Four Common Criticisms of Possession Stats and Why I Think They’re Totally Whack

Doug Murray thinks stat nerds should put their spreadsheets away and watch a game

Photo from Michael Miller via Wikimedia Commons

It’s been a big summer for analytics in hockey. Analytic minded Kyle Dubas was hired by the Maple Leafs as their assistant GM and Team President Brendan Shanahan admitted that some in the organization were scared of certain words and ideas without deep investigation into their merits. The Devils hired Sunny Mehta, a former poker player and Edmonton Oilers blogger, as their Director of Analytics — a position that reports directly to Lou Lamoriello. Two of the best hockey analytics bloggers out there in Eric Tulsky and Tyler Dellow got hired by NHL teams.

With all this going on, those staunchly opposed to analytics, who may or may not have an understanding of why they are considered useful by many, are coming out to defend the merits of actually watching hockey — which apparently stat nerds don’t do.

Below I take on four common criticisms of possession stats, seemingly the most hotly debated aspects of analytics.

1. They don’t tell me who the best players are

I actually don’t think Jake Muzzin is the best defenseman in the NHL. I don’t think Justin Williams is better than Sidney Crosby. Outside of the oft-referenced Corsi-dependent strawman (who I have yet to meet), I don’t know of anyone who does hold those opinions.

Possession stats aren’t supposed to measure all contributions a player makes to winning and losing hockey like WAR and other stats do in baseball. They measure how many shot attempts a team gets relative to their opponent when a player is on the ice. Yet many criticize Corsi and Fenwick for doing a poor job of evaluating what no one claims they evaluate.

I can’t remember the last time I heard a complaint that the stat goals scored doesn’t measure defensive clears, but for some reason possession stats are expected to encapsulate everything.

2. They ignore the intangibles like grit, heart, love and democracy

It’s often said that Corsi and Fenwick ignore the impact physical play and play away from the puck. This is a shallow criticism which I don’t think anyone who’s thought about it for more than a moment would make. To the extent that they impact the play, both would definitely be reflected in a player’s possession stats.

Winning a puck battle, by definition, means you get the puck. Players that wins puck battles frequently then, will see their team getting more shot attempts than their opponent, all else equal. Kyle Dubas made this point the day he was introduced as assistant GM of the Leafs.

So it’s not that Corsi ignores Douglas Murray‘s physical play, just that he’s terrible in spite of it.

What is the goal of strong play away from the puck? If you think it’s putting yourself in as position to win the puck back or stifle and opponent’s shot, then strong play away from the puck would absolutely be reflected in possession stats.

3. All shots are unique, like snowflakes

It’s true that possession stats make no distinction between a high quality shot and a low quality shot. I’m not sure why puck possession stats should be expected to, but let’s address the issue of shot quality anyways.

Shot quality is often the argument made for a bad possession player or against a good one. Of course Michael Del Zotto allowed less shot attempts than Kevin Klein, but they were all breakaways, as the common argument goes.

In reality though, suppressing shot quality is a skill that barely exists over the long term. Garret Hohl showed the defenseman in the top 10 percent of relative on-ice save percentage over a two year span kept less than one-fifth of their positive effect on save percentage in the following two seasons.

On offense, players can certainly have more of an impact on shot quality — I mentioned this in my Round Two Preview regarding Sidney Crosby and Evgeni Malkin — but I still feel the impact is overstated by most fans, and that offensive shot quantity trumps shot quality.

4. They credit players on bad plays penalize them on good ones

A few years back, Ron MacLean trolled the analytics community by showing a clip of Ryan Johnson diving in front of a shot to block it, and remarked to Don Cherry that he would get a negative tally towards his Corsi rating. The old-school NHL coach from the 1970s predictably did not react favorably in what was likely his first interaction with the stat.

But Johnson didn’t get the minus the moment the puck hit his shin pads. He got it the moment his unit of five players let the opposing team in their zone and take a shot at their goal. All five of them did. That’s what Corsi measures: which team is getting the shot attempts when particular players are on the ice. It’s likely that some of the five didn’t directly contribute to the opposition getting a chance at their net. Luckily, players can accumulate upwards of 2,500 Corsi events in a season, so over the entire data set the noisy ones should be drowned out.

Similarly, when you take a shot that goes six feet wide, you get a plus the moment it left your stick. It’s not rewarding you for the shot, but rather rewarding you and your teammates for getting into a position to take a shot towards the net.

I do watch the games, by the way.

4 comments

  1. dan says:

    Have to argue though that maybe stats community is biased against physicality that doesn’t necessarily effect possession. The “grind”those teams are looking for is very valuable in a series and my guess is Kopitars poss. Stats would drop had he be less willing to take hits to gain the zone

  2. Jared Sexton says:

    In what way would it be valuable outside of helping possession?

  3. Centerman21 says:

    Advanced statistics are used to show how a player performs while he is on the ice and how his team plays when he is not. They don’t say that Justin Williams is better than Sidney Crosby. Williams may be the better possession player over Crosby. Anton Stralman had one goal and thirteen points on the season. Yet because of his 56% on ice Corsi, he got a five year $22.5 million dollar contract from Tampa Bay. Someone believes in those darn Stats.
    You can watch what a player does skating up and down the ice and doing the little things all you want. However, in my opinion you can’t come up with a full evaluation without looking into his stats. They also go much deeper than just Corsi & Fenwick. There is also (QoC) quality of competition and zone starts. Does this player start 60% of his shifts in the offensive zone, playing against top competition and has a 45% corsi? Or does he play lesser QoC and start most of his zone starts in the defensive zone with a 52% corsi%
    These stats may not show who the best players are in the NHL but the best players have good or great possession statistics. Sidney Crosby is one of the best possession players in the NHL.

    Jared this is Robert. We have been in touch via Email. Lclnmbr1@Hotmail.com> My twitter handle is the one I used to comment here. @Centerman21.

  4. […] Sexton at Rangers Unlimited talks about four criticisms of possession stats and why those criticisms are straw […]