March 26 2012 07:29AM
Oh, you want more? Ok. Bear with me because this is a very rudimentary statistical analysis, but it's my first time with graphs and stuff. I don't even know how many of them will make it into the post, but here we go!
So, first of all, let me say that I'm not the first person to do an analysis of the shootout, and further, this is certainly not an improvement on past articles, at least in terms of statistical analysis and rigour. But if you are like me, your eyes glaze over when the graphs come out. That is not to say that I don't "believe" in statistics. I would even say that I understand them, generally. Still, despite what some mathematical purists might say, I think there's some benefit in vulgarizing the more advanced work. The conclusion doesn't change, but it's explained for a more impatient audience, let's say.
As with most aspects of analysis involving a little math, the mainstream media gets it wrong. There's an easy narrative that says: hey, a shootout specialist, gotta have one of those, the game is changing. E.g. Pierre McGuire's quote in this, late December 2011, peice from the NYT. But is that really the case?
Ok, so let's start with some clear, but not intuitive numbers. First of all, the two graphs below tell us that shooting percentage in a shootout and save percentage in a shootout are not connected, statistically, in the least.
Ok, so have a look at those graphs. What one might think to be the case is that, the higher a team's shooting percentage in regular time, the higher it would be in a shootout. The same, seemingly obvious assumption might be made about a goalie's save percentage (I used overall save percentage, not EV SV%, but the effect is the same). But that's just not the case. A positive correllation, that is, if your shooting percentage (or save percentage) in the shootout improved as a function of your regular shooting percentage, the points would rise gradually to the upper right of the graph. For those statistically inclined, the r for the first graph is .08892493, and for the second graph it's 0.02421527. For those not statistically inclined, r is the number that describes the relationship between sets of numbers. It's a number between -1.0 and +1.0. The close to +1.0, the more of a relationship there is. Those numbers are low, in terms of describing a relationship.
So what helps you win shootouts? I present two more graphs, that are actually pretty obvious, too, but still worth looking at as a segue into the next part of this.
Here we have shooting percentage and save percentage in the shootout, each graphed against winning percentage in the shootout. Now we can see that gradual upward slope that indicates a relationship. Well of course! If you have a good shooting percentage or save percentage in the shootout, you're gonna win more! The r for the first graph is 0.77029147, and for the second graph it's 0.74884594. So that's pretty close to 1.0, so we can say there is a strong positive relationship. (there are a few more math steps to be done to confirm that, but let's not complicate things. Mostly because I don't know how to do them yet).
Ok, great, so how do we get these things that ensure us shootout victories? Leaving aside the goaltending issue (so nice not to talk about that!), let's focus on the goal scorers. If we could snag a player who was really good in shootouts, he could be a kind of specialist, and help us win more games. But does such an animal actually exist? One way that I thought would be useful to check is to graph, in a similar way, the number of shootout goals scored by the top player on each team (presumably the shootout specialist) with the shootout winning percentage. One would guess that if a team has one guy who scores a lot on a shootout, that guy is a specialist, and if he's scoring a lot, it's helping his team. Here's the graph:
Hmm. Well, it looks like there is some correllation. I mean, that's Ilya Kovalchuk way up top with 11 goals, and the Devils certainly do a good job in the shootout. But the Canes' top shooter, a guy known around the league to be a specialist, Jussi Jokinen, has only 2 goals in 5 attempts, and his team has a winning percentage of 0. Ouch. The r in this case is 0.24790228. That's a much weaker connection than just generally good shootout scoring. And, if you take out Jokinen (because his team is so bad) and Kovalchuk (because his team is so good), the r is 0.17457076.
So basically, the performance of your best shootout shooter is not strongly correllated to your shootout win percentage. But wait, you might say, this is a small sample size of one season minus about 8 games; if you look at lifetime stats, maybe there's a better relationship.
I picked season stats not lifetime stats because the result of a shootout, that is, a win or a loss, does not carry over from season to season. So what is more relevant in deciding whether carrying a guy is worth it? His lifetime stats or how he stacks up this year? Probably in making the decision to begin with, you look at lifetime stats, but when assessing whether you benefited from it in a particular season, you can only look at that season. Maybe it will even out over time, but ask the Carolina Hurricanes if they are ok with that. Jussi Jokinen has 28 career shootout goals. The Canes have not won in the shootout yet this year.
As for the Leafs, who do you think is the best shootout option for Toronto this year? Well, surprise, Tim Connolly has both the most goals (2) and the best percentage (he's perfect). Phil Kessel has 14 career shootout goals in 41 attempts, for 34%. He's 1 for 7 this year. Lifetime, Connolly is 7 for 22, or 31%. Nazem Kadri is 1 for 3 this season in the NHL. Hard to call him a "specialist" at this point.
The question of how much to pay a shootout specialist is another one entirely, but it seems to me, before you start writing cheques, you'd better see if the position is worth paying for. Seems not to be the case here.
As for the shootout itself, a brief editorial: the shootout is a result of (North) America's need for closure, finality, and of course, for some one to be declared a winner. The value of something, these days, can only be measured in relation, or in opposition to, something else. It is sad that we can't value a hockey game for what goes on during the 60min (or 65min), and that we value more the two points in the standings at the end. Am I alone in recalling that one of the greatest hockey games ever played was the 1975 New Year's Eve match between the Montreal Canadiens at the height of their powers, and the Central Red Army team that rivaled them as the best club team ever assembled? The game ended in a tie. A fitting end to a contest between such strong and equal sides.
A telling description of that came, from Joe Pelletier, international hockey author, is that “artistically, it was an absolute joy to watch.”
Yes, there is something special about an overtime goal, and in the playoffs, nothing is better. But in terms of the long haul of the season, is it not fair, after 65min, for two teams to skate off the ice and say, ok, nothing was decided tonight, we'll come back and do it again next time?