Hockey Roundtable: Advanced Stats

By: Liam McGuire

image

I’ve always been curious about advanced statistics in hockey. While I’m admittedly no expert, I feel they give us a great way to track and evaluate players and understand their performance in a new way. To give those - like myself - who could use more insight on the topic, I’ve asked five writers/bloggers to answer six questions on the topic in a roundtable discussion. Full disclosure all of these writers are proponents of advanced stats. In a follow up piece, I’ll look at the opposite side of the argument.  

The panel: 

Sean McIndoe, writer for Grantland and Down Goes Brown

Jeff Veillette, Marlies/Leafs writer for The Leafs Nation. 

James Mirtle, reporter for The Globe and Mail and co-host of Leaf Report Podcast

Steve “Dangle” Glynn - Hockey YouTuber, Managing Editor for Leafs Nation and Steve Dangle Podcast. 

Thomas Drance - News editor for The Score and hockey writer for Canucks Army. 

1) For those that don’t know what advanced stats are, why are they important? 

McIndoe: For the same reason that old-fashioned stats like goals, points or wins are important – because they help us keep track of what’s happening on the ice, including key aspects that we’d otherwise lose track of or even not notice at all.

Advanced stats also have the added bonus of being better than the old-fashioned ones at predicting the future, at least at a team level. Once you know which key stats to look for, you’ll become much better at figuring out which teams are likely to be better or worse than expected in a season’s second half, or which teams should be the favorite in a close playoff series. It’s an easy way to look smart in front of your friends.

Veillette: Advanced statistics are yet another means of player evaluation. Professional sports are a very cutthroat business with little margin for error, so any competitive edge or additional data you can get is a positive. The most popular advanced stats of the present tend to be based around attempted shots, as they are frequent events that tend to correlate with possession and the progression of play.

Mirtle: What they allow for is better (and I’d argue proper) evaluation of teams and players. The reality is the statistics that were available in the game prior to these analytics the last few years were wholly inadequate in terms of projecting how teams would do in the future. Analytics add another layer of insight into how well a team and/or player are doing. 

Glynn: Advanced stats are important because they give you another way to view the game. There’s a misconception by some that people who advocate advanced stats don’t watch the games or all they care about is numbers. Wrong. They’ve changed how I watch the game. If anything they make me pay even more attention. 

Drance: The so called ‘advanced stats’ are really just shot-based metrics. There’s nothing ‘advanced’ mathematically speaking about addition and subtraction, but that’s really what Corsi and Fenwick numbers are -  a tally of shot attempts (or unblocked shot attempts) for and against, usually expressed as a differential or as a percentage. Essentially it’s a different type of note taking, and a puck possession proxy, that focuses on which players spend more time in the offensive end of the rink than they do in their own zone. 

The shot based metrics have some limitations - they don’t really account for higher percentage shooters like Steven Stamkos and Sidney Crosby, they don’t factor special teams play into the equation, and they don’t really account for the quality of a club’s goaltending. What they do is provide a simple quantitative snapshot that allows us to judge which teams (and which individual skaters) are most often able to control the flow of play. All things being equal, the teams that generate the most shots will also generate the most scoring chances and will tend to - over time - outscore their opposition. This particularly matters come playoff time, when the referees put the whistles away relative to the regular season.

2) What are your thoughts on some NHL teams recent declarations of hirings and support of advanced stats? 

McIndoe: It’s not surprising. We’ve know that many NHL teams were doing this for years, including some that weren’t acknowledging it publicly. And any sports fan who was paying attention already understood that this was where we would wind up. It happened in baseball, it happened in soccer, in basketball, football, etc. The idea that this stuff would somehow skip over hockey was always a pipe dream

Veillette: As I mentioned in the opening question, you want to take advantage of any help you can get. The last thing a team wants to do is fall behind on methods of potential improvement. Even if you aren’t going to take this data and make it your central focus (few if any will), there is no point in withholding information from yourself, especially if your opponents are striving to have it in their repertoire. 

Mirtle: I’m not surprised. There’s a lot of value there that many organizations were missing and many of the league’s better teams have already invested in this. 

