So, About Those Goalies: Short Leash Edition

There have been a lot of questions about the Tampa Bay Lightning goaltending these past few weeks. We may even have the beginnings of a...dum-dum-DUMMMM... goaltender controversy. (We probably shouldn't, but that's another post.) I got to thinking and I decided to try to take on one of those questions, namely: Is Guy Boucher favoring Dwayne Roloson? More specifically, does Mathieu Garon have a shorter leash and less room to make mistakes than Roloson? The corollary to that—the fearsome if unstated implication—is, "are the coaches are putting wins at risk by trusting Roloson or by not trusting Garon?"

The short answer? Not really. The long answer begins with a critically important caveat: the very small sample size involved makes discovering trends iffy at best. However, an examination of the situations in which Boucher makes a goalie switch shows a bit more about what gets a guy pulled and what keeps a guy in. To my surprise, I found no evidence that suggested there was much difference in when either goalie was pulled.

There've been four games where the goaltenders were switched. Garon was pulled in the October 27 game in Nashville and the November 16 game against Florida. Roloson was pulled in the October 13 game against the New York Islanders and the November 14 game at Winnipeg. The average save percentage for the pulled goaltender in those games was .753. Garon's save percentages were lower than this average (.706 and .750) when he was pulled, and Roloson's were split, one lower than the average and one higher (.722 and .818). The goal differentials involved were -3, -3, -4, and -3. Although we're talking about only four games, there's no readily apparent difference between the two in terms of when they got pulled. (All data comes from hockey-reference.com)

But since the real question is whether Roloson gets to stay in when Garon would get pulled, let’s look at the games where they didn’t get pulled. The fewest number of goals a Lightning 'tendy got pulled for giving up was 3, so our cutoff will be 3 goals against. Roloson has given up 3+ goals in 7 of 11 starts (63.6%). Garon has given up 3+ in 5 of 8 (62.5%). Again, there's no obvious difference between the two.

(In the table, highlighted games were ones with a goalie switch. The numbers are individual, not for the entire game.)

To understand what might keep a guy in the game in these situations, I considered offensive support (goal differential), timing of the goals, and shot volume (save percentage). What I discovered was that the single most important factor was the timing of the goals. When the goals against were spread out over the course of a game (as in the St. Louis game on Nov. 12) or if the goals came late in the game (as in the Florida game on October 17), a goalie stayed in the net. When the goalie gave up a series of quick goals, early enough in the game, he could get pulled, if (factor 2) the team was not providing sufficient offensive support—that is, when the goal differential was large enough to warrant a change in strategy. As long as the team was scoring goals, Boucher waited until the opposing team was up by at least 3 goals to switch goaltenders.

Surprisingly, shot volume appears to be an uncertain factor in whether a goaltender got pulled, or at least a factor whose dimensions aren't clear in this sample. In the one instance where a goalie gave up 3 goals in a single period (here the 2nd period) without offensive support and was not pulled (Nov. 19 against New Jersey), the goalie (Roloson) had a save percentage of .850 after giving up the third goal. He was seeing a lot of shots, and thus might be expected to let in more goals. The highest SV% for a 'tendy who got pulled was .818, and that was Roloson rather than Garon. The lowest save percentage for a goalie who wasn't pulled was .788 (Roloson), still well above the average SV% when goalies were pulled (again, .753).

All of this suggests that the coaching staff may be relying less on the intangibles of trust or reading a goalie's mental make-up than they are on how the game is unfolding. They may be considering mostly when the goals against are occurring and whether the team is likely to be able to score enough to overcome the deficit, as opposed to any other factors. There is no evidence that Roloson is more likely to stay in the game after giving up a series of quick, early goals without offensive support than Garon.

It will be interesting to see how this dynamic plays out as the season progresses, and whether the markers that are showing up now change as more games are played.