Welcome everyone, to what is my first analytics post here at BeerLeagueHeroes. I go by the pen name “G Money”, and for the past year I’ve been building up some pretty cool analytic data and visualizations and publishing them on my personal blog, OilersNerdAlert.
The subliminal messages I embedded in those visualizations finally tricked, er I mean convinced, the folks here at BLH into thinking I have mad stats skillz, and so here I am!
(Rather than bore you now, I’ve put a bit more biographic info at the end of the post – so I can bore you later)
My plan here at BLH is to do some deep dives into aspects of the game we all know and love, on topics that are amenable to an investigative/analytic approach … and in so doing, hopefully entertain, educate, and enlighten.
Famous last words! So on to the topic of this post … one Mr. Leon Draisaitl
Dr. Drai’s Forgettable February
Much (virtual) ink has been spilled of late about Leon Draisaitl’s recent slump. Since late January or so, he hasn’t been as effective as he was prior to that, and by extension, neither has his entire line. You can see this by eye, by scoring, and by the stats.
Here are the most common of the theories that I’ve heard floated:
- He’s wearing down from a long season of facing first line duty in Nuge’s absence
- He’s fatigued (which is really just the same as wearing down) because he’s young and not used to this much TOI for this long a season
- He’s injured
Short of putting a listening device in the trainer’s room, is there anything we can use to help figure out which of these it could be?
Why yes, I do believe there is!
Specifically, let’s see if we can test the idea of whether or not Dr. Drai is fatigued or otherwise wearing down. How would we do that?
Let me float this hypothesis: if Dr. Drai is fatigued, he’s going to become less and less effective over the course of the game. Fair? Even tired people can start strong. It’s in the later periods (or rounds, or whatever) that fatigue tends to assert itself.
So, here’s the idea: let’s look at how effective Draisaitl was at driving shot metrics over the course of the game before February, and how he does during February, and see if there’s a difference.
If he really is fatigued or worn down, we should see that he gets weaker over the course of games in the month of February than he did before February.
If on the other hand it’s just (ha ha – “just”) his overall effectiveness that is compromised, we’d expect that his pattern will show him struggling over the entire course of the game, rather than weakening in the later stages of the game
Does that make sense to you?
It does to me. So let’s test that.
We can’t just look at Leon’s trace over the month of February vs the rest of the season. What happens if the entire team went into a funk in February (which it kind of did for a while)? What if the entire team was tired? In that case, we’d see a weakening in King Leon’s shot metrics over the game, but it would be reflective of the entire team’s weakness, not his.
So we have to look at Draisaitl’s relative shot metrics – how he did compared to the rest of the team during those two different time periods. That way we can isolate Draisaitl from the overall team trends.
But wait! One more wrinkle!
Is February really the right comparison timeframe? If we don’t pick the right timeframe, we could end up with mixed data, where we don’t adequately separate the good and the bad timeframes. That might be enough to hide differences in the data, so it would be good to confirm whether we have the right timeframe. So let’s take our first quick look at some Draisaitl data:
* Click to embiggen
What is this strange looking duck of a chart?
What I’ve done is looked at three datasets: shot metrics, shot metrics relative to team, and scoring rate. There’s both the raw data and a smoothed version showing for each. Highlighted in purple are the points where it looks like a downturn could be argued to have started, at a peak on either the raw or the smoothed data. In orange are some points where that downtrend might have ended.
It’s not cut and dried (it never is, despite the precision with which we often associate numbers-based analyses), but it does look like February is a separating factor for “good Draisaitl” and “bad Draisaitl”.
Using February vs ‘not February’ as our basis for comparison seems like a reasonable choice.
* Data for this chart was downloaded from war-on-ice.com
Draisaitl, before the walls fell
Here’s the centrepiece chart for this analysis:
This is visualization of Draisaitl’s shot metrics over the course of games excluding February. These are ‘score and venue adjusted’, meaning they try to account for the score state in each game and whether it’s a home or away game.
It’s “Corsi over the course of the game”, so of course I’m calling it Coursi, as a matter of course. What else could I call it?
First impression? Man, is Draisaitl a strong player! Every period, the team is better with him than without him. And apart from a couple of short letdowns early in both the first and second periods, he remains strong throughout. In fact, 51 to 56% means the Oilers are pretty much an elite team when he’s on the ice. Quite encouraging.