Glynn: It’s nice! I think it puts to rest the argument that NHL teams don’t take advanced stats seriously. As if. They’re a new tool in an ever-evolving and highly competitive game. Teams were using them and the ones that weren’t were probably figuring out what to do next. A few of those teams (Oilers, Leafs) have taken big steps in the right direction. They’re not going to magically become great overnight but they should improve with a new mentality.

Drance: I have a lot of admiration for the work that Darryl Metcalf, Cam Charron, Rob Pettapiece, Eric Tulsky, Sunny Mehta and Tyler Dellow have put together over the years; and I think, generally speaking, the hockey world did a surprisingly good job identifying and hiring some of the best, most innovative analytic talents in hockey blogdom. 

I’m not surprised that those guys got hired, though I’m a bit surprised that it happened so quickly. It took baseball 25 years to give Bill James a full-time position, whereas we’re less than a decade removed from the heyday of Irreverent Oilers Fans - and there has undeniably been a rush this summer on talented quants from the hockey blogosphere.

3) Why do you think there has been such a heated back and forth between certain bloggers and members of the media regarding advanced stats? Is it a failure of acceptance? Sensitivity?

McIndoe: Some of it is sensitivity. There are guys on both sides, bloggers and traditional media, who are amazingly thin-skinned and just can’t handle being told they might be wrong about something (which always strikes me as incredibly bizarre, given how much time we all spend criticizing players, coaches, GMs, etc — we can all dish it out, but we can’t take it).

So that’s part of it. But I think that another aspect is that there’s this “old school” side of the debate, which would be best represented by veteran scouts and longtime front office types and various other, for lack of a better term, hockey lifers. And the mainstream media guys typically have lots of access to those guys, and bloggers have almost none. So there is a side of the story – maybe the wrong side, but still a side worth hearing – that only half of the debate is directly talking to. And I think that skews things.

Veillette: The current “movement” represents a massive shift from the status quo, from a media perspective. The current model of sports journalism needs arguments to succeed. Statistics, occasionally have the ability prove the descriptions that “catchphrase journalism” give to players and team to be inaccurate, and can at times bring an argument to the point where it’s not worth having it. This is a direct threat to those in the media not willing to adapt, so they reject it on principle. “Bloggers”, unhappy with being brushed off without merit, get angrier and pushier, which makes the “media” types angrier and pushier, leading to a vicious circle of animosity. To me, a modern media member should be well versed in any means of forming their point, be it visual, statistic, emotional, etc, and you don’t see much of that. As well, you’re seeing more writers concerned about being accepted than being right. It’s close to impatience, though it happens for good reason. Either way, the drastic shift and open platform for the two sides to converge on each other has lead to a rather avoidable mess.

Mirtle: a) People don’t like math and often misunderstand how probability works. This is one big barrier to acceptance and understanding. 

b) People haven’t done the necessary research into how these stats were developed, how they work and how they’re interpreted. 

c) Much of the criticisms of analytics are therefore off-base and are predictably met with blowback from those that do use and understand them. 

Glynn: Oh boy. Long story.

Drance: Having come up in the blogosphere perhaps I’m biased, but I think it’s mostly just a resistance to change. Some people are reluctant to change the way they’ve gone about their business for 20 or 30 years and instead of evolving and incorporating shot-based metrics into their work, have dug in their heels, and criticized these new numbers and new ways of thinking about hockey as faddish. 

Those writers tend to get the bulk of the attention, but really, if you look a bit deeper, you’ll see that old school story tellers have been incorporating new data in their work for years. Whether it’s Jon Rosen, or Adrian Dater, or Jason Botchford, or James Mirtle, or Fluto Shinzawa, or Jim Jamieson, or Elliotte Friedman, or Bob McKenzie, or Ray Ferraro - there’s any number of media personalities, working in a variety of mediums, who seem to have a feel for using “advanced stats” to inform their work.

4) What is – if anything – the biggest misconception of advanced stats in hockey?

McIndoe: That they’re advanced. They’re not. For the most part, they’re really simple. Stuff like WAR and UZR and PECOTA in baseball are deeply complicated, or have secret proprietary formulas behind them, or both. That’s advanced. Corsi is just plus/minus with shot attempts instead of goals. PDO is shooting and save percentage added up. Zone starts is just a percentage. This stuff is incredibly basic, and I know that because I understand it and I’ve never been good at math.  

Hockey’s advanced stats aren’t advanced. They’re just new.

