Read on Twitter
So, Tony DeAngelo is on waivers, which is interesting to me because, one way or another, there is something about him and his teammates which is confusing my model so I have a small thread about it which I warn you, goes nowhere.
DeAngelo is one of those players who I don't post much about, because he's a terrible, terrible person and so everything I post is perceived either as glorifying a person who doesn't deserve it or inviting a pile-on about how he's a terrible person and neither one interests me.
That said, his model estimates are Extremely Interesting to me for technical reasons because they suggest something curious is going on and I don't know what it is. This is his career to-date in terms of my 5v5 xG model.
Rookie year in Arizona, just awful defensively, nothing strange there for a 21yo whether he's actually good or not. The next couple of years close to net-zero but skewing heavily to fun, very Rangers-y, no surprise. All of a sudden in 19-20, very strong offensively.
(The 20-21 estimate is naturally very close to that, he's only played six games so far this season so the estimate of his ability as of the end of last season can't change that much)
In terms of finishing impact his numbers were very ordinary, close to average, until 19-20, when suddenly they jump up to +6%, very, very strong. So, what happened in 2019-2020? Why the sudden change?
There's also the matter of how this season (and in the playoffs last year) he's looked, mostly, like shit in a sock. So if you want to take the opinion that he's "just bad" somehow, you have to find some other way to explain the rangers offensively with him in 019-2020 at 5v5.
The easiest way to fool a regression model of any stripe is to take two different things and staple them together so tightly that there's no way to tell them apart. If that's the case, then the biggest confounder for DeAngelo in 19-20 is ... Marc Staal.
The "most obvious" explanation for why maybe DeAngelo might not be as good as his model estimates suggest is that Marc Staal was carrying him. Out of the frying pan and into the fire, imo. Let's keep looking.
The next big confounder is Panarin, DeAngelo's second-most common 5v5 teammate in 2019-2020. However, even at that he's just playing 375 of his 1035 minutes at 5v5 with Panarin, just over a third of his minutes. So explanatory power here is limited to begin with.
That said, it does have some legs. DeAngelo with Panarin the rangers are FEARSOME offensively, generously score- and zone-deployed, outscore opponents 31-16. Dominant.
And DeAngelo /without/ Panarin is sub-par offensively and right some shitty defensively, contextually not as generous by zone or score, but still.
This can't be the /whole/ answer because my model is clever enough not to be fooled by something quite so simple. Perhaps one part of the puzzle is that DeAngelo plays (played?) to what few strengths Marc Staal had, and also had an unusual chemistry with Panarin.
Anyhow, this thread, as promised, goes nowhere, but whenever I see model estimates that "feel wrong" (to me) I spend a time thinking about why and what I might change in the summer to get a subtler, better bead on players/coaches/context. This one has been on my mind a while.
