Thursday, August 13, 2009

The New School of Hockeynomics


Yesterday, James Mirtle penned an entry in “From the Rink” discussing the “new” statistics in hockey. In it, he referenced an e-mail he received asking about the current state of statistical analysis applied to hockey. The key questions in the message were:


“What are some of the most important new metrics?"

“What are their limitations?"


These questions really get to the crux of the matter. Personally, I’d hate for hockey to go the way of baseball, where countless “statistics” have been constructed that have little use other than to give “seamheads” something to read and argue about over message boards.

There is a distinct difference in my mind between “data” and “information.” In baseball, a lot of folks latch onto a piece of data – on base plus slugging percentage, batting average with runners in scoring position with less than two outs in the seventh inning or later in weekday day games, whatever – and assume those data are meaningful.

Well, are they? Do they inform an evaluation of a player’s effectiveness? Do they credibly explain a team’s success? Can one demonstrate a relationship between that statistic and a result? If not, then all those statistics are, is a collection of useless data. They aren't “information.”

The school of hockey statistics is well served with the question Mirtle asks to close his entry…”if [he] was to put together a list on this site of the five to 10 key ‘new’ metrics you'd like to see creep into the league's stat books, what would it include?”

But in it lies a danger, that hockey will use baseball as a data analysis template in an effort to create a generation of “puckheads” that is analogous to “seamheads.” Frankly, I don’t see that as being particularly useful in developing a body of analytical tools to evaluate performance. Those diligent souls who are at the forefront of developing those tools for hockey have a relatively clean slate to make those statistical tools meaningful, not just a collection of empirical gibberish.

To take a cue from Mirtle’s question, I’d be more interested in learning “why” those metrics should be included and “how” they inform an analysis of player and team performance.

6 comments:

  1. That's a pretty good read. All I know is Ovechkin's VORP is way better than Crosby's day/night split.

    ReplyDelete
  2. Snktimoniuz10:57 AM

    "There are 3 kinds of lies: lies, damned lies and statistics."

    I believe loosely attributed to Mark Twain.

    ReplyDelete
  3. exwhaler12:14 PM

    Thanks for this. The purpose of statistics is to measure an aspect of performance. It's a tool to help evaluate a player's ability and peformance. However, the tool must work in the first place.

    I'm both a hockey fan and baseball fan. While sabremetrics has revolutionized analysis in baseball(it's how the Oakland Athletics remained competitive for so long despite a low payroll), people usually miss the reasons why it's worked so well--baseball is an individualized sport in which the player's performance can be easily measured in the position he plays. A shortstop is rarely affected by what the outfielder does, and a pitcher is not affect at all by the position players. Aspects of the game can be measured clearly and variables--the destroyer of accurate statistics--can be easily identified and accounted for.

    Hockey, however, is a completely different game. Outside of the goalie, the players have cross-cutting responsiblities and the level of their performance depends on by their linemate's. Things like the Corsi Rating don't take into account multiple factors and in the end, do not tell you anything about the player's peformance. It doesn't measure anything but itself, which makes it a worthless statistic.

    I'm not sure if a sabremetric approach to hockey will ever work, unless the method in which data is collected is completely changed. It'll have to be a proactive review of every game with eletronic tracking rather than the traditional and somewhat subjective methods they use now. I just wish those that are pushing the new statistics would use little bit would use a little bit more caution; there's far too many variables in hockey that are not in baseball, and in this case, watching a player perform will probably tell you as much as a new but unproven metric.

    ReplyDelete
  4. @ snktimoniuz...

    I keep this quote on my wall in my office...

    "The government are very keen on amassing statistics. They collect them, add them, raise them to the nth power, take the cube root and prepare wonderful diagrams. But you must never forget that every one of these figures comes in the first instance from the village watchman, who just puts down what he damn pleases."

    -- Sir Josiah Stamp

    ReplyDelete
  5. Flying Cloud12:34 PM

    I've been reading "The Drunkard's Walk: How Randomness Rules Our Lives" by Leonard Mlodinow over the summer, with great pleasure despite being math-challenged. I started it in hopes it would help to better appreciate your insightful analysis, since it's been many years since I took a statistics class, not to mention never having actually played hockey even in my dreams. It's quite illuminating and entertaining. I think the author would agree with you, Peerless.

    ReplyDelete
  6. 1) "What are some of the most important new metrics?"

    2)“What are their limitations?"

    Aren't the answers pretty simple?

    1) Wins & Losses.

    2) Sometimes the best team doesn't win, that's why in the polayoffs they play 7 game series...

    That's also why Ovie is so careful and insistant he "hasn't won every prize in hockey ... yet..."

    LETS GO CAPS!!!!

    ReplyDelete