What save percentage measures, and why it beats goals-against average
Save percentage is the probability that a shot on goal does not go in. Because it divides by shots faced rather than by time, it separates the goaltender from the team in front of them in a way goals-against average cannot.
Consider two goalies who each allow three goals in sixty minutes. Both have a 3.00 goals-against average. If the first faced 20 shots and the second faced 45, the first stopped 85% of what came at them and the second stopped 93.3%. GAA says they were identical; SV% says one had a rough night behind a team that gave up little, and the other was excellent behind a team that was overrun. Every serious goaltending evaluation leads with save percentage for exactly this reason.
Save percentage still is not shot-quality neutral. A breakaway and a point shot through three bodies count the same. Modern analytics answer this with expected-goals models that weight each attempt by its historical conversion rate and then compare actual goals allowed to that expectation. Goals saved above average, which this calculator reports, is the simple version of that idea: it uses league-average save percentage as the expectation instead of a shot-by-shot model.
The formula and its three definitional traps
The calculation is SV% = (SA − GA) ÷ SA. Saves are never counted directly — they are derived, because every goal is by definition a shot on goal, so the shots you did not stop are exactly the goals you allowed.
Three definitional questions decide whether your figure matches the published one.
What counts as a shot on goal? Only attempts that would have entered the net had the goaltender not intervened, plus attempts that did enter. A shot blocked by a defenceman never reaches the goalie and is not a shot against. A shot that misses the net is not a shot against either. The broader counts — Corsi (all attempts) and Fenwick (unblocked attempts) — are team possession measures, not goaltending denominators.
Which goals are charged? Empty-net goals are not charged to a goaltender because no goaltender was in the net. In the NHL, shootout attempts are excluded from both terms; the shootout is decided separately and its shots never enter regular-season save percentage. Penalty-shot attempts during regulation do count.
What scale? Ice hockey reports SV% as a three-decimal proportion (.915). Soccer reports save percentage as a whole percentage (70%), and its denominator is shots on target rather than shots on goal, but the arithmetic is identical. Field hockey, lacrosse and handball all use the same ratio.
Goals-against average is the companion rate: GAA = GA × regulation minutes ÷ minutes played. Sixty minutes for ice hockey, ninety for soccer. It answers a different question — how many goals per game did this team concede with this goalie in net — and it is a team statistic wearing an individual's name.
Worked example: a full season of 1,500 shots
A starting goaltender faces 1,500 shots, allows 120 goals, and plays 3,600 minutes across the season. The league average save percentage is .905.
- Saves. 1,500 − 120 = 1,380.
- Save percentage. 1,380 ÷ 1,500 = 0.920000 → .920.
- Goals-against average. (120 × 60) ÷ 3,600 = 7,200 ÷ 3,600 = 2.00.
- Goals saved above average. (.920 − .905) × 1,500 = 0.015 × 1,500 = +22.5. A league-average goalie facing the same 1,500 shots would have allowed 142.5 goals; this goalie allowed 120.
- Games played, implied. 3,600 ÷ 60 = 60 full games, so about 25 shots faced per game.
Now test the target. To lift .920 to .925 you need x consecutive saves, where (1,380 + x) ÷ (1,500 + x) = 0.925. Solving: 1,380 + x = 1,387.5 + 0.925x, so 0.075x = 7.5 and x = 100 saves — roughly four more shutout-quality starts without conceding once. That asymmetry is why goaltenders talk about save percentage as something you protect rather than something you chase: late in a season the denominator is large enough that five thousandths costs a hundred consecutive stops.
How to read a save percentage
Read SV% against the league, because goaltending baselines move with rule changes and equipment. NHL league-wide save percentage sat around .915 in the early 2010s and has since drifted lower as shot quality has risen; in most seasons of the last two decades the league figure has fallen in a .900 to .915 band. Against a .905 baseline, .920 is a clear starter, .910 is serviceable, and .895 is a backup's line. Junior, college and European leagues each carry their own baseline, and youth hockey figures are usually far lower because shot quality is inconsistent.
