It’s inherently flawed to argue that what someone saw cannot be right because a summary statistic suggests otherwise. In the big picture, I hypothesize this is exactly why professional clubs employ both conventional scouting and statistical-based scouting approaches.
An interesting exercise would be to ask the posters here after each game to score their overall opinion of a player’s game on a scale of 1-10, and then plot that against the players GAR/xGAR. If you did this for all 82 games in a season, would you expect to see an R^2 of 1.0? I doubt it.
Are there any situations a player could have good moments in a game but have a poor GAR and xGAR score? I ask that from a genuine curiosity, because I respect you as a poster, and suspect you’re knowledgeable enough in this area that you can probably provide an interesting opinion.
From the 20 minutes of research I have done, it seems that GAR/xGAR, while being highly regarded as a valuable statistic, still has multiple limitations. In my research, there were two limitations that stood out to me. First, the model has a difficult time measuring the impact linemates have on the score if those linemates spend almost all of their time together. Second, the score is heavily reliant on plays that actually happen in a game, versus what should have happened, and this is particularly true for passing plays.
I highlight those two limitations because the majority of the comments I have read from the eye test crowd have focused on two things: (a) he is making passes to his linemates that they are failing to convert on, and (b) his entire line, which has played together since Dach went down, is struggling right now for reasons that are beyond Slafkovsky alone. Interestingly, both of those critiques would not be captured well by GAR/xGAR.
Bottom line, is I think this reenforces a part of the opinion in my post you quoted. Slafkovsky should be moved onto a line with different players (Monahan in this writers opinion) and be given a period of time to see if he looks better AND has better results. If that’s not happening after 10 games, then considering the entire body of work over the last 1.5 years, it’s time to have a rethink of the development approach.
