirens' only goal earlier in the night. The Frost converted 1 of 3 power-play opportunities. In a 4-1 game, that single PP goal isn't the swing factor the way the regulation-time finishing run was, but it is part of the picture: a discipline gap of more than three to one in penalty minutes is the kind of self-inflicted loss vector that shows up in the analytical post-mortem regardless of what the rest of the night looked like. New York's best non-goals The Sirens' six highest-xG attempts that didn't score. Every entry here is a chance the average PWHL shooter converts at least 8% of the time. The presence of three shots above 0.14 xG reinforces the read on the run of play: New York wasn't the team that didn't generate; they were the team that didn't finish. P2 7:30 · Kristin O'Neill · xG 0.19. P3 3:53 · Anne Cherkowski · xG 0.18. P2 16:20 · Casey O'Brien · xG 0.16. P2 3:22 · Kayla Vespa · xG 0.13. P3 3:56 · Casey O'Brien · xG 0.13. P1 6:08 · Clair DeGeorge · xG 0.13 What this read isn't Not a claim Minnesota didn't deserve to win. They scored four goals; New York scored one. Score is the only line on the schedule. The point of the analytical read is that the underlying process was much closer than the score, which has implications for the next time these teams meet, not for what happened tonight.. Not a claim Kayle Osborne played poorly. A -1.4 GSAx night is a bad night by the model's accounting, but the bulk of that is three goals on shots the model graded under 0.15. Single-game GSAx has wide variance on small samples, especially when a few low-percentage shots find space.. Not a claim the model is wrong. xG is a probability over a single shot. Three improbable shots in a row are improbable, not impossible, and they happen many times across a season. The right test is the season-long log-loss against actuals, not the single game where the rare sequence shows up.. Not a claim discipline was the cause. One PP goal on three opportunities does not account for a three-goal margin. The discipline note is a process flag for next time, not a re-explanation of tonight's loss. 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/103 .