Colorado Avalanche Analytics: Stats Useful, but Not Advanced

Mar 3, 2016; Denver, CO, USA; Colorado Avalanche head coach Patrick Roy talks to his players during a timeout out called in the third period against the Florida Panthers at the Pepsi Center. The Avalanche defeated the Panthers 3-2. Mandatory Credit: Ron Chenoy-USA TODAY Sports
Mar 3, 2016; Denver, CO, USA; Colorado Avalanche head coach Patrick Roy talks to his players during a timeout out called in the third period against the Florida Panthers at the Pepsi Center. The Avalanche defeated the Panthers 3-2. Mandatory Credit: Ron Chenoy-USA TODAY Sports /

Colorado Avalanahce fans, call them analytics, call them advanced stats, even call them fancy stats, but please whatever you do, don’t call them magic.

There has been a discussion on this site about the value of so-called “advanced” stats, especially in relation to the Colorado Avalanche. My editor, Nadia Archuleta, considers them only partially useful. My colleague, Mark Kinz, (we’ll call him The Other Mark) is much stronger in his support of them. Now, despite my using them in nearly every article I write, I actually fall somewhere in the middle.

My educational background has involved a lot of training in the scientific method, logic, and critical thinking. What has been hammered into me is that if you want to make a strong point, you need to back it up with evidence.

More from Avalanche News

So if I want to claim that Erik Johnson is an elite defenseman, it’s a much stronger claim if I can back it up with some evidence. That’s all stats are to me – evidence I can use to back up my points.

Most people have no problem using goals scored or even plus-minus as evidence, but when it comes to “analytics” there seems to be a big divide among hockey fans.

Ever since the bestselling book and film Moneyball, everyone involved in hockey has wondered if the same thing can be done for this sport. If you talk to a big proponent of things like Corsi, it will seem like advanced stats are hockey’s Moneyball, but they’d be wrong.

Bill Bean led the advanced stats revolution in baseball
Oct 24, 2014; San Francisco, CA, USA; MLB ambassador for inclusion Billy Bean before game three of the 2014 World Series between the San Francisco Giants and the Kansas City Royals at AT&T Park. Mandatory Credit: Kyle Terada-USA TODAY Sports /

The Moneyball stats in baseball were a complete game-changer. Simply by adjusting the way they used stats to evaluate players, Bill Bean’s Oakland A’s were able to win more games than traditional thinkers thought would be possible.

A revolution of that kind has not happened in hockey since the introduction of things like Corsi, and it’s not likely to happen.

This is because baseball and hockey are fundamentally different sports. Baseball can quite easily be simplified to a one-on-one competition between a batter and a pitcher.

Hockey, on the other hand, is an ever-changing, fast-paced jumble of ten players constantly interacting with each other – it’s much more complex.

An advanced stat would be something that can simplify that ten-player interaction into a few simple numbers – hockey analytics are nowhere near that point. In fact, they’re actually quite simple.

The most well-known of these stats is Corsi, which is often simply called possession. You’ll hear someone say “so-and-so is a great possession player”. But Corsi doesn’t actually measure possession. All Corsi measures is shot attempts for and against. That’s it.

The idea behind Corsi is that if you’re making shot attempts you have the puck, and if you’re having the puck shot against you, you don’t have the puck – thus, possession. It’s also been found that the best teams *surprise, surprise* have the puck more often.

The other so-called advanced stats generally aren’t much more complex than that. All they’re measuring is the kind of thing that we can see when we watch the game anyway.

So to me, there’s no such thing as advanced stats in hockey – there’s just stats. And I for one think they’re very useful, but not a complete replacement for actually watching the games.

I find if you actually watch the games you’re going to end up with a pretty accurate view of how good each player on your team is. Where these stats come in handy is in two situations: when you’re particularly biased and when you’re evaluating players on other teams.

When you’re biased against a player, you’re more likely to notice when they make a mistake than when they make a good play, and the opposite is true when you’re biased in favor of a player. Stats can help even out that bias.

For example, I found many Avalanche fans were biased unfairly against Nick Holden, but the stats showed that he wasn’t actually all that bad.

The other area stats are extremely useful is evaluating players on other teams. Without stats, your opinions of these players will likely be based on highlights, one or two games you’ve watched (in which they may have had an uncharacteristically good or bad night) and the opinions of pundits and other fans. Stats can help even this out.

What I like to do is look at the stats for the Avalanche players, and see how well they line up with my own opinion after watching all the Avalanche games.

In this way I can figure out which stats (or combination of stats) most accurately reflect how I evaluate players. I can then look at those same stats for players on other teams and get a good idea of how I’d likely evaluate them if I had watched all their games.

Even so, there’s still no real replacement for the combination of actually watching all a team’s games and looking at the stats for that team. In my opinion that’s the best way to go about things.

More from Mile High Sticking

And for this, the new stats that have come out under the title of analytics are extremely useful. Before things like Corsi or zone start percentages, there was no way to effectively evaluate defensemen without watching them play.

Points totals and plus-minus tell you almost nothing about how well a player actually defends. How many shots the other team gets while a defender is on the ice vs. how many shots they get when he’s off the ice – that’s a useful number to have.

But that doesn’t mean these stats are some kind of miracle cure. I have a friend on Facebook I often argue with because he tends to act as if things like Corsi are the be-all, end-all of evaluating hockey. When the Avs were winning the Central division, he was constantly harping on how flukey it was that they kept winning and how they were destined to fail.

To him, the loss to the Wild vindicated the predictive powers of “advanced stats”. To me, the loss to the Wild was due to injures and running into an incredibly hot goaltender. The difference is I based my opinion more on watching the game than on the stats.

The problem I continue to find with advanced stats advocates is they tend to put the cart before the horse. They look at a team like the Pittsburgh Penguins, see that they are a good possession team, and say that’s why they won the Stanley Cup.

I look at the Penguins and say they’re a good possession team because they have players like Sidney Crosby, Evgeny Malkin and Kris Letang, and it’s because they have players like that and good goaltending that they won the cup.

To me, possession is a by-product of having good players playing in a system that suits them. To advanced stats advocates, winning is a by-product of good possession numbers.

But that’s not reality. Advanced stats aren’t fancy. They’re not magical. They’re not revolutionary. They’re just new. Once people understand these stats, they cease to be advanced.

Had plus-minus been introduced now instead of in the 1950’s, it would be called an advanced stat and traditionalists would view it with suspicion.

colorado avalanche
Feb 7, 2015; Saint Paul, MN, USA; Colorado head coach Patrick Roy in the first period against the Minnesota Wild at Xcel Energy Center. The Minnesota Wild beat the Colorado Avalanche 1-0. Mandatory Credit: Brad Rempel-USA TODAY Sports /

People get on coach Roy about he reluctance to use analytics, but I seriously doubt he wasn’t already using them in some form. He’s not an idiot.

The thing is, NHL teams have access to much more useful and accurate stats than we do as fans. They can get data on actual puck possession, not some proxy for it. They also have film of every NHL game they can watch repeatedly and break down in detail.

They don’t need to rely on stats to tell them why a player is good or why a team won a game – they can see it for themselves.

But we as fans do need stats for that. Most of us don’t have time to watch every game for every team, so if we want our opinions to be backed up by evidence, we have to understand stats, and that includes the new ones.

Next: Hockey IS Slipknot

So, if there’s a demand for it, I’ll write an article or two explaining as clearly as I can what these new stats are all about and how they can be used effectively to evaluate players. Until then, please don’t be scared of these stats but also please don’t act like they’re some miracle cure that can make a bad team suddenly good.

As Nadia said, “Hockey is NOTbaseball”.