GSAX
also: goals saved above expected · GSAx · GSAA
Expected goals against minus goals actually allowed: how many goals a goalie saved that an average goalie would not have.
GSAX = expected goals against − goals againstGSAX leaders, 2026-27
Goalies with 8+ games, updated daily. The same numbers get_goalie_stats returns.
| # | Goalie | Team | GP | GSAX | SV% |
|---|---|---|---|---|---|
| 1 | Jesper Wallstedt | MIN | 16 | +13.8 | .922 |
| 2 | Jacob Fowler | MTL | 10 | +11.6 | .925 |
| 3 | Jeremy Swayman | BOS | 13 | +10.6 | .920 |
| 4 | Linus Ullmark | OTT | 8 | +9.4 | .945 |
| 5 | Pyotr Kochetkov | CAR | 12 | +7.5 | .920 |
| 6 | Daniil Tarasov | DET | 8 | +6.3 | .919 |
| 7 | Ilya Sorokin | NYI | 12 | +6.0 | .914 |
| 8 | Jake Oettinger | DAL | 13 | +5.7 | .906 |
| 9 | Yaroslav Askarov | SJS | 13 | +5.6 | .889 |
| 10 | Joel Hofer | STL | 12 | +5.5 | .919 |
Save percentage treats every shot the same, so a goalie behind a team that gives up slot chances looks worse than one facing point shots, whatever either of them does. GSAX fixes that by scoring each shot for quality first: face harder shots and more is expected of you.
Positive is above average. Around +20 over a season is an excellent year and the best reach the high twenties or thirties. It is calibrated within each season, so league GSAX sums to zero every year and a goalie is measured against his own peers rather than a different era.
As a check on whether it is measuring anything: the season leader here is a Vezina winner or the consensus analytics pick in every season tested, including Price in 2014-15, Gibson in 2017-18, Hellebuyck in 2019-20 and 2023-24, Shesterkin in 2021-22.
Reading small gaps as real. The model slightly over-rates the most dangerous chances, so a goalie facing an unusually high-danger workload is flattered. A two-goal difference between two goalies over a season is noise.
Where to get it
Related
GSAX from the API
500 free credits every month, no card. REST or MCP.
Get a free key