Oilers and bears and fancystats, oh my!

Last night’s Oilers-Bruins game provides some interesting fodder for the world of hockey fancystats.  Because the eye and the fancystats appear to be directly at odds.

Who’s right?

What say you?

What the Experts Saw

Ignoring the numbers for a moment, we have a lot knowledgeable observers, including this rather respected one, saying things like:

“We weren’t at all pleased with how our road trip ended in Toronto,” said Edmonton head coach Todd McLellan. “To come back and play and engage in a game against a big, physical team, was something we needed.”

Pretty clear, right?

He’s saying the team was ‘engaged’ against a tough team.  That’s good. (My eyes happen to agree)

He’s also saying this was a better game than the Leafs game, despite the fact that the embarrassing loss to the Leafs looked really good by the fancystats (CF% 52% and DFF% 59%) … while this thrilling win over the Bruins game looked pretty bad stats-wise.

Two busts in a row.  Obviously, fancystats have no value.  Right?

What the Fancies Say

On the other hand, here we have some unequivocally lousy numbers:

  • 5v5 Corsis = 43%
  • 5v5 Dangerous Fenwicks = 41%
  • On the teams, the short handed goal is nice, but otherwise pretty much nothing was generated on the actual powerplay.

These numbers aren’t ‘telling us’ something because they hate us.  They’re basically counts.  They’re objective and they don’t forget parts of the game because they were ugly, or because they went to the bathroom.  And the counts tell us that the Oilers got badly outshot at even strength. That’s a fact.

These clearly say: the Oilers got owned.

Add to that the fact that Nilsson had to play his heart out (a huge part of that successful penalty kill) in order for the team to squeak out a SO victory, and it’s pretty clear that anyone who saw a “good game” from the Oilers – even the knowledgeable ones – are suffering from recency and outcome bias, creating the distorted recall that Steven Novella speaks of:

When someone looks at me and earnestly says, “I know what I saw,” I am fond of replying, “No you don’t.” You have a distorted and constructed memory of a distorted and constructed perception, both of which are subservient to whatever narrative your brain is operating under.

So again I ask: which is right?

Got your answer?

Well then … let me give you my answer.

Writes and Wrongs

For those of you who feel that the fancystats are clearly indicating that the optimistic observers of this game are unable to assess it objectively, and their positive vibes are simply self-delusion … well, I think you are wrong.  The eyes are telling us a lot.

On the other hand, for those of you who feel the poor Corsi (or other shot metrics) last night do not reflect the game, because the Oil won, and you saw ’em try hard and work hard … you are also wrong.  The numbers are, objectively and thoroughly, telling us a lot.

Which means whatever side you chose, you are wrong … but you are also right.

Before I explain what I mean, let’s step back a bit.

Despite the artificial antagonism that the Luddites of the world throw at fancystats as a lens for looking deeper into the game of hockey … the reality is that the eyes and the stats often tell us the same thing.

We don’t need Corsi to tell us the Kings and Hawks have been good the last few years, do we?

Or that the Oilers and Sabres have been bad?


They usually agree.  In fact, my rule of thumb is that if a stat always fails the eye test, the problem is usually with the stat, not with the eye.

That said, assuming you have a stat that meets both the eye test and relevant statistical tests … what happens when the eye and the stat disagree?

Well, I’ll say this to you: it’s exactly when a well established stat (or stats) and the eyes (or results) disagree that things get most interesting!

For example, when a team loses a game and has a high Corsi, the numbers aren’t lying to you.  They are telling you, accurately and objectively,  a simple fact: that the team heavily outshot its opponent.  It’s still possible to lose when that happens. In fact, it happens all the time.  It’s usually not that tough to figure out what happened – goaltending (good or bad), special teams, reffing, or plain bad luck.  Doesn’t change the fact that shots are good.

The real question to ask is … what happened to cause that divergence?  When we get the answer to that question – and we usually can – it often tells us something interesting.

