clock menu more-arrow no yes mobile

Filed under:

Average Goaltending, Expectations and Aggregates

A look at the Caps’ goaltending and why it might not be what you think

NHL: Washington Capitals at Montreal Canadiens David Kirouac-USA TODAY Sports

“If only the Caps could get even average goaltending...”

You’ve heard it. You’ve probably lamented it yourself.

Here’s the thing: the Caps have actually gotten better than average goaltending... from a certain point of view.

But first, what do we mean when we say “average goaltending”? A League-average save percentage in all situations (currently .909)? Or at five-on-five (.919 at present)? Maybe we mean some amorphous idea of “making almost all the saves he should make and a few he shouldn’t.”

The problem with those first two suggestions is that they’re fairly dependent on the defense in front of the netminder. Surely a .919 save percentage playing behind the Rangers defense that’s yielding nearly a scoring chance for every shot on goal they allow (0.98) isn’t the same as posting that same mark for the Avalanche (0.83 ). (For the record, Igor Shesterkin has a .940 five-on-five save percentage at the moment... hand him the Vezina now.)

So it turns out that “amorphous idea,” conceptually at least, is a better way to think about “average goaltending” in that it compares reality (“making the saves”) to expectations (“he should make”). And we have models that quantify those expectations based on numerous factors, mostly related to shot location and designations such as rebounds and rush attempts. They’re not perfect, but they do a good job of defining expected goals for the sake of conversations like these.

As a point of reference, per Evolving-Hockey, in Thursday night’s win over Montreal, Ilya Samsonov had an expected Goals Against (xGA) of 2.60 at five-on-five - based on the shots he faced, we’d have expected him to allow 2.6 goals at fives; he allowed one. If we look at all-strengths, his xGA was 3.62; he let in two. So his five-on-five Goals Saved Above Expected (GSAx) was 1.6 and in all situations is was 1.62. Over at MoneyPuck, the Habs had an expected Goals For (xGF) of 2.16 at fives and 2.88 overall, making Sammy’s GSAx 1.16 and 0.88, respectively. Different models, different numbers, similar story, and we’d certainly all agree that Samsonov gave the Caps above-average goaltending in Montreal.

That, of course, hasn’t been the case all year. To wit, Evolving-Hockey has Samsonov’s year-to-date GSAx at -0.41 at five-on-five and -1.16 overall; MoneyPuck has him at -0.3 and -3.2, respectively; HockeyViz has him allowing 66 goals... against 65.9 expected goals. In other words, in aggregate, Samsonov has allowed almost exactly as many goals as we’d expect, perhaps underperforming a bit on the penalty kill (which tends to be a bit more volatile in terms of goalie numbers due to sample sizes and other reasons).

Of course, the phrase “in aggregate” is doing a lot of heavy lifting in that sentence. And if you feel like the “when” has mattered with Samsonov, well, it has. Of the 45 goalies in the League who have played 250 minutes with a lead, Samsonov has the worst save percentage of the lot at .887; Samsonov’s netmate, Vitek Vanecek ranks 20th at .918. Dial in on five-on-five play (and lower the minutes requirement to 200) and Samsonov is 44th of 45 at .889 (Vanecek is 16th at .933). Protecting a one-goal lead? In all situations (minimum 125 minutes), Samsonov is 41st of 45 at .888 (Vanecek is 14th at .925), and at five-on-five up one (minimum 104 minutes to keep our tidy 45-goalie sample), Sammy is 34th at .907 (Vanecek is ninth at .945). And this all comes on a team that has played very good defense when leading (particularly when leading by one):

via HockeyViz

Again in aggregate. And aggregating numbers doesn’t account for consistency (which may not actually be as desirable as you think - more on that later), which seems to be a four-letter word to Samsonov and, to a lesser degree, Vanecek. Neither of them has been able to hold on to the number one goalie designation for more than a handful of games at a time - though it’s worth noting that Vanecek has been among the League’s best in terms of GSAx per sixty at five-on-five (minimum 20 games):

via MoneyPuck

Clearly, the Caps have among the most volatile goaltending situations on a night-to-night basis in the League, no?

Well, no.

