What Analytics have shown me, and what I still don't support

Analytics are slowly becoming more accepted in hockey so I thought I’d take some time to write out my own feelings on them; if nothing else it allows me to sort out my thoughts on the matter and grapple with some of the thought processes here.

It hasn’t been a secret that I’ve been pretty against certain “advanced” analytics, but I will admit there are a few things I have come around to and I’ll gladly acknowledge them.

There are two main things that stats can do that I really like: they place the focus on what is important, and they cut through external crap that is not always important like style points, “dressing room qualities,” and a few other things I’ll touch on below at some point.

An example that I think is fairly basic would be Dion Phaneuf on the power play. I watch every single Leaf game once, if not twice, every year and yeah I would agree that Phaneuf does make head-scratching decisions on the power play, gets tunnel vision, and can drive me nuts, but hey guess what? This year he was fifth among defensemen in the entire NHL for points on the power play while last year he was tied for sixth. However, he can’t hit the net? He was tied for fifth among defensemen in goals this season and tied for sixth the year before.  

So yeah, Phaneuf isn’t as agile with the puck as I’d like to see, and he sometimes rips shots that go way over the net, or makes really bad plays, but at the end of the day he is an elite power play option on the point right now yet a lot of people in Toronto act as if he is the black hole of power play’s where pucks go to die.

Overall, the Leafs power play has had the 14^th and 10^th highest conversion percentage the last two seasons too, so it’s been a good overall unit.

That’s a pretty basic “analytic” example, that isn’t really analytical, but to me that’s an instance where you can use numbers to say “hold on a second, the facts are right here and you’re without a doubt wrong.”

When we get into newer numbers such as CORSI, FENWICK, PDO, etc. then things can become a little trickier to use because not everyone values those numbers the same way. That said, there is a time and place to use numbers because our eyes can deceive us, and many of us often have personal biases towards players that we might not even be aware of.

Furthermore, as I alluded to above, analytics try to narrow down what is important and I appreciate that. Now I’m not here to argue that they have found what is important and are properly gathering data toward that- whether they are or aren’t is not my point here. My point is that we tend to look at the things that appeal to us and value them way above what we should be actually valuing.

I’ll stick with Phaneuf here as an example because he is a lightning rod of criticism. A lot of people have remarked to me that he is “not the same guy he was in Calgary” because he isn’t throwing massive checks, he isn’t putting up insane point totals and he isn’t the human highlight reel in general that he once was. 

However, if you look back at his time there, was he playing against top lines while putting up those numbers and making those hits? No; at the very least not regularly, let alone every single game like he does with the Leafs. Is his role to be all offense all the time, or is it to be a shutdown guy that produces when he can, particularly on the power play? Obviously it’s the latter.

Yet people bemoan that he isn’t the “Calgary Phaneuf” even though a top pairing defenseman that can actually play against top lines is substantially more valuable than a second pairing guy who throws big hits and racks up points.

It shouldn’t even take analytics to show us that, but as I said, sometimes we need a narrower focus on what’s really important and analytics can do that for us.

There’s also a third thing that I appreciate analytics for, that I have not mentioned yet, and that is their ability to project point totals over an extended period of time. There are a few stats that can combine to give us this sort of indicator and to me that’s valuable- shooting percentage, PDO, even CORSI and FENWICK and of course scoring opportunities, all show us if a player has been fortunate, whether he produces scoring chances regularly, and if he has the puck in general.

So to me, if I’m thinking of signing a guy to a five year contract for 4.85M per year like the Devils did with a turning 31 year old Ryane Clowe, I’m definitely pulling out my analytic team and asking a few questions such as:

1. What kind of production do we associate with a 4.85M cap hit?
2. Let’s say the answer is 20 goals and 50 points, how viable is it that Clowe produces that in each of the next five years?
3.  When the cap inflates and Clowe’s percentage of cap space taken lowers, what should his new expectations be and is that attainable?
4.  What does he bring beyond simple goals and points that we can associate with value and how much longer is it realistic that he can do that- because in this case we’re talking about his ability to bang bodies, fight, etc.

