The advance of hockey stats: Analytics continue to grow, evolve in NHL
About halfway through last season, the Los Angeles Kings started posting statistics internally. Not traditional stats like goals, assists, points and plus-minus ratings. So-called “advanced” stats. Coach Darryl Sutter and his assistants – John Stevens and Davis Payne – explained them to the players.
“They’re like, ‘This is what we look at after a game,’ ” said winger Justin Williams. “ ‘It’s not the be-all, tell-all, but it says where we are.’ ”
Not everyone absorbed the specifics. Even Williams, a darling of the analytics community for his outstanding possession numbers, said he couldn’t remember exactly what the stats were. (He thought they were Corsi and Fenwick but wasn’t sure.) The coaches soon stopped posting the stats, and the talk faded.
But the larger point was this: The Kings value puck possession. They use analytics as one of their tools to evaluate players and design strategies, whether the players understand that or not. They judge the process with a long-term view; they don’t just judge short-term results.
“We lost the game, and they looked at the stats,” Williams said. “And they’re like, ‘You know what? We probably deserved better in that game. We keep playing like this, we’re going to get our wins.’ ”
A few months later, the Kings won the Stanley Cup. They became the second team to win it twice in the salary-cap era – after the Chicago Blackhawks, who play a different style but also value puck possession and use their own brand of analytics.
It was no coincidence, and it’s no coincidence that more and more NHL teams have begun to embrace analytics, too, some hiring bright, data-driven bloggers over the summer to bring fresh perspectives to their front offices.
The most visible example? The Toronto Maple Leafs, who went from dismissing analytics to creating an analytics department under new president Brendan Shanahan. As Maple Leaf Sports and Entertainment president Tim Leiweke put it in a talk at Ryerson University last week, Shanahan decided the Leafs would be “smart from now on.”
“What sport doesn’t that happen in, right?” said Kings general manager Dean Lombardi with an exasperated laugh. “You know it’s coming. You now have the target on your back. But it’s your job to keep pushing in every area to try to make it better.”
The question is not whether analytics can help. Clearly they can. If analytics have helped teams win Cups, we should be beyond the simplistic, dualistic debates – good or bad, right or wrong, skill or character, data or eyes, geeks or scouts, bloggers or mainstream media.
But questions remain: Exactly how have successful teams used analytics? How have they weighed new information against the old? (Answering that is easier said than done when those teams wisely try to protect their secrets.)
How will analytics evolve? How much influence will the new guard have in organizations like the Leafs, who still have some of the old guard in place? Can the copycats catch up, and if so, how quickly?
“Sure, everybody else is doing it now, but there’s a lot to learn there,” Lombardi said. “You’ve got to be careful. There’s a difference between data and knowledge. … It’s not only what you use, but how you use it.”
* * * * *
Let’s start with data. Analytics experts say advanced stats in hockey aren’t that advanced. Generally we’re talking about percentages of shot attempts (Corsi) and unblocked shot attempts (Fenwick), combining shooting and save percentages (PDO), where players are deployed (zone starts) and against whom (quality of competition).
We want to know which teams and players have the puck most often, because over time, the teams and players who have the puck most often have better chances of success. We want to know if abnormal shooting and save percentages are skewing the results, because if they are, the odds are they will regress to the mean. We want to keep it in context.
But remember: Teams not only value those numbers differently, they come up with their own numbers. They track everything from simple scoring chances to sophisticated data sets on zone entries. And they aren’t about to share their methodology or results for obvious reasons.
Blackhawks GM Stan Bowman is the son of legendary coach Scotty Bowman, but before he broke into the NHL, he was a business consultant. He worked in “process reengineering,” helping struggling companies become more efficient, introducing new software packages, redesigning the way they did business. He brought those skills to hockey.
Bowman will tell you the Blackhawks don’t use Corsi and Fenwick. He will tell you they use something similar but different. That’s about all he will tell you. “I think it’s better, but I guess it’s a matter of opinion,” said Bowman to the Chicago Sun-Times. “It’s also a competitive advantage.”
