I think Pionk has been head and shoulders the Jets best defender... I also think the lack of skill and performance he's comparable to on the team is making him qualitatively by us and quantitatively by some models look better than he would if he wasn't the only guy playing even to average level.
I like EW's GAR and xGAR. I use it all the time here, and I think it's best when you blend them with xGAR having slight more weighting (from my experience and testing). But, it's good to know models blind spots when using them too, as I spoke about a bit in a recent Hockey-Graphs post.
The two issues I have with GAR and xGAR in the short season come from sample size with using 0 as the season start prior instead of past seasons as at least some of the equation and using only performance throughout the season as determiner of blame/credit of good/bad events.
1) Small sample size is worse when you have a model that refreshes each season. There's pros and cons to that with tradeoffs but one of that is that it makes it volatile and whatnot, especially GAR that's influenced in part by PDO. That's because the twins wanted it to be something that says what you've done, not what you will do and who you are. That's fine and 100% useful, but it's not great and answering the latter two questions.
2) With it only using performance of that season as info (plus environmental impacts) it doesn't know to bias itself for forwards or defenders in the right situations. We know from studies that forwards drive offense more and defenders drive defense more, but this model methodology will take a while to get that correct, ESPECIALLY in situations where you have a team with very different levels in quality and depth of forwards vs defenders (both when F>D and D>F).
An additional problem it means comparing players outside of each bubble is useless. You don't have crossover, so a 2GAR in the North Division != 2GAR in the East, etc.
All this to say I think he's legit been the best defender on the Jets. 100% and the models show that. I'm cautious to it saying he's actually a legit #1D calibre defender, especially when a model that is stronger at those specific problems I had above paints him as middle of the road (but still doing better than normal):