If we consider a game in which a goalie allows within half a goal of what’s expected at five-on-five to be, well, as expected (or, to circle back to the top, “average goaltending”), the Caps are near the top of the League in the percentage of games in which they’ve gotten what you’d expect from their goaltender(s):

Note: Score and venue adjusted numbers were used here, but we don’t need to get into that now.
via Evolving-Hockey

Now, let’s be clear - being further to the left on this chart doesn’t mean “good,” but rather that the team more often has gotten an average-ish goaltending performance. It reflects the goalies’ play and nothing more - a team whose goalies always gave up as many goals as expected would be at 100% here, whether they were expected to give up one goal per game or five.

Where the rubber meets the road, then, must be when the goalies’ play doesn’t align with expectations, so let’s look at that. For the purposes of this exercise, we’ll consider a team whose variance from expected is between half a goal and one goal to be good or bad - save three-quarters of a goal at fives and you had a good game; allow 23 of a goal above what was expected and that’s bad - and above one goal to be very good or very bad (Ilya Samsonov was very good on Thursday night). Here’s how the first of those two groupings shake out:

data via Evolving-Hockey

Okay, so what the hell is that?

Along the horizontal axis is the percentage of “good” games - those games where the team saved more than half a goal above expected. The vertical axis are the “bad” games - those games where the team allowed more than half a goal above expected. Taken with the chart above (of “as expected” games), we should have all of a team’s games accounted for... and we do. For the Caps, they’re at 37.5 percent “as expected” (goals saved relative to expected between -0.5 and 0.5), 35.3 percent “good” games (more than half a goal saved above expected) and 27.1 “bad” games (more than half a goal allowed above expected). That puts them near the middle of the pack in good and bad games (slightly better in the latter), which you might expect given their relatively high rate of “as expected” games. Of course, Seattle was in that “as expected” zone even more than the Caps and, well, yikes - when their goalies haven’t been “average,” they’ve been bad four times as often as they’ve been good.

As for very good (and bad) games? Let’s take a look:

data via Evolving-Hockey

You’ll see the Caps similarly positioned here, though near the top of the League in “very good” games. Huh. Oh, and now Seattle has company, isn’t that sweet?

To bottom line this, the Caps have gotten “average goaltending” at among the highest rates in the League and, when they haven’t, more often than not it’s been above average... at five-on-five (overall, they’ve been in the black on GSAx in 56.3% of their games this season). Fix the penalty kill and that “at five-on-five” disclaimer might disappear entirely - per MoneyPuck, the Caps have saved the 8th-most goals above expected at five-on-five, as does Natural Stat Trick; Evolving-Hockey has them tenth. This is a good defensive team getting average-to-good goaltending (say it with me) in aggregate, particularly from Vanecek.

This is where you might expect something pithy like, “but NHL games aren’t won in aggregate, they’re won in moments.” Except they are - literally - won in aggregate. Every save is a “timely” one. As the Caps approach the trade deadline, they need to look at what each of their two netminders have done this year, and weigh their options, cost/price considered. Is there a goalie available on the market who is a clear enough upgrade over Vitek Vanecek that he’s worth the acquisition cost? Perhaps not. Can Ilya Samsonov build off his recent wins over Pittsburgh and Montreal and be a positive contributor (or even a reliably “average” goalie)? Guess we’ll see (again).

All of this isn’t to say that goaltending isn’t a problem for the Caps, but rather to say that it isn’t the problem. And that it’s not as easily solved as it might appear - it’s not as simple as “if they could just get average goaltending.” Because they are and have been. In aggregate.


Post Script

We tend to take for granted that consistency is desirable... but is it? Let’s say you have two goalies, each of whom gives up 30 goals over a ten-game stretch. Goalie A gives up three goals every night, while Goalie B gives up one goal in five games and five goals in each of the other five. If you’re a team that scores 2.5 goals per game, alternating 2- and 3-goal games, you’re going to lose half of Goalie A’s games and tie the other five (we’ll be generous and give you three shootout wins in those five), for a 3-5-2 record. With Goalie B in net, however, you’re going to win five and lose five. Sometimes “consistency” isn’t all it’s cracked up to be, especially for a team that’s outmatched.