30-16-8-4-2-1

First off, I want to apologize for being so late with a new blog entry. My daughter’s health has been more down than up in the past while, and when that takes a bad turn my writing falls away. I’m scrapping my promise of weekly posts until things improve. Thank you for your patience, and for continuing to check here. It is kind of strange, but there was an increase in traffic here when I stopped posting. I will try not to take that personally.

Since my last post, two developments caught my attention: the hiring of Kyle Dubas assistant GM in Toronto, and the interview Travis Yost conducted with Dallas Stars GM Jim Nill. As usual, I will look at each of these items from the perspective of how different types of hockey analysis can be integrated. I will probably go on for too long, because that is kind of what I do, before I get to why the article is named as it is.

 

The Toronto Factor

The Toronto Maple Leafs play in the largest hockey market in Canada. More importantly, they exist in the central media hub of our nation. As a result, every move they make is heavily scrutinized, every game is analyzed with a high intensity microscope, and every success and failure is questioned by legions of fans. When the Leafs made it public a year ago that they have no use for fancy stats, the analytics community reacted loudly and twitter was on fire on a daily basis. The team succeeded coming out of the gate, and was well positioned until the Olympic break. The effect was polarizing, as those who were pro- and anti- analytics stood toe to toe. The anti-analytics crowd were thumbing their collective noses, saying “I told you so”, while the analytics crowd moved from initially predicting a massive fall to making weak arguments based on riding shooting percentages to explain why the Leafs were winning despite losing the Corsi battle on a nightly basis.

Then, seemingly out of nowhere (or out of a very obvious somewhere, if you are in the analytics crowd), the Leafs faltered badly and started to pile up loss after loss. The extent of their fall was astonishing, as they finished behind the Senators and 9 points out of a playoff spot. Twitter was once again a battle zone, as parts of the analytics crowd were taking their turn saying “I told you so”, while the anti-analytics crowd looked for excuses.

The big problem, as I see it, is that the extreme positions on each side had it wrong. Toronto is a single team among 30, and whether or not they made the playoffs does not confirm or refute hockey analytics. Every year teams with strong corsi numbers (like New Jersey) miss the playoffs while teams that are below average make it. Stats are based on probabilities, not absolutes. The only reason people gave a shit about this one case is that the team in question was Toronto. The fact that Toronto was notoriously anti-analytics was an interesting sidebar, but it really wouldn’t have mattered much in the grand scheme of things if it was a small market team (quick question: without looking it up, is Carolina pro- or anti- analytics? The correct answer is: “who gives a fuck?”). The point I want to stress here is that single cases do not matter in stats. When you focus on a single case and use it to try to “prove” or “disprove” anything you are engaging in philosophy, not empirical science.

 

Enter Kyle Dubas

The interesting outcome of Toronto’s fall last season was the hiring of Brendon Shanahan. He examined the franchise from top to bottom, and ended up turfing two old school hockey guy (Dave Poulin and Claude Loiselle). In their place he hired Kyle Dubas as Assistant GM working under soon-to-be-unemployed Dave Nonis. Dubas is a bit of an analytics hero based on two factors. First, he collected and drew upon fancy stats while working as a minor league GM. He embraces hockey analytics, and for this reason the analytics crowd embraces him. Second, he is young. Do not underestimate how the divide between “old school” and “new age” analysis is roughly (but not entirely) generational. Specifically, younger fans are more likely to embrace fancy stats than the older generation who have looked at the same types of numbers and analysis for decades.

Steve Simmons wrote an excellent piece on this hire that is worth quoting at length:

The overly sensitive analytic community will hail the hiring of Dubas as a great victory for their world. It is — but that’s only part of what happened here. He was hired by Shanahan and agreed to by Nonis because he provides the Leafs with a bright young hockey mind, a progressive thinker, a front-office version of Roger Neilson.

In so many front offices in the NHL today, there is a disconnect between the statisticians and the old-style coaches or general managers. One side doesn’t care for or respect the other. And that works both ways.

Now, the Leafs look at it differently: They have a hockey man in whom they believe who is also a stats man. And they have a stats man who is a hockey man. And the disconnect that exists on some of the best teams in the game shouldn’t necessarily be a disconnect here.

First off, I have never enjoyed Simmons’ writing more than when I read “the overly sensitive analytics community…” That is a first class troll job, and I say that with the upmost respect and admiration. Well done!

Moving on though I think it is interesting that Simmons, who is well known for not being on board the analytics bandwagon, frames the hiring of Dubas as being similar to Roger Nielson. Nielson is rightly credited as being exceptionally innovative in his age for his use of video to analyze the game and prepare well informed game plans. The important part here is that Nielson was a hockey man through and through, and Simmons brings the connection home by stressing the Dubas is a hockey guy as well as a stats guy. The strawman that is created here is, of course, the computer nerd who looks only at numbers and does not watch actual hockey games at all. You know, the guy that absolutely no one on either side would actually advocate hiring.

