For a better browsing experience, please upgrade your browser.

## Sports

With 17:25 left to go in the first period of a December National Hockey League game between the Philadelphia Flyers and the Edmonton Oilers, the Oilers gathered the puck in their defensive end and passed it ahead to Jordan Eberle, who was headed up the left side of the ice. The Flyers’ defensemen — Kimmo Timonen and Braydon Coburn — were well positioned, but Eberle decided to challenge them. He cut diagonally across the ice towards Timonen, drove wide to the far boards, put on a burst of speed and beat Timonen to enter the offensive zone.

Somewhere, Jessica Schmidt was watching. She has spent the last two seasons tracking each entry into the offensive zone with a spreadsheet open in front of her. A 26-year-old diehard hockey fan, she had read some articles I wrote about the Flyers’ zone entries in the 2011-12 season and the usefulness of zone entries in assessing a team’s performance. When the Flyers missed the playoffs in 2013, she wanted to know what had gone wrong and volunteered to try recording the zone entries herself.

That information doesn’t come easily, however. Schmidt estimates that tracking a game takes her about 90 minutes, which means that it would take an incredible amount of dedication and effort from several people to collect a year’s worth of data for a handful of teams. It would take a whole platoon of volunteers to track zone entries for every NHL game, and even then they would be capturing only specific pieces of select key moments. An NBA analyst wouldn’t need someone like Schmidt to put in hundreds of hours tracking zone entries; that sort of information — and much more — is easily gleaned from the NBA’s automated video tracking system, SportVu. But hockey lacks the position-tracking systems that many other sports use, even though there are hugely important lessons their data can teach.

Instead, it has people like Schmidt. As Eberle drove around Timonen into the offensive zone, Schmidt made a note in her spreadsheet: “1 17:22 C Opp 44.” Translation: in the first period, with 17:22 left, there was a carry-in (C) by the Flyers’ opponent defended by the Flyers’ No. 44, Timonen.

That last part — who defended the play — was a new wrinkle Schmidt and I added this year. Previous analysis had given us important insight into the performance of the puck-carrier, but assessment of the other nine skaters on the ice relied largely on inference. This year, I asked Schmidt to include some off-puck information — things like which player retrieved the puck when the Flyers dumped it in, or who had primary defensive responsibility on the opponent’s entry.

Tracking by Schmidt and others has helped explain that a team’s entry into the offensive zone has a big impact on its shot differential. Carrying the puck into the offensive zone leads to more than twice as many shots and goals as a dump-and-chase play does, even after removing plays like odd-man rushes and dump-ins that are made just to buy time for a line change. These results have even made an impact on strategy.

Schmidt’s data from this season allows us to evaluate individual defensive contributions for the first time, which we couldn’t do with traditional box scores and play-tracking. It’s clear, for example, that the Flyers have generally controlled play when Timonen and Coburn were both on the ice, but it’s hard to tell how much each player contributed on the defensive end.1 We can try to infer individual ability by comparing how two players perform when one is on the ice and the other is not. But the data is complex, confounded by differences in their deployment, who else was playing during those minutes and simple variance.

But with Schmidt’s data we can track what happens off the puck, and gain enough granularity to examine defenders’ performance as opponents are advancing into the Flyers’ zone. Timonen and Coburn are the Flyers’ top defensive pair by any measure — they play the most minutes, face the toughest competition and have the best Corsi ratings (a plus/minus stat based on shot attempts) among the team’s defensemen. Timonen’s Corsi is actually higher than Coburn’s, as is his plus/minus, the metric most used historically to assess defensive success. Timonen is paid a lot more, too, carrying a $6 million cap hit as compared to Coburn’s$4.5 million. But while Timonen is undoubtedly a more skilled puck-mover, Schmidt’s data provides the first strong statistical evidence that Coburn may be making the bigger defensive contribution.

Opposing teams seem to be attacking Timonen’s side of the ice somewhat more often than Coburn’s. Through 70 games, it was 25 percent more common for Timonen’s man to attempt a zone entry when both Timonen and Coburn were on the ice together. There’s good reason for teams to stay away from Coburn: Opponents are 16 percent less likely to successfully carry the puck into the offensive zone when attacking his side of the ice — and more than twice as likely to see the play broken up altogether.