Remember last year when the Flames were heavily outperforming their lousy underlying metrics?  Some used that to point out that clearly fancystats don’t work. An entire mythology (“comeback kids!” “Hartley for the Adams!” “Heart!”) sprung up to explain why a lousy team was winning.  An entire city convinced itself that giving up a big lead was somehow a deliberate – and winning! – strategy.

Meanwhile, others pointed out it was inevitable that the Flames would regress.  Lousy teams can and do outperform lousy numbers in the short term all the time – even for an entire season.  But it doesn’t last.

Who was right in that case?  Or about Toronto before them?  Or about Colorado before them?

Last Night’s Game

So let’s look at last night’s game.  The numbers and the eye have clearly diverged.  Is this just a case of the two randomly not agreeing, or is there something more interesting there?

Well, on further investigation – it turns out both sides are right (and wrong), and maybe for understandable reasons.

To explain, let me throw this chart at you (from naturalstattrick.com), which shows the shot attempt differential between the two teams.  I’ve annotated five segments of the game for you:


A. So we have a rare good start.  The first minute or two, the Oilers were the better team.

B. Then the Big Bad Bruins kicked into gear, and started using that gear to kick Oiler ass.  Twelve minutes or so to drive a +2 shot differential all the way down to -10. That’s a beating! The “seen it good” folks may perhaps forgotten about this part, yes?

C. But wait!  Our Oilers are not so easily cowed. Anders the Giant kept them in it, and now it’s pushback time.  And not a small pushback either. Even though the slope of the line may be gentle, it is real and it is spectacular, and it lasts all the way to the middle of the third period.  For thirty five straight minutes of game, the Oilers are the better team, clawing all the way to a 2-1 lead, not to mention all the way back up to a -2 shot differential.

Eight shot attempts better than the Bruins in other words. When the folks who say ‘the Oilers looked good’, this is what they’re referring to.  They did look good.  They were the better team for most of the game.  Todd’s not fibbing.  *Lot’s* of good things in that game.

D. Except oops. At this point, the game went quiet, and the Oilers sagged. They had a chance to go up 3-1 by scoring on a powerplay. Two listless minutes later, and the Bruins went into hyperdrive.  Completely dominating the Oilers and dragging them all the way to -16 and tying the game – in about two minutes!

Holy crap. Surely the seen ‘im goods didn’t forget this, did they?

E. And then finally a bit of a pushback by the Oilers after the tying goal, and yet another pushback on the pushback by the Bruins – leaving the teams more or less where they finished after the Bruins scored the tying goal.

All topped off with a thrilling 3v3, and a satisfying end in the SO. And a heavy dose of thanks for Anders.

So Let’s Review

So the Corsi folks are right – the Bruins overall badly outshot the Oilers. It took a stellar effort from the goalie to grab two points.

But the observers are also right.

For about two thirds of the game, the Oilers were actually, measurably, the better team. That’s what the eye remembers.  The Bruins earned their shot differential in about 16 minutes.  A crazy 16 minutes, with the Oilers in a shell and the Bruins out for blood.  But only 16 minutes.

The point of explaining all this should be clear. The eyes told us one thing. The fancystats told us (or perhaps, objectively reminded us) of a whole lot more.

We understand the story of this game better because we incorporated both sets of information.

I don’t remember who said this (I wish I did so I could give them credit), but it goes something like this: there are no hockey stats people and non-stats people.  There are only people who use more information, and those who refuse to.

And that applies to both sides of this entirely artificial and unnecessary divide.

Got that? If you care to understand what happened or is happening with your team, stop picking sides.  Use MORE information, not less!



6 thoughts on “Oilers and bears and fancystats, oh my!

  1. I like your approach of illustrating the game from the perspective of re-integrating the storyline through the stats measure.

    To harp on your theme, the rejuvenated version matches more or less the live version and we’re left with a pretty good take on what “actually” happened.


  2. Great article G. I appreciate the tone as well – a defense and a criticism of both eye and measure. A balanced approach with no favorites; the bias of both sides of the argument has always struck me as odd, especially the bias coming from the so-called “objective analysis” party.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s