Toronto Maple Leafs: Why Analytics Matter In the NHL

TORONTO, ON - FEBRUARY 23: William Nylander #29 of the Toronto Maple Leafs celebrates his goal against the Montreal Canadiens in an NHL game at Scotiabank Arena on February 23, 2019 in Toronto, Ontario, Canada. (Photo by Claus Andersen/Getty Images)
TORONTO, ON - FEBRUARY 23: William Nylander #29 of the Toronto Maple Leafs celebrates his goal against the Montreal Canadiens in an NHL game at Scotiabank Arena on February 23, 2019 in Toronto, Ontario, Canada. (Photo by Claus Andersen/Getty Images)

The Toronto Maple Leafs have famously given the keys of their franchise to then 31 year old Kyle Dubas.

It was a bold move, but the Toronto Maple Leafs front office felt that Dubas had a unique blend of “hockey guy” background and analytics background that overcame the fears about hiring someone so young.

One thing I think we ought to keep in mind about Dubas is that he must be one hell of a bright guy to earn the top job of one of the biggest sporting organizations in the world.  To get all the necessary parties in a billion dollar company to sign off on what – from an outside perspective – is guaranteed to look like a very high-risk move, the guy must be impressive.

Now that doesn’t mean that he gets a free pass, or that you shouldn’t criticize him, because he shouldn’t and you should. But it does means that if he has some ideas that go against traditional thinking, he has at least earned the right for  a window of time to implement them, and try to change the mind of people who might write those ideas off at the outset.

And if you watch a few of his youtube videos where he speaks at analytics conferences, I think you will be impressed.

But so many people have already written him off.  I know this because my job is to write these articles and then promote them through engagement with the readers. I work hard to speak with people of all sorts of ideas and not restrict myself to a group of like-minded people who are predisposed to like my ideas.

So if you are well versed in analytics in hockey you can quit reading now. If you ever find yourself saying things like “stats are stupid” or “stats don’t tell the whole story” then this article is for you.

The Toronto Maple Leafs and Advanced Stats

The point of using statistics to measure things in hockey is to find inefficiencies where you can gain an edge over your opponents.  It’s a salary cap, professional league, and this means that the margins to win are tiny.

So if you can find any kind of inefficiency to exploit, you can have an advantage over your opponent.

The Leafs are a multi-billion dollar entity, they are said to be expanding their analytics department by 50 million dollars, they hired a 31 year old GM (the hockey equivalent of a blue moon) because he is an expert in analytics.

More from Editor In Leaf

They aren’t doing this on a hunch.  It is 100% clear, beyond a shadow of a doubt, indisputably, that using statistical analysis will lead to better decisions than the eye test or traditional scouting. Since they know this and are the richest team in hockey, the Toronto Maple Leafs are going to exploit a market inefficiency in finding market inefficiencies.

Statsception!

Because of fan resistance, Dubas and his ilk will make concessions to fans that are meaningless, and purely political, by saying things like “Analytics are just one tool in our tool box.”

They say this to quell outrage and angst among their fans. They don’t mean it.  The NHL is a highly conservative place and change is slow.  People don’t like it when you tell them they are doing things the wrong way – especially when they are supposed to be experts.

So sure, the Toronto Maple Leafs have other tools – they would be stupid not to.  But they give  a message to the media and fans that implies that they use a 50/50 mix of analytics and the eye-test.

That is nonsense.  Stats do, in fact, tell the whole story, and as such, no one is flipping a coin when the data contradicts the visuals.

An example would be William Nylander.  The fans and media turned on him.  The counting stats from this past year (goals and assists) were low.  Traditionally, when this happens the player is traded.  I would guess in hockey history, he gets traded almost 100% of the time.

But I bet you almost anything that Dubas is sitting there wryly laughing.  He got Nylander signed for probably three million dollars cheaper than he’ll have to sign Marner for.  But at 5v5, statistically, they isn’t three million dollars difference between them. More like two bucks difference.  Dubas knows that he just got a massive bargain on Nylander and that he’s likely to look very smart a year or two from now.

Stats Resistance

One thing that I frequently see is that statistics are “not the whole story” and this is both true and false.  If you look at a single statistics – points, goals or Corsi – then it is true.

But if you look at a wide range of statistics, then they will in fact tell the whole story.

Take for example plus/minus.  If a player is a minus, you would think it would be hard for him to have helped his team in a positive way.  But there are only seven to ten goals for every 100 shots taken.

This means that goals are fairly rare, and that because of this it takes a very long time to build up a sample size where the results would be reflective of the play.  Therefore, plus/minus is a bad indicator of a player’s play.

But if you include plus/minus (which is just basically goal differential) into a wide range of different stats, you can get a much more accurate player evaluation. If a player is a minus, but his team is over 50% in all the things that results in goals when he is on the ice (shot-attempts, shots, scoring chances) then he is probably still good.  And if the inverse is true, then he probably isn’t as good as he seems. (See Ron Hainsey and his gaudy plus/minus stats this season).

This should be common sense, because the more information you have, the more accurate your assessment will be.  And this is a rule that applies to everything, not just hockey stats.

Still, you will never have enough information. All analysis may be flawed.  But you can get a very good idea. Situations change, and a coach might find a way to use a statistically flawed player to advantage.  There might be things the statistics aren’t measuring.

This is all true.  But it’s said in a way to discount all statistical analysis whenever a person finds the results of said analysis to go against what they previously thought.

And in all cases, the statistical argument should prevail. The reason? It’s backed up by actual evidence that is derived from measuring.

The eye-test (as it is called) is extremely flawed.  First of all, it would be impossible for a person to watch enough hockey to accurately have an opinion on every player in the NHL.  Since we all pretty much do have opinions on every player, it’s reasonable and obvious to assume that most of our beliefs about NHL players and teams comes from reputation.

Secondly, even if you could watch enough hockey, there are over 10 000 events in every game that are tracked.  You could not organize all of this information in your head and properly evaluate it. No one could.

Third, hockey is a subtle game were small things that often go unnoticed can have big impacts over the long term.  This gets lost when highlights show only big plays.  If we see Gardiner make errors on Sportsnet, we start to notice when they happen in games because we are looking for them.  This is called confirmation bias and it forces viewers to over-weigh big plays and under-weigh small plays.  It can lead to a very big discrepancy in how a player is viewed.

Jake Gardiner has the advanced stats of a top pairing player, and he makes no more errors than your average player who plays that much.  But even though he’s one of the better defenseman in the NHL, some fans think he shouldn’t even play.

But this isn’t a case of two disagreeing parties where the best answer is in the middle.  The people who hate Gardiner are wrong. The people who dismiss statistical analysis are wrong.

It is exceedingly clear that the eye test is so problematic that it can’t be trusted. Hockey is a fun game to watch and you should continue to do so. Contrary to popular opinion, “stats people” do watch the games.

They just do so with a knowledge that they are very likely to draw bad conclusions based only on what they see.

Video analysis is extremely important. Scouting is important.  But information is power, and in hockey, like everything else, that information comes from measuring and analyzing what you measure.