Why do good goalies have crap seasons? (Trending Topics)

Why do good goalies have crap seasons? (Trending Topics)

There was a terribly interesting paywalled story on the other day that attempted to explain why Kari Lehtonen, who was eighth in Vezina voting last season and probably deserved to be higher, is having a crap year. 

His career average, at this point, is .916, and he hasn’t been below .911 since 2005-06. That is, until this year, when he’s stopping just 90.8 percent of the shots he faces in all situations. So the question is simple: what’s wrong with Kari Lehtonen?

The short answer is that it’s incredibly difficult to say; Dallas has seemed rather porous this year in comparison with last year, and that’s showing up across the board. Backups Jussi Rynnas and Anders Lindback both have save percentages south of .864, and so the obvious answer to hockey people is that the shot quality the Stars are giving up must be quite high.

But here’s the problem: Given what we know about the NHL these days — i.e. how shots are tracked and the kind of work that goes into such efforts — unless you’re sitting there doing things your own self for every single game every single night (which is impossible) you’re not going to be able to come up with a satisfactory answer for the “shot quality” equation. As cited in that Custance piece by former bad backup Steve Valiquette, there are major discrepancies from one building to the next in terms of how shots are counted.

We know that Florida, back when Tomas Vokoun was playing there, dramatically overcounted shots to such a ludicrous extent that they probably inflated his save percentage by as much as 10 or 12 points per year, which is about the difference between Steve Mason and Henrik Lundqvist’s career numbers.

But Valiquette’s assertion that, “Save percentage is a joke,” shows a profound lack of understanding of what it actually tells you. As much as it’s nice to make fun of a .850 save percentage in any individual game, the fact of the matter is that a goaltender could allow five goals in a game and be at fault for none of them; they could all be off deflections and 2-on-0 breaks and any number of other things for which we know by watching that a netminder cannot be held accountable.

Hockey fans understand this intrinsically. This is true over one game, five games, 10 games, 40 games, even an entire season in most cases. The single-season record for the number of shots on goal faced by a netminder is 2,488, set by Roberto Luongo in the post-lockout 2005-06 season. At the time, he played for Florida, which again is known for dramatically inflating these numbers, but even if he faced 500 fewer shots than he actually did, 2,000 shots against is still enough to slot him into the 58th-busiest season of all time.

(Valiquette’s system of evaluating goaltender performance is better explained in this lengthy video, but never fully explained, nor is his data proven out anywhere. “Royal Road,” “Green shots,” and “Red Shots,” are all basically terms he made up for what we’ve always called “the slot,” “quality scoring chances” and “shots from the point.” It’s nonsense, and it only serves to illustrate one thing: 24 percent of league goals this season have been shots from the point, which he treats as a somehow insignificant number.) 

Some very smart people who do this kind of statistical analysis in exchange for money have found that goaltending statistics don’t get really and truly predictable until a player has faced a shot volume in that 2,500 to 3,000 range. Most goalies probably only face about 1,500 shots per year these days, meaning that outliers present themselves in this league regularly.

Eric Tulsky had a fascinating breakdown of this sort of thing for SBNation ahead of the start of last season, and to some extent these fluctuations are predictable. (The animated gif of Luongo’s career is particularly instructive for how helpful the simple passage of time can be.) The reason it’s important to gather that much data as a means of evaluating and comparing goaltenders is that, by the time they’ve faced 3,000 shots, the sum total is probably going to be of roughly equivalent quality. 

For this reason most guys who keep close tabs on this stuff are willing to say that Steve Mason posting a save percentage 16 points higher than his five years in Columbus (.919 to .903) is not really all that sustainable even if he’s faced 2,690 shots for the Flyers to date. 

The likelihood that he’ll regress to anything resembling his career number shrinks by the day, but going from .903 to .919 — especially behind this particular iteration of the Flyers defensive corps — is extremely unsustainable, even if it has been 92 games. 

The truth is that all kinds of goaltenders are seeing volatile changes in their numbers so far this year. Lehtonen’s is down appreciably from his career average, and Mason’s remains up. Ondrej Pavelec enjoyed probably the best two months of his time in Winnipeg to start the year but has been slowly slipping back to that guy the stats have always said he is (and he’s consequently having his job usurped by Michael Hutchinson). Pekka Rinne posted 67 games well below his career norms over the last two seasons but this year he’s on the fast track for the Vezina because his numbers are now 14 points above his career average. Marc-Andre Fleury has dazzled in Pittsburgh at .928, up from his .911 career save percentage. You can go on and on and on like this. Tuukka Rask is way down, and Jonathan Quick is a little bit up (and has been since last year, in fact).

To what can we attribute this? Is save percentage a fundamentally broken stat all of a sudden? Are years of using it to evaluate goaltender performance about to go out the window? The answer to these questions is, sorry to say for the Valiquettes of the world who have wasted hours trying to crack the “shot quality” conundrum: Sample size, no, and no.

Cory Schneider is the busiest goaltender in the league this year — and another guy who’s perceived as dramatically underperforming so far — and he’s faced only 1,009 shots in 35 games. That .918 is down a lot from his career average of .924, or at least enough that people are starting talking like his contract was Another Classic Lou Lamoriello Gaffe.

(The thing with goaltenders is that they play so many minutes and they’re so important to winning and losing that goaltenders who can reliably post save percentages even a point or three over the league average all but guarantee you a playoff spot and for this reason can be viewed as paying for themselves in terms of win value within a season or two, easily.)

And that’s certainly within the realm of the standard fluctuations from one year to the next.

Moreover, we’ve come to understand that goaltender performance can be better evaluated by way of looking at even-strength save percentage rather than the overall number, because that controls for more factors; of course a netminder is going to give up goals on a higher percentage of shots on the power play, and so forth. And by this measure, we can look and see that Lehtonen, to stick with that example, has seen his ESsv% come in at .922 so far this year. And a further breakdown shows that he’s gobbling up low- and medium-percentage chances, but getting smoked on the top-quality ones; his save percentage on high-quality shots is actually at the lowest it’s been since 2011-12, but more in line with his career numbers. For another example to explain this kind of thing: Rask’s has dropped off even more precipitously.

And here’s the thing: Lehtonen is not facing a larger number of high-quality shots on goal this year than he did last year, on a per-game basis. In fact, he’s facing fewer (5.32 per night this season versus 5.62 a season ago). Rask’s hasn’t really changed at all (4.59 to 4.6).

So what does that tell us? If we know high-percentage shots against Lehtonen are down, but a larger number is going in, then it’s not so much that the Stars are more porous now than they were last year; they’re actually tightening things up. Lehtonen just isn’t making the stops, and whether it’s just bad luck (which, again, I’d doubt because the numbers he’s posting now look very similar to his career averages), simply a matter of regression, or both, it doesn’t really matter. And again, it’s so early. Lehtonen has faced just 707 shots at even strength, meaning he has a long way to go before we can declare him washed up or a victim of bad bounces. 

No one is saying shot quality doesn’t exist. It very much does, and demonstrably so. But all the evidence has always suggested that it’s not controllable for the most part. Goaltenders are good, average, or bad, and the only thing that can change is the sample size in which their performance is viewed and the quality of the team in front of them.

Ryan Lambert is a Puck Daddy columnist. His email is here and his Twitter is here.