Veillette: That they’re advanced! Sure, there’s a few snake oil statistics out there that seem to be attempting to be the catch-all, WAR-like number for hockey, but the ones that are breaking ground are relatively easy. Most use low-level addition or division, not unlike, say, points or shooting percentage.

Mirtle: That they’re being blindly applied with no due diligence done to determine their utility. And that they’re used context free by people who don’t watch hockey games. I often find the opposite is true. 

Glynn: That people who like these stats don’t watch the games. It’s ridiculous to the point where it doesn’t even make me mad anymore. I go on Twitter right now, in August, and it’s filled with people talking about re-watching hockey games so they can record data from it. Maybe they enjoy a different brand of hockey, maybe you don’t see eye-to-eye with them, maybe you think a lot of them are jerks, but don’t kid yourself by saying they don’t watch or aren’t dedicated.

Drance: The idea that anyone thinks that Corsi is an “all-encompassing statistic that can be viewed in a vacuum” is the straw-man argument against advanced stats that I find most obnoxious.

5) Do you think advanced hockey stats will one day become a part of our every day hockey vocabulary? – such as goals, assists ect.

McIndoe: It will look like it does in baseball. You can watch an MLB game and not see the truly advanced stuff showing up for every batter like HRs or RBIs does, but some of the basic stuff like OPS is starting to appear, and every now and then they’ll get into an explanation of something more complex. It will be slow but sure, because it has to be, because not all fans are up on this stuff and you can’t just carpet bomb them with numbers that they don’t understand.

It’s also possible, maybe even likely, that stats like Corsi and Fenwick won’t even be part of the conversation in ten years. Remember, those numbers are just proxies for the stuff we really want to measure, but can’t right now. Once we get access to better information, we should be able to replace them with better stats.

Veillette: Absolutely. We’ve seen more television broadcasts integrate them over the past year, we’ve seen more mainstream outlets with articles involving them, and obviously teams are hiring people who know their way around them. There are people in their teens and early twenties who don’t flinch at these numbers and will carry this knowledge well into their adulthood - many of them with potential to bring them into relevant careers. It shouldn’t take long for the most popular stats to become generic hockey lingo.

Mirtle: Absolutely. I find this is already happening, especially among younger fans. As more media adopt the terminology, it’ll become more widely understood and accepted. Last season was big for this. This coming year will be even bigger. 

Glynn: Absolutely. Like anything new and exciting though we have a difficult time picturing something like “Corsi” as a norm. Some of the more complicated stuff maybe not but something as simple as Corsi, which is really just a new interpretation of “shots”, could easily become mainstream to the point where it’s in video games or on hockey cards.

Drance: I absolutely do. I still hope that we’ll call these metrics something more descriptive as we move onto more complicated fare though. I’ve long called Corsi “shot attempt differential” in my own writing, and generally refer to Fenwick as “unblocked shot attempt differential”. I remain optimistic that, over the long-term, simpler, descriptive terms will be more widely adopted for these statistics. 

6) Who are some of the best/most helpful follows on social media for those wanting to learn more on the topic?

McIndoe: That’s getting tough these days, since every keeps getting snapped up by NHL teams. But some of the guys whose work I really enjoy include Jonathan Willis, Travis Yost and Scott Cullen.

Veillette: My biggest direct and indirect influences have both been hired by NHL teams, so passing their info on wouldn’t help much, haha. Jen from Second City Hockey made an incredible database of stats-focused writers a few weeks ago that should do the trick. I’m even on it, somehow! You can find it at: http://www.secondcityhockey.com/2014/8/8/5984435/nhl-stats-focused-writers-guide 

Mirtle: This is changing as many of the best follows have been hired by teams. But I’d start with @mc79hockey, @bsh_erict, @behindthenet, @shutdownline, @robvollmanNHL, @67sound, @kyledubas, @travishehateme, @jonathanwillis. And me I guess. 

Glynn: Haha well unfortunately a few of them just got hired! There are still lots of good ones out there. @RegressedPDO made a handy list for that this summer.

Drance: I learned a lot from @jonathanwillis, @behindthenet, and @kent_wilson in my early days. I think @Thats_offside@travishehateme, @regressedPDO and @dimfilipovic are some of the best analysts around. If I was looking to get into advanced stats, those are the seven accounts I’d follow to start with.