Terry Appleby is a retired Canadian board-game developer, an economist, and 20-year member of baseball’s ultimate nerd group, the Society of American Baseball Research (SABR). Since a child he has been fascinated with the statistics of sports, and a lifelong basement researcher, crunching numbers and thinking about the games we watch in unique and wickedly smart ways.
When hockey writer Mark Spector, now at Sportsnet, filed a lighthearted newspaper column in October 1998 on how the Oilers’ new lineup should yield eight more victories come season’s end, little did he know that he would trigger Terry’s beautiful mind.
“If the purpose of an NHL hockey player is to help the team win, why aren’t we measuring him that way?” Terry asked himself. It was as if a Zamboni had cleaned an ice rink rutted with traditional statistics. Terry began to make fresh strides in the way he gathered and compressed goals and assists, creating new formulas and developing composite “super stats” focused on finding winners.
His fantastic findings, however, were known only to Terry and his hard drive. And his son.
For Christmas 2008, Terry’s son, Marc Appleby — an entrepreneur who impressed Canadian audiences and high-rolling investors by successfully pitching EcoTraction on Dragon’s Den in 2009 — bought his father a website to showcase all of his number-crunching. Once the father and son worked out the bugs, they brought the PowerScout system here to Sportsnet. After 13 years of scribbled notes and a gazillion hours of one passionate hobbyist’s data entry, a revolutionary way of evaluating hockey players and teams (moneypuck?) is now upon us.
“I’m PowerScout’s biggest superfan because I watched my dad develop it year after year, and I saw the results. I thought it was incredible: you’re really telling the story behind what people think they’re seeing. And it reinforces a lot of the clichés of hockey,” Marc explains. “They talk about defence wins championships and you want strength down the middle. These things come out in the research. So now we’ve built a system that has done all the research, and we just track the numbers for people and let the numbers speak for themselves.”
We caught up with Marc to better understand the NHL’s new math and what it means to fans.
How is PowerScout’s approach to hockey comparable to Moneyball’s approach to baseball?
The statistics that come out of it and the research that we’ve done is based on one fundamental principle, and that’s winning. And Moneyball had the same principle, basically looking at, OK, if you do this in these situations, you’ll be more successful. That our research is based on winning, it helps teams make better decisions, (and) it helps fans understand what’s going on with their team or their players. It gives a third category for people to follow. There’s the scoring race, based on goals and assists; there’s fantasy, which is everyone’s pools where power-play goals and penalty minutes are worth something; and we’re more reality-based. It’s an honest look at what’s on the ice, why teams are winning and losing. It’s founded on research that says, OK, if you do these things, you have a higher chance of winning games, winning championships.
What first inspired this research?
When Mark Spector’s piece in 1998 came out and said, hey, everybody’s looking at goals here, but everyone cares about winning. They don’t care about goals; they want to win, they don’t care how it happens. So he says, “Why aren’t we measuring players this way?” And that was three years before (Michael Lewis) came out with his baseball (book) Moneyball. His research predated (the publishing of Lewis’s book in 2003), but of course, (my father has) been in his basement for 13 years. My dad’s been part of the Society of American Baseball Research for 20 years. He’s a member of SABR, which is the baseball nerd group, of which Bill James is a member, and my dad sits on the Business of Baseball committee and the analysis committees. He’s a massive baseball fan, so he knew a lot of the research that was being done in Moneyball. He was more inspired by Mark Spector and his article in the Edmonton Journal taking a look at the Oilers in the upcoming season and saying, I think they’re going to be eight wins better, based on all the players coming in or going out. That was the light bulb. (Terry) said, “Wait a minute. This is the way we need to evaluate players.” So he went back and said, How do we do this statistically? And once the real-time statistics came out and you could get ice time per player, that allowed him to say, OK, for the amount of time this guy was on the ice, this is how much he contributed.
PowerScout’s current MVP rating has defencemen in the top three spots, led by Brian Campbell. In your opinion, then, should Campbell be voted MVP of the league?
