An EKG for the game?

Weird Obsessions

I’m kind of obsessed with defensemen right now.

Not in a creepy way, honest!

But I use a lot of my spare brain cells, all three of them, to mull on how we can better understand which defensemen played well, and which didn’t.

If you can’t objectively assess how well a defenseman played in the last game, you have no hope of making a good decision about who is going to play well next game, yes? Never mind assessing for trades and signings.

The eyes we know can be easily misled, influenced by anchoring, confirmation, recency, and outcome bias (to name a few).

On the other hand, the fancystats as we use them today to my view haven’t proven to be consistently useful at assessing defensemen either, certainly not in comparison to their usefulness for forwards and teams.

Plus of course, the Oiler D (often through no fault of their own, such as youth) have generally not been very good, and the Oilers have made some notably terrible signings, which amplifies the need to have better assessment tools.

That’s one reason I came up with and publish “Dangerous Fenwick Against” rates for defensemen and defensive pairings.  These show unblocked shots against, weighted by the ‘danger’ of the shots taken. Hardly a proven methodology yet, but it has matched up and complemented reasonably well so far the eye test and coach’s TOI.

Now, how to make that information better, clearer, and more understandable to more people?

Charts, that’s how!

I spent a lot of time thinking on how to visualize data.  The same way a (good) picture can be worth a thousand words, a good visualization can be worth a thousand cells in a spreadsheet.

One of my favourite charts is the Game Flow chart at naturalstattrick.com.

Unlike most of the commonly available in-game shot charts, which show the shot metrics for both teams as separate lines, this one shows you the difference between the two. I find it gives a really terrific snapshot of the flow of the game.  So naturally, doing this for my Dangerous Fenwick seems natural (punny sentence unintended).

So that’s what I’ll be publishing. It looks a little something like this (this is from the last Dallas game):

Cumulative Danger Fen 5v5

You should be able to see pretty clearly the ebb and flow (mostly ebb…) of the game.  The line shows both Dangerous Fenwick and regular Fenwick; the divergence between the two in the first period shows the really deadly chances the Oilers gave up early.  You can also see the deadliness of the chance that resulted in the goal against late in the second.

Now – can we apply this to individual players?  Yes, we sure can!  Here’s what it looks like for the six defensemen that played against Dallas:

EDM vs DAL Defensemen Game Flow

Some of the information there is no surprise.  Sekera and Nurse had a good game, but got snowed under a bit – well, a lot – in the third.  Some of it is surprising, like Gryba’s chart.  Lots of observers didn’t like his game, but it sure looks like McLellan’s playing him as much as he did was a sensible decision after all.

The chart looks complicated, but it isn’t as scary as it looks.  Here are the key elements:

  • Just like with the game flow chart, the yellow and purple lines show the progression of Fenwick and Dangerous Fenwick.  The difference is that it only shows it for when that specific defenseman is on the ice.  In between it flatlines.
  • The purple bars show you the shifts for the player in question.  A lot of shifts, nothing good OR bad happens, and the flatline continues.  That in itself is valuable info.
  • The goals for and against are shown as dots.  Only goals that occurred while that player was on the ice are shown.
  • The header gives you the Dangerous Fenwick % as well as the rate against for the player in question.  So that shows you the end result, while the chart shows how that player got there, which can be just as or even more important.

After each Oiler game, I’ll be showing these charts not only for the defensemen, but also the top 3 defensive pairings and the top 4 lines.

I’m hoping this will act kind of like an EKG for not only the game, but for the key player units that the Oilers deploy.  Interested to hear if you agree.

Advertisements

7 thoughts on “An EKG for the game?

  1. Graphical representation of statistical measures is more important than the stats themselves for argument. This is super! One of the anti-stat arguments in hockey is that each game has its own story, and ebb and flow that can’t be captured in a numerical measure. Well, BS! Express that measure in relation to another variable – time! This is a far more convincing summary of the game than the box score. Please keep up with this project.

    Like

  2. Very nice. Since any specific player is on the ice only for a short part of the game, I’d dispense with the dead time and compress the time scale to only TOI. This would add information as well: a longer EKG indicates more TOI.

    Is DFA per player adjusted for quality of competition? (Don’t need to adjust for the overall chart, obviously, which btw is sick).

    Thirdly, is it possible to calculate a normalized slope of that team EKG? So you can see where the momentum shifts are very clearly and perhaps what (er who) caused them? I’d expect a lot more EKG volatility in the second period, for example, with the long change.

    Like

    1. Hi New, and thanks. Compressing the timescale is tougher than you’d think, especially with the multiple plots per panel that I use. I might see if I can get it working, just to see what it looks like.

      DFA isn’t adjusted or QoC yet, partly because it’s hard and partly because I’m not a fan of current QoC measures, so if I’m going to do that work, I’d like to solve both problems if I can.

      As far as normalized slope, were you thinking of say a moving average? Or maybe just adding a rate of change line? You’re definitely right, it’s the rapid changes that represent the most ‘interesting’ situations in the game.

      Like

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 )

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s