Wanna blog? Start your own hockey blog with My HockeyBuzz. Register for free today!
 

Stats, the Eye-Test, Leadership and Love

March 4, 2019, 11:35 AM ET [48 Comments]
James Tanner
Blogger • RSSArchiveCONTACT
Yesterday in the comments section, I came across a post by a guy known as Njuice, which I sure as hell hope stands for nectarine juice, his beverage of choice.

I thought it was a thought provoking comment, and when I went to answer it last night, I wrote a book and a half, so I decided to polish(ish) it up and post it today, as I think it might be interesting.

Perhaps not.

His quote came after I stated my belief that a player who isn't a positive differential in the majority of five major stats catagories isn't a good player.

Those catagories were shot-attempts, shots, scoring chances, high-danger scoring chances and goals.

[quote=Njuice]He doesn't understand anything about hockey. He mentions the five things he thinks make a hockey player a player - but using his own logic - those things are only results of other ACTUAL things which he knows nothing about.

Shots...high danger chances for and against....those are results of so many things... the things that make hockey players hockey players... NHL players are elite talents so I believe that positioning, decision making, effort/heart are the most important.

Ron Hainsey is not terrible. Ron Hainsey if a highly effective NHL defenceman who is straight up underpaid for what he provides...still I hope we can sign him for 1-2 years at $3mil.


I think you've hit the nail on the head here, Mr. Nectarine Juice

When you measure shot-attempts, shots, scoring chances, high-definition scoring chances and goals you are measuring the results of a hockey game. I 100% agree that a million things go into these results.

Things like leadership, chemistry, faceoffs, special teams, psychology, coaching, toughness etc.

When talking about hockey statistics and analysis, a lot of people complain that the so-called 'stats guys' ignore these things. That they don't understand them, and are subsequently misguided in their analysis.

I subject that this isn't true. Just the opposite, in fact.

They are included in the analysis.

Whatever intangible things exist in a hockey game, things that can't be measured directly, like leadership or desire, these are included in the final results.

It is not necessary for me to explain to anyone that good leadership results in wins. In hockey this is a foundational philosophy of virtually ever analyst on TV.

But just like leadership factors in the final results of wins, it also factors in the results known as shot-attempts, shots, scoring chances or goals.



Other than doing a large study where you took players known to be great leaders and measuring teams results with and without them, there's no way to measure leadership.

But you can measure the things that it potentially results in.

So when I say that Team X has a season long corsi-for of 53% that number includes all the intangibles people think we don't care about.

I suspect that almost universally, 'stats guys' grew up watching hockey and being huge fans of the game. We too were raised by people who strongly valued hitting, fighting, leadership, coaching, chemistry etc.

The best way to describe me when I started writing about hockey in 2013 (besides terrible) was as a Brian Burke, Dion Phaneuf, Tyler Bozak fan.

I was ecstatic when the Leafs first acquired Roman Polak for Carl Gunnerson.

I can't speak for anyone else, but my interest in stats came when I realized that I couldn't possibly watch enough hockey to actually have any true insight to the game, even if I watched three games per night, coupled with the fact that some people who I knew were smart and/or respected were saying basically the opposite of everything I believed.

For instance, how could I know with any accuracy if player a with 100 points was better than player b with 80 points based on a what I thought player B's defensive game was worth?

I realized that I was guessing, and at the same time, I was reading Twitter and hearing about how things I always thought were true were in fact quite wrong. Things like the value of faceoffs, leadership or stay-at-home defenseman.

I decided to base all my future analysis on statistics.

And since I was basically learning on the job, I'm sure this led to some extremely questionable interpretations of players, stats and other things. (But not on my views of the Foo Fighters as the worst example of the most boring, corporate, garbage rock in history. Those remain unchanged).

But regardless, you are supposed to learn and change your mind as time goes on. If you don't, or can't do that, you're not going to be very happy or successful in anything that you do.

Getting back to the post above, I probably don't understand that much about hockey. At least, I'm not going to defend myself about it. What I don't know could fill a book and I'll keep trying to learn.

What I do know is that a person cannot hope to keep track of all 10 000 things that happen in a hockey game. Our human minds cannot help but overrate high-lite reel plays, or recent events. That is why we need statistical analysis.

I think much of the pushback on statistical analysis comes from the term "biased." People take it as a purgative and get defensive, but in fact 100% of humans are perceptible.

This doesn't mean you shouldn't watch the games, it just means you should be wary of things you see that are contradicted by statistics,. By the same token, you should be wary of statistics that don't seem supported at all by visual evidence or common sense.

