A question posed by Ryan over at Lowetide … where are elite D drafted? Is it really as low as everyone thinks?
First – we have to answer: what is an elite D? Not that easy a question, actually.
Here’s a couple of lists using three years of lookback.
[All data sourced from Vollman and war-on-ice]
GP > 40, EV TOI > 18, EV Corsi > 49, EV CorsiRel > 0.
Weird part: if you add a QoC filter (CCorsi>50), guess who drops out?
Some of those are obvious (Campbell, Yandle). Some are shocking (Chara).
Another definition – simplify to GP > 40, ALL TOI > 23. I like this list, so this is where I’ve added the draft position overall (blank means undrafted):
That looks like a decent list. Here’s a breakdown (undrafted treated as #210):
There’s your answer, Ryan! Where are elite D drafted? Not often very high…
P.S. Whenever I do any work with draft data, it always comes back to the reality that you run into a classic filtering problem of signal vs noise.
Set the filter too loose, and you get noise in your results.
Set it too tight, and you filter out signal (which in this case would be dropping near-top D off the list).
And it’s almost impossible to find the right balance, because the signal and noise are so mixed.
My end response is to set the filter loose, and let my brain do the filtering.
The main conclusion I read off this list: you can get good D anywhere. But you can also get good D high in the draft (e.g. Hedman, Pietrangelo, Doughty, E Johnson, Bouwmeester).
What this chart doesn’t show, though, is comparative risk. It’s not just a matter of picking a D, remember that you are also forgoing another player, likely a F when you do that.
That’s where the risk aspect comes in – the data (I’ve crunched numbers previously) do show that drafting F high has a higher success rate.
So on the topic of Hanifin – not a mistake to draft him given where he’s rated and the roster issues of the Oilers, but must recognize there is some risk there. Drafting Hanifin comes down to confidence in scouting I guess. I sure do like the Pietrangelo and OEL comparisons.
Lastly, Ryan asked a really good question to ask: is scouting of D getting better? The data is pretty thin, but if I have time, maybe I’ll see if I can put a time lens over top of it that might indicate if results are improving over time.