Yes, I've read his analysis a billion times. The preambles, the blurbs in the rankings, the blog post from 2009.
It hasn't helped me make heads or tales of his weighting of the billion different variables (based on what I hope is correlative analysis of some kind), or why the weights/important stats seem to change week to week, nor does it provide any insight into what he's assigning as arbitrary values/weights for things like divisional strength, nor does it explain cases where Team A takes a dive because their star is out ("team that is going on the ice that night") but Team B does not (despite otherwise similar stats).
It seems any time someone expresses confusion for a certain ranking of Team X, compared to another Team Y, there is one set of criteria explained as the difference-maker -- but when that same set of criteria fails to adequately explain a difference between Team Y and Team Z, we move on to a different set (which then fails to explain Team X and Team Y). Obviously he's got a formula (linear or otherwise) that gives him an output value, especially since he's using "just the stats" and therefore must have a way of mathematically determining divisional strength. Perhaps he's got something in place to determine an ELO-esque rating system for each team, assessing match outcomes night-by-night, building cumulatively throughout the season, allowing him to account for quality of competition and gauge divisional strengths -- perhaps he just looks at total points and adds some kind of multiplier to the end result. Either way, I don't know.
There are a lot of variables that need to be accounted for, so I'm just not sure how he's accounting for them and determining those weights -- obviously I can't do it at a glance by comparing two or three teams and comparing 'key' basic/advanced stats, because that's why things fall apart and lead to headscratching in the first place.
I just want more detail on how he's coming to them, if he feels they're valid. That's all. Until I see that, no amount of hand-waving will validate the output of whatever formula he's got in place. I might as well just roll dice week-to-week and slide people up/down 1-3 slots if all I'm getting is "my super sekrit method takes every stat, recent trend, and head-to-head performance factor into account".
Edit: Actually, that might be fun to do and compare...