My Take On Advanced Stats.

 

The Globe & Mail’s James Mirtle recently explained how he become intrigued in advanced stats in hockey, particularly the debate over how they applied to the Toronto Maple Leafs this season. 

Mirtle’s piece reminded me of my own discovery of advanced stats, or “hockey analytics”, and how over time I also came to believe in their validity.

Do advanced stats like Corsi provide an accurate measure of performance?

Do advanced stats like Corsi provide an accurate measure of performance?

Around the turn of this decade I began hearing terms like “Corsi”, “Fenwick” and “PDO” kicking around the hockey blogosphere. Googling these terms, I discovered Corsi and Fenwick were  measurements of shot attempt differential, except that Fenwick doesn’t measure blocked shots, while PDO is the addition of shooting percentage and save percentage at even strength.

As I understand it, these stats measure puck-possession, the theory being that the team which controls the puck generates more scoring chances. Corsi and Fenwick are not only used to measure a team’s possession numbers but also to evaluate individual players.

I confess I was skeptical of these statistics, not really see how these measurements had any significant impact upon the NHL game. Like Mirtle, I didn’t really begin to pay attention to them until this season. What sparked my interest were comments made last summer by Maple Leafs forward Joffrey Lupul and Toronto Sun columnist Steve Simmons dismissing the validity of advanced stats (particularly Corsi). When analytics bloggers predicted the Leafs poor puck-possession could hamper their playoff hopes, the debate was on.

As Mirtle observed, early in the season the Leafs were among the top teams in the league despite their poor puck-possession numbers. That had critics like Simmons scoffing over the validity of Corsi.

The analytics crowd, however, stuck by their guns, claiming the Leafs couldn’t possibly maintain their then-torrid pace. They were ultimately proven right when the Leafs struggled with consistency over the course of the season before collapsing down the stretch and missing the playoffs. By season’s end they had the league’s second-worst Corsi numbers, ahead of only the woeful Buffalo Sabres.

Looking further, I was amazed to see that, of the 14 NHL teams which failed to make the playoffs, the Corsi numbers of all but four (Ottawa, New Jersey, Phoenix and Vancouver) were in the bottom half of the league.

What finally won me over on Corsi was the numbers for the 16 playoff teams following the first round of this year’s playoffs. Of the teams in the bottom eight for Corsi numbers, all but two (Montreal and Anaheim) were eliminated from the first round.

I now believe advanced stats are valuable tools for evaluating team or player performance. Bear in mind, of course, advanced stats aren’t infallible. While they shouldn’t be taken as unshakable gospel, neither should they be blithely dismissed. They belong in the equation for measuring performance.

For those of you interested in learning more about advanced stats, check out the following:

“Advanced Hockey Stats – An Introduction”. SportingCharts.com.

“Frequently Asked Questions about Statistical Analysis in the NHL”. Arctic Ice Hockey.

“Fancy Stat Summer School” – Habs Eyes On The Prize.

For more information on advanced stats, check out the following sites: ExtraSkater.comBehind The Net, Hockey Abstract, Department of Hockey Analytics, NHL Numbers, Arctic Ice Hockey, mc79hockey.com and  Stats.HockeyAnalysis.com.

5 Comments

  1. I like advanced stats. I think there is a place for them in hockey but do have an issue with them. With all statistics, especially ones that try to predict the outcome on things like hockey games which involve a lot of luck, aren’t as full proof as they seem. First off, one has to understand that for the most part, good stats need a lot of data, meaning one game or one week of stat collecting won’t give you a good picture of what is going on, you need a lot more than that. That in itself is a problem because stats need consistent variables, ie, same group of players playing on the same line on the same team for the same coach for a good chunk of the season which almost never happen. Also to keep up with the apples to apples comparison, players on other teams are rarely utilized the same way by different coaches and have different support staff playing a different style of hockey. To me that’s a hard sell. For an example the ulitization of phaneuf this past season, facing the toughest mins and most dzone starts than anyone else in the league won’t be a fair comparison to say someone like karlson.

    I think trying to predict the outcome of sporting games is great if you’re a person who loves to place bets on games. Otherwise I can’t see how the information is transferable from one team to another, in the event of a player trade or how one year results translates to another year of similar results simply because so much will change that ultimately have to affect the numbers.

    • There’s an old saying about three kinds of lies – there are lies, damn lies and statistics. Having said that, the correlation between advanced statistics in hockey and team performance is extremely strong. No question that, for individual players, advanced stats should only be examined in context of many other factors, as you have correctly pointed out, Ron.
      But I’m sure of one thing – under the Brendan Shanahan regime, the Leafs will use hockey analytics a lot. You can take that to the bank.

    • Since Phaneuf was brought up I’ll comment on him and his effect on the Leafs….

      - if he plays two many minutes it is a combo of both his and Carlyle’s stupidity. He obviously has a say in it and most team players accept lesser role when overwhelmed
      - early in the year his poor coverage was shielded by Bernier’s play
      - Phaneuf starts many shifts in his own zone due to poor ability to control the puck and breakout properly; I have zero confidence that he’ll do a proper breakout pass
      - Phaneuf plays vs toughest opponents? Well he also plays with Kessel line, can’t imagine his +/- without all of their even strength goals
      - His image as a tough defender with punishing hits is laughable. Teams all comment how little fight they get from T.O. and there is no fear dumping it in vs us.
      - Package him, and propect, and our top pick for Shea Weber. Weber, Gardiner, Reilly and Finn go forward will be best top 4.

      • Imagine if he came off the ice or refused to go on for a face off in his own end? How on earth are the minutes or zone starts Phaneufs stupidity? Look at Kessels minutes the look at Dionsalong withtheir zone starts, now how is he always on the ice with Kessel? And why if he is so bad would the Preds want to trade a prospect and Phaneuf for Weber?

    • Good teams use them,and they have proven valuable in ways to improve a team more than predicting outcome Blackhawks actually have their own variation of fancy stats, which plays a big role in every aspect of their team including negotiating contracts. The Blues employ an analytics scouting/coach no surprise that the 2 teams that rely heavily on the info they provide are near the top of the league in these stats, and subsequently overall standings, but when it comes down to it they are just tools like any other stat and not necessarily a crystal ball.