Well, it’s been a wild ride on twitter for the past few days. If you follow the hockey analytics crowd on twitter you probably know what I’m talking about, though for those of you that don’t, let me fill you in. I think the best place to start is the beginning of this recent road and it’s a bit of a bumpy ride so buckle up.
As I’m sure many, if not all, of you know, on Thursday afternoon the Leafs placed Grabovski on waivers for the purpose of buying him out. I think it’s pretty safe to say that this news was a shock to most of us, but none more so than those heavily involved in the “advanced statistics” community.
NOTE: As a slight detour, I placed that in quotations because I, like some of the others that use these numbers (Corsi, Fenwick, etc.) prefer to call them possession statistics or just statistics. The reason is relatively simple, “advanced” is a term that differentiates those numbers from the ones tracked by the NHL and in order to help the possession numbers become more widely used and more understood, the end goal is for the two to be treated homogeneously (with the appropriate trust in the different statistics based on the merits of each).
Why was this move such a shock? Well, because Grabovski is a good hockey player that signed an expensive (pretty much accepted that he’s overpaid, the degree to which is open for debate) contract because he put up points. This season however, instead of playing Grabovski in an offensive situation, Carlyle gave Grabovski some of the toughest minutes in the league. He started the vast majority of his non-neutral zone shifts in the defensive zone against opposing teams’ top players, making it very difficult to generate offense. On top of that, he played with line-mates that weren’t exactly going to set the world on fire offensively themselves, exacerbating the problem of generating offense. On top of that Grabovski had some bad puck luck and a gastro intestinal issue that’s he’s been dealing with for quite some time. In past seasons when he’s been used in an offensive role, he’s been a possession monster for the Leafs. This season the Leafs were a train wreck possession wise but Grabovski was still good relative to the team, which is very impressive considering his role.
Grabovski wasn’t placed in a position to succeed offensively, was instead given some of the toughest minutes possible, dealt with bad luck and an injury, and was still above water possession wise relative to the team. So his buyout didn’t really make a lot of sense, regardless of the re-signing of Tyler Bozak that took place the next day.
The outrage this buyout generated amongst the hockey analytics twitterverse was met by opponents (or non-users) of possession statistics with cries of melodrama and claims of understanding the buyout because of Grabovski’s lack of production given his cap hit. Harsh words were exchanged on both sides, prompting a tweeter by the handle @smithdanielj to claim “Grabo’s buyout may prove to be the Archduke Franz Ferdinand assassination of the war between Stats guys and Traditionalists.”
The next day wasn’t much better as the Leafs spent big money bringing in David Clarkson and re-signing Tyler Bozak. The prevailing opinion amongst the hockey analytics people is that Bozak is an inferior player to Grabovski, an opinion not shared by most mainstream media types (as evidenced by TSN placing Bozak as their #4 UFA with Grabovski as #9). Before noon, when FA Frenzy officially began, news of Grabovski clearing waivers was yet to be announced and the opponents/non-users of possession statistics were forwarding ideas that if Grabovski cleared waivers the proponents of possession statistics were wrong abut Grabovski. This certainly didn’t help ease tensions between the two groups.
Saturday saw the intensity of the dialogue brought down to a simmer (-ish situation) but Sunday would easily change that.
On Sunday morning, Leafs forward Joffrey Lupul tweeted “contracts aren’t awarded by this CORSI i am hearing all about. They are awarded for an equal value of skill and depth (at a certain position” and “If you bring certain attributes and you play to win. I’ll take you on my team 7 nights a week. Lets not look at this like Moneyball.”
And with that, all hell broke loose. The dialogue developing from Lupul’s tweet once again pitted “traditionalists” (opponents/non-users of possession statistics) against “advanced statistics” people. Quietly tucked away in the debate are clusters of productive discourse but for the most part the claws came out and the name-calling began with earnest. Typical comments coming from the “traditionalist” camp were along the lines of “watch the games instead conjuring up all-knowing numbers from behind your abacus”, while responses from the “advanced statistics” people took the form of “show us something better Luddites”.
These are of course exaggerations of the comments (in most cases, some were actually like these) and there were a relatively large number of people saying, “Actually it’s possible (and preferable) to use possession statistics AND watch hockey (correctly) to evaluate players”. In fact, the purpose of using these numbers is to shine the light on what’s happening on the ice, a shared goal with video analysis. Both are typically used in good analysis to reinforce each other and explain abnormalities.
The ride is almost over, I just want to take a minute to say that I enjoy using possession statistics to augment my understanding of the game and that they’ve helped me focus my attention to useful activities while watching hockey games. These statistics aren’t “advanced” or sorcery at all, and in fact are really about finding ways to quantify (albeit with funny names) ideas that we hear hockey players talk about all the time (great visualization of these ideas by Blake Murphy called “EXPLAINING THE GRABOVSKI-BOZAK “STATS” DEBATE IN SHINNY HOCKEY TERMS”).
The point I’m trying to get across here is that these stats don’t paint the whole picture and don’t replace watching the games, and nor should they. However, they DO tell a story about what happened and when used with the knowledge of other variables like “character”, “skating ability”, etc., a more complete picture of a player begins to take shape. They don’t tell the complete story about what’s going to happen in the future either (something Cam points out in article below) but there are other measures that help with that (like knowing the peak scoring years for forwards, for example).
There have been some posts about this already (see Cam Charron’s Corsi and Moneyball and Dirk Hoag’s NHL Advanced Stats Aren’t So Advanced, But They’re Better Than Nothing) and you should definitely check them out. On top of that fellow MLHS writer @mORRganRielly has an open invitation to discuss possession statistics on reddit. I would also recommend reading Jesse Spector’s piece about the history of Corsi titled Who is Jim Corsi? Meet the name behind the number.
Wow, what a wall of text! Thanks for sticking around to the end and I look forward to your comments about possession statistics and their role in the NHL.
Taken from @mORRganReilly’s reddit thread are some learning links on possession statistics:
- Overview of Corsi
- Overview of Fenwick
- From Vic Ferrari via James Mirtle on Zone time and Corsi
- Statistics Article Index — links to multiple pages of statistical evidence
- Zone starts and player usage