I had a long anti-fighting diatribe all typed up for today’s post to be in response to my evening/weekend TBS co-hort’s first article here at The Shelf a week ago, but, in the wake of the Arron Asham incident I put that one on the literal shelf until we know a little bit more about the practicality of fighting in the game.
The guy who wrote the post, Daniel Wagner, whose Trinity Western Spartan hockey team also lost a 5-3 decision to my Thompson Rivers WolfPack Friday night, had another wonderful post Thursday that had a philosopher’s name in the title and everything. It also drew a very well-reasoned response in a discussion about the limitations of statistics and expansion on what we already know as hockey fans.
With David Staples of The Cult of Hockey also writing a little about the minimal impact statistics had in shaping a hockey team, I figured this was as good of any as a subject to tackle today. The basic, most absolute problem that exists in turning new people over to the statistical world and joining what has become a very fine discussion, is that it’s almost like they expect us to champion “one” number as the measure for every hockey player in the world.
Staples indicates that “if Oilers GM Steve Tambellini had relied only upon Corsi plus/minus to determine which players to cut from last year’s team, the first three defencemen to go would have been Jason Strudwick, Theo Peckham and Ryan Whitney.”
“No argument on Strudwick, but cutting the tough and promising Peckham and the superlative Whitney? Something’s wrong here.”
Corsi, as it bears repeating every time its mentioned, is an advanced +/- statistic which tracks not just goals for and goals against while a player was the on the ice, but also all saved shots, missed shots, and blocked shots. It’s application is primarily that of a “possession” statistic, which means that a player with a positive Corsi rating is very likely to have held the puck in the offensive zone when he was on the ice.
The problem with a statistic like Corsi is that it’s not supposed to be the be-all and end-all in a statistical discussion, and we understand that all these numbers need to be looked at in context. On anybody’s list of top forwards in the NHL last season, who would have had Torrey Mitchell or Mason Raymond? In what universe is Alexandre Picard the best defenseman in hockey?
That’s not exactly what Corsi is for. Corsi can’t tell us which players should be cut (although I’d advocate against having Ryan Whitney in my lineup) Corsi just tells us where the puck was when a player was on the ice. Does it qualify what the player had to do with the situation? Was possession turned into chances? Did the player get into shooting position? Was this against easy competition? Tougher competition? Did the player have the benefit of starting the shift in the offensive zone? Is the player working within a system that allows his style of play to succeed? Does the score situation favour the player on the ice? Is the player good at special teams?
These are all questions that are vitally important to the value of a hockey player in the statistical world that Corsi alone doesn’t measure. We have ways of quantifying some of these things, and some websites like timeonice.com and behindthenet.ca soak up whatever information they can. If you’re a stat skeptic, it may be worthwhile to accept the limitations of what single numbers alone can offer. No number in the next ten years will ever tell you just how good a hockey player is, just as no single number in baseball (I often hear that baseball is more adept to statistical analysis because it is more linear. In fact, the major difference is that it is slower and easier to quantify what happened) can describe a player’s value in a single, objective way.
“In a way, we’re where baseball was 15 years ago,” Gabe Desjardins wrote in a FAQ section for his website. “They could evaluate offense, didn’t quite have pitching figured out, and didn’t really have the data they needed to evaluate fielding. Technology and improved record-keeping will allow us to close the gap in a few years less time, but it will not happen immediately.”
We’ve had offense figured out for a while now. We know that 30 goals is really good, 20 goals is really good for a second-liner, 10 goals is really good for a third liner, and anything you get out of your fourth line is a bonus. We don’t have pitching in hockey, but we’re getting there with defense. We know that a player with possession is probably less likely to give up a lot of chances against, but we can break down the Corsi number to snuff out the number of shots a player had against him when he was on the ice, and that can tell us something.
This is way more than we could say five years ago. It was one thing to say “well, this player’s coach trusts him on the penalty kill and gives him a lot of minutes, so he must be a good penalty killer” and defer to the coach’s expert authority. We can now cross-reference that with actual data pulled from NHL games to see whether or not it actually helped the team succeed on a man down. Adam Hall and Samuel Pahlsson play similar PK minutes, and Pahlsson comes with a higher pedigree, but I can say objectively that Hall was on the ice for one recorded shot fewer per six minutes than Pahlsson.
Moneyball was a long research paper objectively about baseball’s on-base percentage, just how “Moneypuck” isn’t a concept about finding players with good Corsi ratings. It’s simply about finding the players who do things we often forget about when they head off the ice for a shift, and crafting a fantasy team around them. More data doesn’t hurt, and neither does objectivism. But let’s look at what statistics really are: just numbers. Quantified measures of what happened when, and we have to take every single available resource we have, whether it’s Corsi, scoring chances, assists, giveaways, hits or time on ice, and put it into its appropriate context for the argument at hand.
By the way, the best player in the National Hockey League is Sidney Crosby.