2013 Report Card: Tuukka Rask

Statline:

What, you actually needed words to explain to you what a sparkly BAMF Tuukka Rask was this year. Were his deeds not enough!? Ok, fine. Once more with feeling:

Entering the season as a question mark in his ability to fill the starting role vacated by Tim Thomas, in spite of having done so the season before last Rask still hadn't been in NHL action long enough to say with any certainty that he wouldn't Raycroft on us. After one more full season carrying the bulk of the games and as the exclusive starter in a long playoff run, consider Rask legit.

In spite of a Vezina snub, he was in the elite ranks of the league all season, finishing the year third in save percentage, second if you consider Craig Anderson's 24 games insufficient - and likely to regress (surprise, he did in the playoffs!). Rask was also second in ESSV with .938, and 4th in GAA. He furthermore clocked in at number 1 in Quality Start percentage, making him the most consistent starter in the entire league by a 3.5% margin over Mr. Vezina Bobrovsky.

Game situation is pretty much meaningless in terms of evaluating individual goaltender performance, but since goalies are always victim to subjective overvaluing of scoreboard conditions, I'll humor this bias. Rask was still "clutch" enough for you. He had a stellar .941 when down one, definitely good enough to keep a team in the game, and an obscene .954 when clinging to a one goal leading margin. Pity he was ONLY .923 when tied. And as established before, he was the best third period goalie in the league.

And if you've any concerns about his ability to carry a starter's load in a full season given that this audition callback happened in a lockout shortened year, remember that he played 58 games between season and playoffs, good for 70% of a normal length campaign. And in a more compressed schedule to boot.

Which brings us to the post season. Once again, a sterling record. He lead the playoffs with a .940 SV%, which incidentally falls the same .017 distance from playoff average as his .929 regular season figure from the season average. He was 4th in GAA, but within a tight pack that landed him .04 GA/game from the top spot. Stupid shot volume! Also, yet again, he was the most consistent around, only delivering five games below league average, even as that average rose from .912 to .923 from the regular season. That QS% actually went UP a touch from his exemplary in-season tally.

But still, that bum let in two goals when it REALLY COUNTED! Trade him!

Hang on a sec: we've been pretty damn fortunate in goal for the past several years, something must be fishy here. Rask is virtually duplicating Thomas' numbers, begging the question of whether they were both being propped up by a top notch defense. Are we seeing goalie talent or defense driving these numbers?

We've covered PDO a good deal around here and it's been taken as accepted that this composite figure does indeed regress toward league average. As Gabe Dejardins showed, after a first 1000 shots, PDO regresses 87% toward the mean for the remainder of the season. See his chart at Arctic Ice Hockey, where he plots regression for this stat at various shot-counts. The two input variables, on-ice shooting percentage and on-ice save percentage, are by definition driving the overall regression figure, both doing the same themselves. Eric Tulsky at Broad Street Hockey recently looked at shooting percentage over 3000 shots and found a 67% expected regression. So essentially, shooters have limited control over maintaining their shooting percentage and manufacturing higher percentage chances in any sustainable way, requiring us to adjust projections according to their out-or-underperformance of the mean. Right Tyler?

And why are we talking about this relating to a goaltender? Because save percentage works the same way - it's an input value in the luck-proxy of PDO and is likewise heavily prone to strong regression on an individual level. Skaters, in essence, have very little control or influence over save percentage - at least not enough to evince a clear repeatable skill of any value to a goalie. They really aren't able to do anything to make a shot less likely to go in, in any sustainable way. Specific to defensemen, we turn again to Tulsky. Looking at all defensemen who got at least 1000 minutes/yr in two three-year periods - a 97 player pool - he found a super low correlation coefficient that resulted in an 85% expected regression. (Forwards: 94%!)

So with league-wide findings in mind, lets turn to the Bruins. I switched up the methodology slightly to look at two-season segments rather than three so as to have a not completely empty dataset of Bruins fitting the criteria, and as a happy accident, to have a season each of Rask and Thomas starting in both samples. Word of caution, this is obviously a small pool and chock full of noise. Among defensemen, I found coefficient of determination of .05, resulting in a 1-r of .77. So based on the huge fluctuations year to year in on-ice sv%, our expectations of subsequent years should be regressed by 77% toward the mean - in a somewhat similar ballpark to Tulsky's more extensive data pool. Since we're looking at only Bruins, let's regress relative to our starters' six-year mean rather than the league average: .934. So a guy posting a .920 one year, .014 below mean, should project up to .931 the following season. Which incidentally is almost exactly what happened for Boychuk between 2013 and his 2011/12 season. Ditto for Seidenberg. Taking two year samples of forwards and D, I wound up with a 1-r of 89% - as expected, the forwards have even less influence and pull that percentage up. Basically, on an individual skater level, our players aren't really controlling their on-ice save percentage compared to anyone else on the team on a year-to-year basis. There's a teeny tiny sliver of skill and repeatability, but it's heavily overwhelmed by chance.

I stress: this sample is not valid as a predictive tool and analysis should be expanded to a much much bigger group of players, it is merely meant to illustrate the concept and that the team as a whole isn't doing much to influence save percentage. (I'm sparing you the scatter plots because the small amount of data points is goddamn comical, but if you want the backup just speak up in the comments) Year to year, individual percentages fluctuate significantly while the overall goalie save percentage has remained a relative constant for us over several years. No one player is making a consistent marked impact relative to their teammates season to season. Skaters can influence GAA by preventing shots from reaching the goalie or preventing them from happening altogether, but when it comes to their ability to exert any real control over how likely that shot is to enter the net, it's pretty much all up to the goalie.

For another, far more glib example, McQuaid's average On-ice SV% is the same as Chara's, and the former a better relative differential between On-and-Off save percentage...

Rask don't care who's on the ice. He's just that damn good.

Summary:

Really? Really? You need it spelled out: A.

And he's getting mad paid for it.