By Conference Finals, goaltending analysis has been thoroughly contaminated by narrative. A starter who won two series on hot streaks becomes “the reason his team is here.” The question the analytics answer is: is that narrative tracking a real signal, or are we labeling variance as performance?
GSAX — Goals Saved Above Expected — is the metric designed to answer that question.
GSAX counts how many goals a goaltender prevented beyond what a league-average goaltender would have allowed facing the same shot slate. It uses an expected goals model (Natural Stat Trick’s methodology) to weight each shot by location, shot type, traffic, and rebound situation, then compares actual goals allowed to that weighted expectation.
A goaltender at +5.0 GSAX through two playoff rounds has saved 5 goals that a league-average replacement would have allowed. That is a genuine performance contribution — about half a win’s worth of goals in a playoff context where a single goal swing wins ~20% of close games.
At -1.0 GSAX with a high SV%, the goaltender has benefited from facing more below-average shots than expected. He is not underperforming; the shot mix has been kind to him. The SV% looks good. The underlying production is neutral.
Profile 1: High SV%, Positive GSAX — Genuine outlier. The goaltender is stopping shots at above-expected rates AND the shots he faces are not unusually favorable. This is the hardest profile to maintain, but it has the strongest persistence into the next series. Back-to-back Conference Finals runs with +3.0 GSAX per round exist — Vasilevskiy in 2020–2021 is the benchmark.
Profile 2: High SV%, Flat or Negative GSAX — Variance beneficiary. The goaltender’s high SV% is primarily explained by shot quality — he is facing disproportionately few high-danger shots relative to the total. Conference Finals shot mix, which skews more slot-heavy than earlier rounds, removes the conditions that produced the high SV%. Regression is structural, not random.
Profile 3: Average SV%, Positive GSAX — The underrated profile. This goaltender is stopping shots at above-expected rates despite facing a difficult shot mix. His SV% is suppressed by opponents generating quality chances, but he is actually outperforming his expected save rate. In Conference Finals, where the shot mix trends harder, his GSAX should improve rather than regress.
Profile 4: Average SV%, Flat GSAX — Neutral performance. The goaltender is providing league-average value relative to his shot slate. Team possession structure will decide series outcomes for him — he is not a multiplier in either direction.
Playoff rounds 1 and 2 include at least some matchups between teams with significant skill differentials. A goaltender posting +4.0 GSAX in Round 1 against a weak possession team may be stopping genuinely low-danger shots more often than the model expects — the model’s expected goals calibration is based on league-wide data, not specifically weak-team shot quality.
By Conference Finals, both opponents in each series are among the four best teams in their conference. The shot slates are calibrated closer to the model’s training data. GSAX becomes more reliable as a signal precisely because the shot quality it is measured against is more representative.
Additionally, Conference Finals GSAX accumulates over 12+ games by the time the series begins. At that sample, the measurement error is substantially lower than after 4–5 games.
The threshold that separates meaningful goaltender surplus from baseline noise at the Conference Finals stage: +4.0 GSAX or better through two completed rounds.
Above +4.0: the goaltender has materially outperformed expectation. His performance contributed real wins to the team’s playoff run. He is a genuine positive in Conference Finals.
Between 0 and +4.0: competent, within normal range. Not a multiplier factor in either direction.
Below 0: the team has gotten to Conference Finals despite neutral or negative goaltending relative to shot quality. Their possession structure and forward depth are carrying the load. This is not unsustainable — Tampa Bay in 2021 had a below-average GSAX goaltender in the Conference Finals and won the Cup. But the margin for error in 5v5 play is narrower.
2026 Conference Finals goaltender GSAX — through Rounds 1 and 2 (source: Natural Stat Trick):
| Team | Goaltender | SV% (R1–R2) | GSAX | Profile |
|---|---|---|---|---|
| Carolina Hurricanes | Frederik Andersen | .950 | +10.87 | Profile 1 |
| Montreal Canadiens | Jakub Dobeš | .910 | +9.39 | Profile 3 |
| Colorado Avalanche | Scott Wedgewood | .914 | +4.60 | Profile 1 |
| Vegas Golden Knights | Carter Hart | .917 | +1.44 | Profile 2 |
The profiles in this table determine whether Conference Finals goaltending is a matchup factor or a non-factor. When two teams in the same series have GSAX profiles that mirror each other (both Profile 1, both Profile 4), goaltending cancels out and possession decides. When one team has a Profile 1 goaltender against a Profile 2, that mismatch is the single most important series variable.
Andersen’s +10.87 GSAX through two sweeps is the most dominant two-round goaltending performance remaining in the 2026 bracket — and one of the stronger two-round marks in recent playoff history. Dobeš, at +9.39 through 14 grueling games across two seven-game series, is carrying a shot mix that suppressed his SV% while his underlying stop rate has been exceptional. The ECF matchup between these two is the most compelling goaltending context of the conference round: Andersen is a genuine surplus goaltender by every measure; Dobeš is the underrated profile, facing harder shots and outperforming that difficulty.
In the Western Conference Final, Wedgewood enters at the bottom of the meaningful-surplus threshold (+4.0) after a Round 2 that dragged his cumulative GSAX down from a dominant Round 1 (+6.19 in Round 1 alone). Hart, at +1.44, is a variance beneficiary on the cumulative ledger — his .917 SV% has been shaped by a difficult Round 1 (-4.82 GSAX) and a brilliant Round 2 (+6.26 GSAX). The trajectory matters: Hart is trending sharply up, Wedgewood is trending down.
Data source: Natural Stat Trick (GSAA/GSAX, playoff goaltender page). Subscribe to The Hockey Analytics newsletter for game-by-game analytics throughout the 2026 Conference Finals and Stanley Cup Finals.