The two players saw about the same number of dump-and-chase plays (25 percent for Coburn, 22 percent for Timonen), but Coburn managed to completely break plays up more than twice as often as Timonen did (14 percent versus 6 percent of the time). Of the Flyers’ seven defensemen to see regular playing time, Coburn has been targeted least frequently, has allowed opponents the fewest carry-ins and has forced the most failed entry attempts.2

Coburn has clearly been the player helping the Flyers obtain possession at the defensive end. Without Schmidt, we wouldn’t know that with any accuracy because Coburn’s results have come from a collection of small plays, unrecorded in the box score and overlooked by most observers.

The magic of analytics is in recording all of the small things lost to memory that add up to something significant. As this sort of micro-tracking spreads and eventually becomes automated, we’ll get better and better at capturing and valuing the cumulative impact of these small plays. Once teams can easily access a season’s worth of data — along with zone exits, passing maps, shooting tendencies, defensive positioning and the rest of it — there will inevitably be significant changes in both evaluation and strategy.

To do that, though, the NHL needs some type of automated tracking system. Some companies are trying to build one, and teams recognize what video analysis can offer. People in NHL front offices are enthusiastic about such systems’ potential and anticipate video tracking making a big impact in a few years.

“The biggest limitation right now, based on the people I have talked to, seems to be the cost and labor involved with mining the data,” said Josh Flynn, assistant to the general manager of the Columbus Blue Jackets.

Michael Peterson, a statistical analyst for the Tampa Bay Lightning, says that in particular, “it’s difficult for them to pick up the puck on the video, so there is a lot of manual processing time that goes into tracking the puck.” Unlike a basketball, a hockey puck is small and rapidly darts between sticks and skates, making it hard to keep it on camera.

Peterson suggested that fully automated puck-tracking might require embedding an RFID chip in the puck. He further noted that tracking stick placement can be quite important to assessing defensive positioning and might also require RFID tagging.

Marc Appleby, the CEO of PowerScout Hockey, one of the companies meeting with teams about advanced tracking, said his company already provides data on all player and puck movements, which he feels makes it “easy to break down defensive positioning.” He added that while stick locations aren’t captured currently, it’s something that could eventually be done via video alone.

One way or another, it seems all but inevitable that the requisite data is coming. Today, manual tracking of zone entries records an important component of how individual players drive their teams’ success. But there’s more to be had. When automated video tracking comes to the NHL, Schmidt and others like her will be eclipsed by systems vastly more complex but with data that’s far more public. And if they aren’t? Schmidt said, “I’d most likely continue to track entries the same way I always have.”

## Footnotes

1. The NHL’s play-by-play data attempts to measure defensive effort by recording events such as hits, blocked shots and takeaways. However, in practice, these plays show no correlation with success, so hockey analysts have had to turn to other methods. ^
2. When dealing with a new stat, it’s important to think about whether, and to what extent, the observed differences can be attributed to skill rather than to simple random chance. One way of assessing that is by comparing players’ performances in odd-numbered games to even-numbered games; if the stat is measuring a repeatable talent, then the two should be highly correlated. In this case, for the six Flyers defensemen who have played at least half the season, the correlations are quite strong — 0.89 for carry-ins against and 0.73 for entries denied. Six players isn’t a large enough sample to draw firm conclusions, but we can be reasonably confident that our stats represent meaningful differences among players. ^

Eric Tulsky is a hockey analyst who has consulted with the Nashville Predators and other teams. More of his work can be found at his blog, Outnumbered.

All Sports

### A Ben Carson Surge May Test TrumpAug 31, 2015

Never miss the best of FiveThirtyEight.

Subscribe to the FiveThirtyEight Newsletter
×

Sign up for our newsletters to keep up with our favorite articles, charts and regressions. We have three on offer: a curated digest of the best of FiveThirtyEight from the past week; The Week In Data, our weekly look at the best data journalism from around the web; and Significant Digits, our roundup of numbers in the news. Enter your email below, and we’ll be in touch.

By clicking subscribe, you agree to the FanBridge Privacy Policy

Powered by WordPress.com VIP