I think the problem was both with the model and the interpretation. The model is static and cross-sectional and doesn't consider a longer trajectory. Similarly, the interpretation ignores a lot of the information based on previous performance and the current context. Put any #1D into a situation where the rest of the D are Pionk, Forbort, Poolman, Sbisa, Bitetto and Beaulieu and you end up getting swamped.
Recognizing that every model has important outliers and being able to identify them is an essential part of the analytical process. Consider another D; Skjei. The Canes traded a 1st round pick to acquire him from the Rangers, despite his dropping value. He bounced back and is now a core player. Clearly, they looked beyond the models per se and made a good decision to acquire and keep him.
View attachment 884377