Arik Parnass is an editor for the SB Nation blog Eyes on the Prize. Follow him on Twitter @ArikParnass
During the pre-season, Justin Bourne published a fantastic series highlighting a unique — somewhat analytic, somewhat systematic — element of each team that might interest readers. For the Florida Panthers, Bourne chose to focus on their new (old) goaltender, Roberto Luongo. Luongo, Bourne claimed, had thrived under ex-Canucks head coach Alain Vigneault in part because of the coach’s preference not to have players attempt to block shots. Vigneault would rather his players not stand directly in the shooting lanes, thus allowing his goalies a clearer path to see shots and make easier saves. Under such a system, defenders box players out in front of the net, preventing them from tipping pucks or getting to easy rebounds. The opposite of such a system is referred to as “fronting,” and it is exactly what John Tortorella brought with him to Vancouver last year. Tortorella demanded that his players block as many shots as possible, meaning fewer get through to the goalie, but those that do can pose a greater problem. Bourne quoted Luongo as describing the system as follows:
“You better block it, because if you don’t, I won’t see it. A couple of times guys were trying to block shots but they weren’t sure if they should or not and they didn’t end up blocking them,” Luongo said. “It goes along with the territory. There will be a learning curve.”
It would be logical to assume that a goalie’s numbers would be adversely affected by a system in which his defenders screen him as frequently as opposing forwards, but is that truly the case?
Last season, I used a metric I called Possession-Independent Shot Blocking (PISB) to ascertain the rate at which teams block shots. It simply divides shots blocked by total shot attempts against (on the road, to account for scorer bias). In order to get a basic idea of how team shot blocking strategy and volume affect goalie numbers, I found the correlations between road PISB/plain old blocked shots and save percentage.
|PISB (Road)||Sv% (Road)||0.15|
For the 2013-14 season, there was a slight positive correlation, in fact, between shot blocking and goaltender save percentage, but nothing nearly concrete enough to make any determinations.
Goaltending is incredibly difficult to analyze because the sample sizes involved are minute. Try to assess the ability of a goalie whose true talent only emerges after approximately 3000 shots, but who works to improve his game every season, adds new saves to his arsenal, gains mental fortitude, and thus renders a portion of his results from the past irrelevant. Put simply, by the time you have an adequate sample size — with a measure like save percentage — to evaluate a goalie, your numbers are out of date because they began half a decade ago. With smaller sample sizes (like in the regressions above), the numbers fluctuate wildly, so even if there is a connection between shot blocking and save percentages, those numbers can’t catch it.
Still, I decided to examine Luongo’s career specifically to see if there was any connection between the teams for which he’s played, in terms of willingness to block shots, and his save percentage.
|PISB (Road)||Sv% (Road)||0.63|
The sample size is small, and thus once again I’d hesitate to make any firm conclusions, but a 0.63 correlation is quite strong. For Luongo specifically, a higher shot-blocking rate has generally led to a higher save percentage. It’s possible that years in which his teams have blocked a lot of shots have happened to correspond to years in which he has played his best simply by luck, but it’s also conceivable that Bourne’s thesis is in fact wrong, and that unblocked shots from the point, and the rebounds that followed, were giving the future Hall-of-Famer problems.
A Major League Baseball executive once asked a panel of students including myself if we knew what the biggest issue front office personnel were having in correctly applying analytics. He answered with an example. Let’s say a player has several years of major league experience in the same city. He then gets traded and his performance dips. How can you gauge the sustainability of that dip? After all, a regression explanation can only take you so far. If a player doesn’t feel comfortable in a city or an organization; if the style of play doesn’t suit their strengths; if he picked up a knock, those factors can cloud numbers in such a way that regression to the mean isn’t linear.
While hockey faces many of the same issues, the idea of goaltending may pose an even greater one. Without puck and player-tracking technologies, how can one accurately assess goaltender performance with every shot constituting a practically unique set of circumstances? It’s a question that has yet to be answered, and one that makes relying on current analytics a tenuous position.
It’s difficult to say with any confidence that shot blocking leads to better save percentages, but until our metrics improve, I certainly wouldn’t assume that Bourne’s claim is correct. In fact, I’d wonder if it might even be the opposite.