Well, I think you have to be on a playoff team to be MVP. How the MVP rating is calculated is: How much better is he, based on his ice time, than if he were to be replaced by someone on the minors? Brian Campbell has a certain performance, but if I replace him with some no-name, that guy will also make some contribution. So it’s the percentage better he his than a guy who could replace him. Right now, everyone’s wondering why Florida’s so successful, and part of it is Campbell’s stability on the back end, creating offence. For Florida, he is their MVP.
A lot of offensive defencemen will lead their team in MVP because they’re contributing to wins. It’s all about contribution to the team, and defencemen play a lot more time than wingers. A winger might play 15 minutes; a defenceman’s on the ice for 20-25 minutes. He’s on the ice for half the game. Plus, each forward plays a third of the ice (surface); a defenceman plays half the ice, and they’re the last line of defence. So defencemen, and particularly offensive defencemen, are rated higher.
Which NHL statistic do fans and media overvalue?
The most overrated statistic is penalty minutes. And for most people in fantasy leagues, penalty minutes are great. You want your guy to get fights and take penalties. In reality, who wants to take penalties? This is a negative. You look at Campbell, he’s disciplined. Where if you look at (Dustin) Byfuglien, some people thought he should’ve been MVP last year, but he was taking almost a penalty-and-a-half per game.
In our research, we’ve found that penalty killing is almost four times more important than power play. That was a massive finding. You’re not going to lose a game if you have a bad power play, but you will if you have a bad penalty kill. Why are the Leafs struggling? It’s their penalty kill.
And in the playoffs, we found that one of the most important stats is goaltending performance against only the top 10 teams in the league, so we look at that and penalty killing: the two major drivers of playoff success.
Power-play goals and penalty minutes are dismissed. Everyone likes goals, so they talk about how the power play is doing. Penalty kill is not as sexy, but it’s almost four times more important.
There’s a misconception that toughness equals penalties, but in reality it’s about discipline. Toughness comes from hits. When we look at players, we look at 12 skills. It’s very much like the quarterback rating in football. Each quarterback is rated between 75 and 150 based on all these stats, and no one really knows how it’s calculated, although it’s on Wikipedia, but based on all these metrics, it gives you a sense of performance. We know 150 is an incredible game, and 100 is about average.
Is the eventual goal to have stats like your MVP rating appear in a player’s stat line when they’re on TV?
Absolutely. If there’s one stat that I think can revolutionize hockey, that would be Game Dominance. It’s the ultimate context to NHL scores. We’re an analytical company, so there’s a niche market there. But every fan needs to get their score; everyone wants to know, hey, what happened in the game last night. PowerScout allows you to see that score but then also see which team outplayed the other in percentage terms. It’s not a new stat that people need to learn; it’s percentages. You can see that a team won 2-1 but was outplayed 70% to 30%. You get the context for the game. That’s what I want to get to for out-of-town scoreboards: “Oh, 2-1 but they’re getting outplayed.” That little bit of extra context that allows media guys to present a better story at a glance. That’s one statistic I think could become part of the box score.
Why should Sportsnet.ca users care about these Game Tracker and Game Dominance graphs?
It’s the single best way to scoreboard-watch, to get the feel of the game at a glance. If you’re a Leafs fan and you’re at a restaurant or in a movie theatre, pull up your phone and you can watch the game in real time and see how it’s playing out. It’s the ultimate platform for fans to get and the score and the context in one second. People want scores. Where do they go for scores? A website. And it’s interesting to go back in the archives. For the rest of hockey history, every game now will be archived in one little picture, never to be forgotten. I find that really powerful.
Explain the Momentum Meter in simple terms.
The momentum meter tracks how fast shots happen in real time to measure team effort. We’re trying to create a visual fingerprint-and I like the word “fingerprint” because every game is unique of every game’s momentum swings for each team. Every 30 seconds, we look at how many shots that team had. If they had a lot of shots, their momentum goes up; if not, it goes down. What you’re able to do is capture how hard teams are trying. Shots are an estimator; it’s the most basic stat you could use to drive something like this, which everybody understands. If you take a shot, you’re in the opposition’s zone. It’s a proxy for zone-time. And it shows how teams react to power plays and fights and goals.
One of our readers wrote that using shots on goal can be misleading. One team could be taking a weak shot every 30 seconds and not generating quality scoring opportunities, whereas the other team could be taking a shot every five minutes but giving itself great chances to score. How do you respond to that?