Once you decide what data to use, you must figure out how to turn the data into knowledge. What is meaningful? What isn’t? What applies to your particular team?
Even among the analytics champions, there is more than one way to build. The Blackhawks are known for speed and skill; the Kings are known for size and relentlessness. Both are great possession teams.
Lombardi has identified possession players, but he also has identified character players – which might explain why this team trusted the process through three come-from-behind, seven-game playoff series and went on to win the Cup.
Not all of Lombardi’s moves are driven by analytics. He re-signed veteran defenseman Matt Greene to a four-year, $10 million deal. He declined to buy out veteran forward Mike Richards despite the six years left on his contract at a $5.75 million cap hit. The numbers say both should have been let go.
“We will never lose sight of the fact that emotion and character is central to what we believe in here,” Lombardi said. “I think emotion in hockey – probably more than any other sport when it rises to a certain level – can close a bigger gap in terms of skill than any other sport, and I even put football in there. That’s why I think it’s the greatest game, because so much depends upon the human will.”
When veteran GM Jim Rutherford took over the Pittsburgh Penguins this summer, he put an emphasis on analytics. He said the team has more detailed data than the public does. He thinks they are a great way to look at matchups – who plays well with whom, who plays well against whom. He wants them to be part of a system of checks and balances.
“You and I could watch and think, ‘Those guys look pretty good,’ ” Rutherford said. “But when you dig deep into it, there’s a missing piece there. That’s when you really start the conversation. Why are the analytics telling us this about this player, and why are we seeing this? …
“If you’ve got a big discrepancy between what the analytics are saying about a player and what our staff’s saying, then we need to have a big meeting. How do I get the information brought into the middle here so I can make a better decision? How come you’re way over here, and how come it’s way over here? …
“At the end of the day, I still have to make a gut decision. I’m not going to solely go on what the analytics say, but it is a very important tool.”
* * * * *
For the most part, when analytics have been used in the NHL, they have stayed on the management and coaching levels. They have not filtered down to the players.
The executives build their rosters. The coaches come up with the X’s and O’s, deployments and matchups. The players aren’t told to, say, “possess the puck more often.” They are given specific instructions to, say, “carry the puck into the offensive zone instead of dumping it in.”
Most players say they know little to nothing about analytics and pay little to no attention to them. “I’m not going to look into it,” Williams said. “People have told me that, yes, I am a good possession player, but I’m not going to try and do more or less. I’m just going to try to do what I do and hopefully that resonates into success.”
Even the Blackhawks’ Patrick Kane, one of the most knowledgeable players in the NHL, doesn’t follow Corsi. “To be honest with you,” he said, “I didn’t even really know what it was until later in the year [last season].”
But there are exceptions, like when the Kings’ coaches posted those stats last season, and it will be interesting to see if a new trend develops. Edmonton Oilers winger Taylor Hall said he had casual conversations about Corsi with coach Dallas Eakins last season. When the numbers didn’t look good, they tried to find the root of the problem.
“ ‘OK, our Corsi’s bad. Why is that? We’re not breaking the puck out properly,’ ” Hall said. “Applying advanced stats to in-game situations and in-game strategies, I think that’s where that next step’s going to be taken that hopefully can help out players.”
The Oilers hired blogger Tyler Dellow over the summer, even though Dellow had been one of their harshest critics at times. Hall said he read some of Dellow’s work before the Oilers took down the blog to keep the ideas in-house.
“I mean, it’s not like Tyler’s going to be on the ice scoring goals,” said Hall with a smile. “But I think you can see that the Oilers are doing whatever they can, and they’re trying to get ahead of the curve. Whether they are or not, I’m not sure, but it certainly doesn’t hurt to try.”
It only hurts not to try.
“I don’t consider myself a hockey nerd, but I enjoy the game and enjoy seeing new ways that the game is improving,” Hall said. “If you’re someone that doesn’t see legitimacy in advanced stats, then you …”
Hall didn’t finish the sentence.
“It’s just probability,” he continued. “It’s just having the odds in your favor. And all it really comes down to is puck possession. Everyone knows that’s going to win you a lot more games than anything else.”