Those are just some of the questions I’d ask my analytic team, and while I’d by no means expect anything close to 100% accuracy, I’m reasonably confident that these numbers can help paint a pretty good framework for what to expect.

So yes I acknowledge some strengths in analytics in which they can help us project the future, they force us to stick to what’s important and the “facts,” and they cut through biases and can even cut through myths.

However, the above example on Clowe is also how I view analytics in a practical sense. I use my eyes to grade players and I always will. For me, I might look at a player like Clowe –although realistically I’d target someone better and not on the downside of his career, but let’s run with this example—and I’d say okay I think he’s a pretty good player. Clowe does this, this, this, this and this that I really like, and I think these attributes will do this, make this better, help this guy get better, and so on. You get the point.

It’s at that point that I bring analytics into play here and say okay tell me how the stats look here. Odds are with almost any player, especially one that is hitting the free market, that something isn’t right. So then I’d go back to the tape and ask why a certain stat is off and I’d try to see the reasoning behind that.

For instance, most people in Toronto acknowledge that Mikhail Grabovski got buried in a defensive role which is why his offense went down the drain, and I’d agree with that. However, he played over 15 minutes a night, saw steady second unit power play time, shot a high shooting percentage for the third straight year, yet still managed only 16 points on a high scoring team. I mean, every Leaf that played the full 48 games like he did had more points than him, so you have to at least raise an eyebrow to that.

When I watch the tape though, I see a guy who played between Jay McClement and Nikolai Kulemin and they were just offensive anchors for Grabovski. There were a lot of passes he made that bounced off each of their sticks repeatedly, and while Grabovski attempted to jump start the offense on that line McClement and Kulemin were more likely to grind in the corners than create passing plays, shooting lanes and scoring opportunities. As much as I’d look at his stats and say these things can improve if he isn’t starting this much in the D-zone, or playing against top lines all the time, I’d need to see that tape that says “you know what, his skating is still there, he can still beat players one-on-one, he played with some pretty bad offensive players last year –God bless them as they both work so hard, but come on—and I think if we flank him with some talent and play him in some favourable match-ups, he should be able to score for at least the next few years.”

So if I’m a team like the Ducks I’m thinking there’s a very good chance Grabovski will produce well and be an asset to my team if he’s playing between, say, Dustin Penner and Jakob Silfverberg.

But I need to have that video and statistical evidence there. I can’t solely have one or the other.

Now this is the part where people will tell me that “nobody only uses analytics” but that’s just a huge lie. I’ve read more than enough stuff online to confidently say that people definitely write “this guy is good/will be good/should be acquired, kept or whatever, because of this number, this number and this number.” And that’s it. I’m not going to call anyone out, I’m not going to link to anything in particular, but it happens and that’s where analytics lose me at times with a huge eye-roll and a page exit. And I’m not even going to bother discussing the arrogance some analytic-types have which really just turn people off of the matter altogether; I will note I’ve noticed some are making an effort to not be dicks and are generally speaking good guys to talk to and I think that’s awesome.

This further spreads to another point in analytics in which people discuss what teams do and don’t use them. First of all, every team uses them to some degree, let’s not kid ourselves. Beyond that though, I really dislike when people mention teams like Chicago, Boston, Detroit, and Pittsburgh as analytic teams.

The Hawks drafted Toews, Kane, Seabrook, and Keith just to name a few; Boston drafted Bergeron, Lucic, Marchand, and Krejci; Detroit drafted Datsyuk, Zetterberg, Kronwall, and I’ll name Lidstrom when they were winning Cups and so on; Do I even need to name who Pittsburgh drafted? I’m sure all of these teams use analytics to supplement that, but naming those teams as primary advocates and shining examples of these methods is such garbage in my opinion because those teams were all primarily built through the draft with elite players either falling in their lap like Pittsburgh, or striking gold like Detroit and Boston.  I don’t think this devalues analytics whatsoever, the point is simply that mentioning these teams that use analytics as proof that they are to be taken seriously is not why I do, nor should it be to be honest.