Absolutely. We’re always looking to refine it. Certainly quality of shots is important. But how do you rate the quality of every scoring chance unless you’re watching every game? Or unless you’re getting location data-where the shot was taken-which the league does not allow us to get. NHL.com is the only site that can access where the shots are taken from and if it’s a wrist shot or a slap shot; they retain that for their Game Center Live. You could get scoring chance data, but shots are the fundamental building block. If we went to scoring chances, you’d be missing part of (the picture). There’s the odd shot from the other end of the ice, sure; usually when you get a shot, you’re in the other team’s zone. You get a sense of who’s attacking and who’s playing defensive by looking at shots. And if you watch a game while you have the graphs going on, it almost mimics the sense of the crowd. It tells you that much about the game; that’s why I find them so powerful. If you missed any part of any game, you go back to that graph, and in three seconds you know exactly how it played out as if you were there.
What’s the biggest flaw of the PowerScout system?
I haven’t heard negative feedback yet. Some people think we shouldn’t use giveaways and takeaways. We’ve been asked why we’re not using plus-minus or game-winning goals or power-play goals.
Why aren’t you using plus-minus?
We’re trying to look at player skills. When looking at an individual player’s statistics, plus-minus is a team statistic that’s applied to a player. We know that the trouble with plus-minus is that he could be just coming onto the ice or just going of the ice or not involved in the (goal-scoring) play. Players on really bad teams have low plus-minus because they lose a lot, not because they’re a bad player. It’s not an individual skill. Much like power-play goals. Goal scoring is the skill. Power-play (and game-winning and shorthanded) goals are just when they happen. But a goal is a goal is a goal in our system. We don’t care if it’s power-play or shorthanded. It’s a skills-based approach. Takeaways and giveaways are a small part of the model, and we’re the first people to incorporate all the individual statistics the NHL uses into one single stat, which is like the quarterback rating.
Why don’t you factor in faceoffs?
Everyone says, “Well, for centres, how come you’re not using faceoff winning percentage?” There’s some faceoffs that are important, but for the most part, they’re not. Blue lines? Centre ice? The best faceoff guy is 60% or 55%; the worst guy is 40%. That’s a difference of one faceoff for every 10. Some faceoffs are important, but say you win the centre-ice faceoff. What do you do? You go and dump it in the zone and give the other team the puck right back. There are so many moments after a faceoff when possession changes hands, it’s obscene to think that faceoffs are a measure of possession. But we measure it because it’s the only set piece that happens in hockey. It’s easy to measure. But in terms of its impact on winning, we feel it’s overvalued.
Why bring this system to Sportsnet? Why not bring it to a small-market team and try to build a winning team for cheap?
From a business perspective, it’s costly to go after fans individually. Trying to do Google ads and trying to attract fans one by one, it’s time-consuming. I knew I had awesome content that could help a Sportsnet take their stuff to the next level. On a whim, I said, I want to talk to the networks. Ultimately we want these graphs on TV. What better way to do out-of-town scoreboard? This is cutting-edge, revolutionary stuff, totally designed for TV. We talked to other networks as well. Sportsnet felt like the best fit because they’re the leader in NHL broadcasts in Canada; they show the most games. They’re regionally based, and they’re in a position for growth. I felt good rapport with the Sportsnet team.
I’m an Oilers fan, I’m from Edmonton originally, live in Ottawa now, and I get a lot of Sportsnet games. I liked where the branding was going. We are still going to the teams, and there’s a challenge around that because we have out information on the web and possibly on TV. The teams might say, “Why do I want you guys? You’re giving away all your stuff for free?” Ultimately, there’s so much that can be done. A small portion sits on Sportsnet, stuff for the fans to understand. If we look at the Trade Maker, Sportsnet will have five variables to post; the teams will get 12. There’s enough content there for everybody.
I’m trying to give a legacy for my dad. He’s a superfan who’s done so much for so long. A brilliant economist and sports fan, and he sees the game in a way that needs to be shared. I want to help you guys be the leading sports broadcast starting with trade deadline day. It was kind of a no-brainer.
For more, please check out PowerScoutHockey.com.