Every situation is different and there is no binary here, it's a very subtle thing to find a balance, or weight to every thing you can know/think/see/compare etc. and it's 100% subjective. (But again, something being subjective doesn't mean there isn't a right answer. Two doctors can have different opinions on the severity of an injury, but either one still has a better answer for you than my Uncle Chuck).

So my theory here is that if you take five major results and look at them, you can tell if any player is any good.

When the player is on the ice, are his differentials in shot-attempts, shots, scoring chances, high-danger scoring chances and goals positive or negative?

If they're mostly positive, that player is having a positive effect, and if not, then he isn't.

What could be more simple?

You could definitely continue to break it down and add more categories, but it's going to get cumbersome and difficult to analyze.

But just because something isn't exclusively mentioned, doesn't mean it isn't an important factor.

I think zone entries and exits are pretty important. I like to know if a player is good at them, but ultimately, just like leadership or character, they are going to show up in the shot-attempts, scoring chances and goals.

If a player led the NHL in zone entries but was a negative differential in the five major categories, I'd have to conclude that zone entries are not a major factor in hockey games. They might help, but seeking out a player with this specific skill would be fruitless.

If you hit, display leadership, character, grit and heart, but every time you're on the ice, you are a negative in all major results, then those things aren't helping either.

Now maybe those things do help, in which case they're showing up in shots, goals etc.

I could be wrong here, and if so, I'll change my mind when I learn new information. But for now, I can't see how it is possible that a player can be a negative differential in shots-attempts, shots, scoring chances, high-danger scoring chances and goals while still being helpful to his team.

Let's take a look at Ron Hainsey - his 5v5 stats are as follows:

shot-attempts 48%
shots - 46%
scoring chances 48%
high danger scoring chances 47%
goals - 60%


Ron Hainsey, who either is or is close to leading the NHL in +/- is a negative differential in every stat except for goals.

A lot of people are happy to look at this and say "well, goals are what matters." In this case, they would be wrong to say that. Goals are the rarest of these five results and as such they are the least repeatable.

Put another way: if you ran a million simulations of a player with Ron Hainsey's stats, he is going to be a negative goal differential player more times than he is a positive one.

Given that Hainsey is a negative in four of the five statistics, we can reasonably conclude that the fact that 60% of the total goals scored while he is on the ice are for his team is a total fluke.

In fact, Toronto goalies are saving 93.5% of the shots when Hainsey is on the ice. Using data going back to 2008, we can see that Ron Hainsey's on-ice save percentage fluctuates all over the place ever year, as if it were random. (It is).

If he was the same every year, we could assume that he does something that makes it easy for goalies to make saves, but this very clearly isn't the case.

The Leafs' results when Ron Hainsey is on the ice are that of a non-playoff team unless he gets hall of fame level goaltending to balance out the goals. (For Context, Robin Lehner has a .929% this year and is a shoe-in for the Vezina at this point).

Now I guess you could value Hainsey's role as a penalty killer, but he plays 16 minutes per game 5v5 and 3 minutes per game short handed. Even if he's the best penalty killer in the world, I'm not sure that forfeiting 16 minutes of a game would be worthwhile in order to have him.

There is of course one complication we must consider before declairing Ron Hainsey a terrible player. How does he compare to his team? All stats provideres, like Natural Stat Trick or Corsica, provide Relative data, which allows you to compare your player to his team, or the players he plays with.

When we look at Ron Hainsey, we see that he plays for one of the best teams in the NHL and that relative to his team, he is very ineffective. Thus we can feel confident that our judgement is correct.

If, however, he played for a bad team, and he negative differentials, but they were better than his team, than that would be an indication that he is helping out. It is important then, if a player has bad stats, to make sure that they're bad compared to his team and teammates before declaring him a bad player.

So anyways, I hope that clears up the rivalry between people who love hockey but do and do not like hockey stats based analysis. I hope we can all be friends now and that I am hailed as a genius for finally bridging the divide, that Jeff O'Neil takes to the air to declare his new found love of Corsi, and Don Cherry makes a rare mid-week appearance to denounce his views of Adam McQuaid.

all stats naturalstattrick.com and 5v5 unless noted.
Join the Discussion: » 48 Comments » Post New Comment
More from James Tanner
» I am Just Curious If This Works
» NHL At Least Tries to do the Right Thing
» The NHL Cannot Remain Apolitical and Must Show Leadership
» Time for a New Coach to Go Along with the New G.M
» Coyotes Eliminated Following Severe Beating