#FancyStats: Goals Saved Above Average

Evaluating goal tenders is tough. GAA (Goals Against Average) is wonky, and is more of a team stat than a single person, as is SV% (Save Percentage). One method for goaltender evaluation that is becoming more popular and is easier to use is the Goals Saved Above Average, or GSAA.

There are no #fancystats that are really "old" in hockey, as even the ones that are almost commonplace nowadays (Corsi, Fenwick, etc.) are only a few years old, but GSAA is one of the newer ones. The stat comes from the idea of measuring goalies against something, rather than just a stat alone. The way it works is: you apply the league-wide SV% to the number of shots faced by a goalie, and compare that to the number of goals allowed by that goalie.

While at first this appears to simply be comparing a goalie's SV% to the league standard, it is not. Take, for instance, this table:

via 668d3eaa831be4d52f0d515f.ingoalmagazine.netdna-cdn.com

While Cam Talbot seems to have a better SV% than Ben Bishop (.935 compared to .933), or even better than Tuukka Rask (.935 compared to .928), by looking instead at GSAA, we see that Talbot saved only 9.6 goals above the average, while Rask and Bishop saved 18 and 23, respectively. By that measure, Bishop and Rask are the better goalies. It's worth noting that Talbot's GAA is better than both Rask and Bishops' also.

GSAA is great at taking out the quality of a goalie's team. A goalie with a terrible team defensively (Jonathan Berneier... Semyon Varlamov...) will still have a good GSAA, even if their GAA or SV% are a little worse. GSAA levels the field... err... ice, and makes it a little clearer what a goalie's quality was or was not.

Another way of saying this might be: GSAA gives you a good indication as to whether a team relies on great goaltending to win games or not. Notice, for instance, who isn't in that top-10 list? Lundqvist, Quick, and Crawford; arguably the 3 goalies most likely to win the Stanley Cup.

The Big But

GSAA is not perfect, however. Like all statistics, it is subject to sample sizes: the bigger the better. However, in this instance, more games played=more goals saved, which will tend to drive this number up. We could eliminate this by turning it into "GSAA/60," but that causes problems as well.

In addition, GSAA does not differentiate between even-strength goals, PK goals, etc. This means that a goalie for a team that is constantly killing penalties will tend to have a lower GSAA than a goalie whose team is not, and that does not necessarily mean that one goalie is better than the other.

Another effect that exists for goalies is fatigue: a goalie that faces 46 shots (John Curry v. the St. Louis Blues) will likely not make as many saves as a goalie that only faces 12 (Bryz v. the Colorado Avalanche). There's no good way to account for this, so again, it's something to keep in mind.

You may see some argue that GSAA is hampered by not taking shot quality into effect. No stat does; this is not a 'weakness,' particularly as there is no statistical proof that shot quality is controllable.

Hockey analytics are far from the world of baseball (frankly, every sport is). GSAA is a newer #fancystat that can definitely help fans and teams evaluate a goalie. It's worth noting, by the way, that noted free agent Optimus Reim's GSAA this season was 11.64.

In case you're curious, Broad Street Hockey has a great breakdown of this season's netminders here.