It's true that all Bruins defensemen have strong expected goal numbers, with their top 6 players' on ice XPG percentages ranging between 55% and 62%. But Carlo's XPG% numbers are best both on the team and in the league when looking at defensemen with over 300 5 on 5 minutes (194 players). The fact that Carlo has done this with the lowest OZ start percentage among all Bruins defensemen by a considerable margin, and in fact the 7th lowest OZS% in the NHL among those 194 defensemen, is incredible. In the past I found a solid relationship between OZ starts and both expected goals and actual goals on ice (and individual expected goals and goals, as well), such that the lower a player's OZ starts, the lower their expected goal and goal percentage. I'd have to look back, which I'm not going to do while the waves are cranking at Quimixto, but I cannot remember any player either forward or defensemen who was top 10 in XPG% and bottom 10 in OZS%.
Carlo has played with Coyle, Pastrnak, Zacha, Hall, Marchand, and Foligno most of all team forwards (over 150 even strength minutes each), so he hasn't been getting mostly first or 2nd line pairings (it's actually been pretty even across the board). Interestingly, his expected goal percentage numbers on ice without these players, and really this is true of all team forwards he has played with, has been consistently very good, ranging from 56% to 66%.
The same is true looking at defensemen. Carlo has played most with Lindholm and Grzelcyk (both over 60% XPG%), but his without XPG% numbers for all Bruins defensemen range between 61% and 63%. No one is "carrying" Carlo this year that I can see statistically.
Finally, re: the bolded, great call. I can't share too much here, but he's been excellent on recovering opposition dump in's and getting them out of his own end.
I trust your eyes very much, which is why I looked at Carlo's data in the first place. But the data I'm seeing thus far this season does say Carlo is a far different player than the past few years, and is a solid asset to a very good team.
One last thought. The publicly available expected goal data is wonderful, and it's getting better all the time. But the expected goal data I have access to has an exponentially higher relationship to actual goals than anything out there that I have seen. They track far more actions than the public systems, and are far more sophisticated and (IMO) accurate. I bring this up because Carlo's Corsi, Fenwick, Natural Stat Trick, and Hockey Reference numbers are very different than what I have. They show a well above average player, but not someone at the top of the league. I think the data I have is more accurate, but half season sample size can be an issue. I guess what I'm saying is, don't discount the eye tests entirely.