Quoting: OilednGreasy
I don't think relative stats should be used when comparing a player from the 30th place team to one from the 6th. The good player on bad team is a bad player on a good team kinda thing.
I'm of the mind that they need both a Janmark and Brown and that neither is easily replaced with a league minimum player (it's more likely to be possible with Janmark, but then you're only saving 225k in cap space).
Yet you're content to compare two players in such situations using rate metrics? Assuming that the player's role is controlled for (I have), I'd argue that rate metrics are more influenced by the quality of the team around a player than relative metrics would be.
If, for example, Murphy manages to keep his head above water (~0 or better in relGA, relGF%, relxGF, relHDCA, etc) in his shutdown role on an atrocious CHI team, then that tells me that, despite most factors being against him, he still played his role effectively. Lehkonen's final year in MTL is a great example of relative metrics revealing a valuable player stuck on a bad team. In comparison, if I read the page you sent, I simply see the rates at which various things happened while Murhpy and Ceci were on the ice without any reflection of what happened on their teams without them- therefore, no baseline by which to compare them. You're just comparing their numbers across two extremely different contexts.
Given the fact that both players played a similar role (primary shutdown RD for ~17mins/night at EV), a comparison using relative metrics is more appropriate than rates because what matters to me is what the player has done in comparison to what his fellow Dmen have done in the same system/circumstances (while factoring in significant differences in deployment).
The reasoning surrounding "good player on a bad team" confusing the data never quite landed with me. Sure, a good player's relative numbers can be inflated when playing on a bad team but it also means he's holding up his end of the bargain despite inoptimal conditions. If this were a sheltered, 3rd pairing Dman (ala Matt Benning in EDM) or a bottom-6 F whose line meshed well then the argument holds more water but such inflation isn't something that commonly occurs in the kind of role that Murphy played. If anything, his role in the one that ought to be among the most negatively influenced by a team-relative comparison.
In my opinion, the best argument against the use of relative metrics in this particular context is that the nature of the shutdown role on a bad team is somewhat different from such a role on a good team (re: expectation of offensive contribution & strictness/rigidness of role). However, we could discuss that for ages and not come to anything conclusive, so I'll refrain from that topic.