Soccer goalkeeping runs entirely differently. Shots on target convert at a much higher rate than hockey shots, so a top-flight keeper's save percentage typically sits around 70%, and 75% is excellent. Never compare a soccer figure to a hockey one.
Sample size is the discipline. A goaltender needs on the order of a thousand shots before save percentage stops being dominated by which shots happened to come. Judging a goalie on ten starts is judging roughly 300 shots, where a swing of two goals moves the figure by seven thousandths. Goals saved above average scales with volume for exactly this reason: it tells you the size of the contribution, not just its rate.
Goals allowed at a given save percentage
| Save percentage | 500 shots | 1,000 shots | 1,500 shots | 2,000 shots | Reading (ice hockey) |
|---|---|---|---|---|---|
| .940 | 30 | 60 | 90 | 120 | Award-winning season |
| .930 | 35 | 70 | 105 | 140 | Elite starter |
| .920 | 40 | 80 | 120 | 160 | Clear number one |
| .910 | 45 | 90 | 135 | 180 | Around league average |
| .900 | 50 | 100 | 150 | 200 | Backup level |
| .890 | 55 | 110 | 165 | 220 | Below replacement |
Ten thousandths of save percentage is worth 15 goals over 1,500 shots — roughly the difference between a playoff team and a lottery team.
Mistakes that produce a wrong save percentage
- Counting blocked shots as shots against. A shot blocked by a skater never reaches the goaltender. Including blocks inflates the denominator and the save percentage with it.
- Including empty-net goals. They are not charged to a goaltender, because none was in the net.
- Including shootout shots. The NHL excludes them from regular-season save percentage entirely. Some leagues differ, so check before comparing across competitions.
- Leading with goals-against average. GAA measures the team's defence as much as the goalie. A 2.00 GAA behind a team that surrenders 20 shots a night is worse goaltending than a 3.00 GAA behind one that surrenders 40.
- Comparing hockey and soccer figures. Soccer shots on target convert far more often, so a 72% keeper and a .920 goaltender are both excellent in their own sport and not comparable to each other.
- Judging on fewer than 500 shots. Below that, save percentage mostly reflects which shots arrived rather than how they were handled.
Expected goals is the modern refinement
Save percentage assumes every shot is equally difficult, which is plainly false. Expected-goals models assign each attempt a scoring probability from its location, type, preceding play and rush context, and then measure a goaltender by goals saved above expected rather than above average. Where the data is public — the NHL, most top European soccer leagues — that is the better tool, and it can move a goalie's rating substantially when they face an unusual shot mix. Goals saved above average, computed here, is the volume-aware approximation you can build from a box score alone.
Where save percentage sits among goaltending measures
Save percentage is the rate, goals saved above average is the volume-weighted contribution, and goals-against average is the team-context figure. A complete read uses all three: a goalie with a .921 SV% over 2,100 shots contributed far more than one with a .921 over 700, and GSAA is the only one of the three that says so.
Beyond those, quality-start rate (a start with a save percentage at or above league average, or with fewer than three goals allowed on a light workload) measures consistency rather than peak. High-danger save percentage splits the shot log by location and tells you where a goalie is losing pucks. And in soccer, post-shot expected goals — which conditions on where the shot ended up on the goal frame — is the sharpest available keeper metric.
The structure is identical to rate statistics elsewhere in sport: a success count over a properly bounded opportunity count. True shooting percentage in basketball has the same shape, as do on-base percentage and earned run average in baseball. The NFL passer rating takes the idea a step further by combining four such rates into a composite. In every case the denominator, not the arithmetic, is where the errors live.
Key terms
- Shot on goal
- An attempt that either enters the net or would have without the goaltender's intervention. Blocked and missed attempts are excluded.
- Goals saved above average (GSAA)
- (SV% − league SV%) × shots against. The number of goals a goalie prevented relative to a league-average goalie facing the same volume.
- Goals-against average (GAA)
- Goals allowed per full regulation game of ice time. Heavily influenced by the team's shot suppression.
- Corsi and Fenwick
- Counts of all shot attempts and of unblocked attempts. Team possession measures, not goaltending denominators.