Before I move on, I want to compare Simmons’ take on the hire with the one put out by James Mirtle. While Simmons sells a story about how things are business as usual because Dubas is a hockey guy, and a hockey guy who has little power under Nonis, Carlyle, etc, Mirtle paints an entirely different picture. Mirtle wrote:

For too long – years and years, really – the Leafs front office has been built on cronyism. It was – and still in many ways is, if you look at the scouting department – a group of older former pro players, two of whom were let go on Tuesday to make way for 28-year-old assistant GM Kyle Dubas.

Not only has this group been reluctant to change in the face of failure, they’ve displayed thinly veiled disdain for the analytics movement for two seasons now, publicly denouncing the kind of innovation that has helped teams like the Los Angeles Kings and Chicago Blackhawks build Stanley Cup contenders while the Leafs imploded late in the year.

[...]

Rather than analytics, however, the real story of the day was the reveal that Shanahan had seen the Leafs for what they were: An organization that had become diseased from the top and that was in need of new blood and new ideas.

To Mirtle, the central issue is that the Leafs had become an old boy’s network who resisted outside influences. This retarded their growth and prevented new innovations and ideas from taking hold within the organization. It is not so much that the emperor had no clothes as it seemed to be the same emperors running the show in the same way year after year. That would not be so bad if the team was winning, but the Leafs over the past decade and a half have been nothing short of pathetic. They have been fortunate to have some of the best and most loyal fans in the world. Those fans have not been fortunate to have had a steady dose of what the Leafs management has offered. The change had to come.

 

The 3-5% Solution

Hockey blogger and analytics guru Travis Yost put out a fascinating article based on his interview with Dallas Stars GM Jim Nill. As expected, Yost focuses on the Stars’ use of analytics, which has in his eyes led them to be the only team who has not made a misstep in the past 12 months. While Yost’s take is interesting, well argued, and thought provoking as always my focus is going to be on what Jim Nill is quoted as saying in the article. First off, Nill talked about Sort Vu coming onto the scene shortly:

There are changes every 30 or 40-seconds. The important part is getting the technology right  when we do it, we want to do it right. And I look forward to that being implemented moving forward.”

As I wrote in a earlier entry, there are all sorts of challenges to bringing this technology in to the hockey world. One of the big ones is sorting the data in ways that make sense for hockey analysis, and making it possible to retrieve the data that is most needed.

Nill also talked at length about how he and coach Lindy Ruff embrace analytics and search out ways of making themselves better using such numbers:

“Well, first, I love the job you guys do. You are part of the community — part of the hockey world. I love what you write and what you read. We are always — we are all trying to get 3-5% better. It’s a cap world and we are limited. We are always looking for the next thing. That’s the best part of the game.”

“There’s amazing stuff in the blogosphere. We sit down all the time and analyze it. Lindy and I are on the plane all the time and looking at this stuff – we look at it and track it to see if there’s something there. Like I said, we’re all very competitive, and we are all looking for the edge. And whatever’s gonna help us is great for the game.”

The interesting thing here is that Nill does not present analytics as an earth-shattering revolution. He is looking for ways to fine tune his team, and to help him make decisions that are as informed as possible:

We do use numbers — we are like everybody trying to figure how they fit in. Sometimes they fit in, sometimes they don’t.

This fine tuning involves getting 3-5% better, which is shorthand for saying not a huge amount overall, but enough to get a competitive edge over the competition. Analytics may provide that edge, but you get the sense that Nill uses numbers as a double check rather than as a guide for decision making. (a side note: I am fascinated by how Yost sees the story as analytics being used to good effect, while I see the story as a challenge of how to integrate new data. We really do see the world through individual eyes.)

 

Relational Improvement

Thirty teams compete each season in the NHL. Of those teams, 16 make the playoffs, 8 make it to the second round, 4 make it to the third round, 2 play in the Stanley Cup final, and one team wins it all. If analytics did not exist those numbers would all be the same. If we still lived in an age where the coaches refused to let players drink water during games for fear they would cramp up, those numbers would be the same. If no crowd watches, the numbers would be the same.

Establishing and maintaining a winning franchise in the NHL is all about gaining small competitive edges on the competition. The most common edge, and the one that is the most easily recognizable, is drafting and trading for the best players. Beyond that is a world where small coaching decisions, roster moves, budgetary decisions, training schedules, and other factors all provide tiny advantages over the teams with which you are competing. As an NHL GM, Jim Nill operates in this world, and it is in this context that his comment about looking to fancy stats for a 3-5% improvement should be understood. Mirtle describes the stagnation in the Leafs’ front office, which allowed other teams to pass them by. When the only measure of achievement is in relation to other teams, standing still is losing strategy.

 

I am going to leave the argument here for the moment. I promised to re-work this all into the realm of integrating hockey analysis, but I don’t think I can get there without making this post a mile long (I’m over 2000 words already, which is my usual marker to wind it up). So, as they say in television, to be continued…

 

Posted in Analytics, Jim Nill, Kyle Dubas | Tagged , , , | Leave a comment