Now, let’s temper this a little bit – Draisaitl has spent most of this season with Taylor Hall. We already know Taylor Hall is elite. We already know Taylor Hall drags pretty much every single player on the Oilers above breakeven when he’s on the ice. But the good news is that Drai was with Hall throughout February, so fortunately we don’t have to worry too much about that particular teammate effect biasing our comparison.
The key chart in some ways is the third one – the purple smoothed line shows Draisaitl’s results when compared to the team as a whole. We’ll get back to that line later!
* Data for this chart was scraped by my own scripts directly from NHL data. Corsi calculations use score and venue adjustment factors provided by Micah Blake McCurdy (hockeyviz.com).
Draisaitl on the ocean
Here’s Drai just in the month of February:
That’s quite a contrast, yes? You can very clearly see that Drai is not doing nearly as well in February. That’s true on an absolute basis, AND on a basis relative to the team – the purple line on the rightmost chart dips below the breakeven line a lot more often than it did outside of February.
Hang on a sec, though. That’s not what we were looking for. We were already pretty sure that Drai had a tough February. What we wanted to see is any evidence of fatigue.
Do you see one? I don’t. There is no noticeable downward slope on that purple line.
Looks to me like our man Drai’s issues, whatever they are, are something that stay more or less steady throughout the game.
We can see that effect even more clearly if we isolate the two smoothed ‘CoursiRel’ lines on a single chart:
Again, I’d say it’s pretty clear. February is lower overall (more time below breakeven). It is also more volatile. But it’s not sloping downward over the course of the game.
I’m going to go out on a limb here and suggest to you that whatever is ailing Leon Draisaitl – it’s probably not fatigue.
One More Control
One last chart for you. When we look at the chart above, we see quite a difference in character between the two curves. Enough that, while it doesn’t appear to be fatigue, it certainly implies something.
Or does it? What if it’s a data artefact?
We are comparing a fairly large data set (all the games played outside of February) with a small one. Maybe it’s just that the small dataset is inherently more volatile.
So let’s look at a comparison dataset. It was suggested to me that Matt Hendricks, who plays primarily as defensive third/fourth line guy facing mostly similar competition through the season, might make a nice comparison. So here’s exactly the same pair of lines for Matt Hendricks:
Certainly, you can see a difference between the two, as you’d expect. February is more volatile, indicative of a smaller sample size.
The overall character of the two curves, though, is similar. In other words, over the course of the game, Matt Hendricks has been pretty much the same player in February as he was prior to February.
And for our purposes, that’s good. That at least provides some reason to believe that the difference we’re seeing with Draisaitl is a real one.
So the evidence doesn’t indicate fatigue. So what could the issue be?
I think Drai is injured. Not enough to keep him out, but enough to impair. He took some vicious (uncalled … naturally) cross checks and a few shot blocks in January that visibly hurt him. That would be consistent with the overall reduction in his game, and also what appears to be a recent improvement in March (7 games in 11 days doesn’t seem like the ideal circumstances to address a fatigue issue!)
Or maybe, just maybe, it’s a garden variety slump! Those happen too.
Either way – I’d say it’s unlikely it was ‘wearing down’. It’s not definitive, certainly. No single study is, especially a relatively casual one like this.
But I feel comfortable in saying that this analysis suggests we should look elsewhere.
What say you?
Post Script and Bio
There you have it – my first article for BLH. I hope you enjoyed it. My plan is to average about one a week. Some lighter, some heavier, but hopefully always at least somewhat interesting!
A quick word about me for those not yet asleep: I’ve been an Oilers fan for uncountably many years – including the glory years. Maybe that’s what has kept me going, even as I’ve spent so much time living in the land of the bovine enemy.
I’m unquestionably a numbers guy, but just FYI, I “watched the game” for about 35 years before I ever got into the whole fancystats thing. I do this stuff as a hobby, but when I’m not watching hockey or talking hockey online (my wife calls them “my imaginary friends”) I’ve got the aforementioned wife, plus three kids and a cabin in the mountains to keep me hopping.
P.S. I’m a huge Rush fan, so a lot of my posts will have a Rush lyric easter egg embedded – though today I went with Star Trek.