Editor’s note: I know there are people who say “nobody in hockey actually uses analytics” so people bring up these examples, which I guess is sort of justification to bringing them up, but to that I say show their value and worth, not a patchwork example of usage.

Beyond that there are a few other slight bones I still have to pick.

The first is that I think a lot of people pick and choose their stats to suit their needs at the time. Some people say that goals and assists aren’t important, it’s possession; but when David Clarkson, who is a good possession player, is signed by the Leafs the focus is on his points, or lack thereof, over his career. Do I think Clarkson drove possession? No, that was probably Elias. Do I think Clarkson has a good contract? No. Do I think he is a player that is strong on the cycle and is a good possession player should we look at it in those terms? Yes. If he plays with a good possession driver in Kadri, which seems like the plan, can we expect a good possession line? I think so, and the stats reasonably show that. His goal scoring, at least this year, also can’t be attributed to a high shooting percentage and his PDO says he’s been unlucky if anything. Considering the analytical theme of the year for the Leafs was that they suck at possession, it’s pretty interesting to hear that a player who is actually good at possession will be a bad signing because his point production will drop quickly over the next few years.

A lot of the time people use stats after they have already taken a stance on something and then they go out to prove it. I don’t respect that use of analytics at all. If someone has a question and then they seek out an answer and are open to changing and redoing their research based on issues with their original work, I have all the time in the world for that. Those people work their asses off and love hockey just as much as anyone else and I respect that.

Advanced stats should, at the end of the day, be about getting a player down to one number. In a perfect analytical system we would be able to grade players out to their exact value and then they would be listed in order of the actual best player in the world to the worst, and then the issue would become acquiring them. Considering we have yet to do that in a manner which is representative of a player’s value compared to his counterparts, we can’t consider them law. I say this because it seems so often people value players strictly by their CORSI or fenwick.

When, for example, Ryan O’Byrne came to Toronto he was touted as “the worst player on the worst team” because his possession numbers were bad. Is Andrew Cogliano the best Duck because he had the best CORSI ON on Anaheim? Is Kane one of the worst Hawks because his number was low? This is how you end up thinking Bryan Little is better than John Tavares.

The final thing I’d like to say, and if someone has done work on this I’d appreciate a link showing me, is that I’m surprised I haven’t read much on the margin of error for analytics, how much of the game they are capable of capturing and so on. In other words, if we agree there are intangibles in hockey that can never be measured, whatever they maybe, how much of that is not accounted for in analytics? For example, does analytics equate to say, 60% of the big picture, while everything else equals 40%? How much of the overall picture are analytics giving us?

I do place some importance on things like chemistry, sticking up for each other, having a good dressing room and so on. Now don’t confuse that belief with me valuing those things over good hockey players, because good players are what win you hockey games at the end of the day, but you do need to have both. No company is successful in any walk of life where everyone has to work together yet nobody gets along. That doesn’t happen. So I mean, yeah, that’s a real thing but none of us can judge that from our TVs and computers so it’s not really fair for us to say either way. That doesn’t mean it doesn’t exist though.

There are countless examples of this kind of stuff either way. I think the Heatley-Spezza-Alfredsson line wasn’t amazing because of any chemistry whatsoever, I just think they were three dominant players in their prime all playing together. Whereas I’d look at the line of Fleischmann-Weiss-Versteeg and think the sum was greater than the parts and that they did have chemistry.

Hockey is just a game and sometimes these stats go way overboard in my estimation- adjusting stats based on zone starts and such is just something I’d be hard pressed to ever support because there are so many variables at play there. However, there are also people who completely ignore all stats whatsoever and are truly oblivious to some fantastic information that is openly available. Really, the people that can combine scouting hockey with at least some form of statistical analysis are truly the smartest ones, and that’s what people should be striving to do.

As I said at the top, writing this article has helped me sort out some thoughts on the matter for myself. I’m not interested in having this piece picked apart and “proven wrong,” as they are just my own feelings on the matter. If you have something constructive to say and aren’t a D-bag about the way you come across in doing so I’ll happily listen and am more than open to having a change of heart on a certain matter, if not, I hope this article was still worth your time.

Cheers.