I had kind of avoided looking too deeply at the Pairwise Numbers, because things can change drastically with a couple results, but in looking over the final numbers, there's plenty of room for outrage. Two things stick out to me: Alaska barely avoided missing the tournament because of the TUC cliff, and in my opinion Minnesota-Duluth got completely screwed out of a tournament bid.
First up is Alaska, since that's a fairly simple one. Alaska would have lost a comparison against Massachusetts, for some dubious reasons, if UMass had been a team under consideration, . One fewer comparison win would have dropped Alaska to 10 total comparison wins and into a tie with Ferris State and Michigan State for the last at-large bid. Ferris State had the highest RPI of the three and likely would have gotten the bid. As luck would have it for the Nanooks, UMass finished 26th in the RPI instead of 25th. Not that any of this was Alaska's fault. It would have been a joke if they missed out on the tournament because of that ridiculous comparison.
That brings us to Minnesota-Duluth. I think they're the real victim here. Adam Wodon probably does a better job of explaining this, in that old article about the travesty that kept Minnesota State out of the 2008 tournament, but the same situation applies here. Basically, the history is that about 8-10 years ago, the committee didn't necessarily look at the standings of who won the most comparisons like they currently do, so much as they just looked at the group of teams that was actually under consideration for the tournament and how they did against each other. If that were still the case, I think Minnesota-Duluth makes the tournament.
Duluth finished the year 12th in the RPI, but 17th in the Pairwise because they lost three comparisons to teams that were under consideration for the tournament in name only: RPI #20 Maine, RPI #23 UMass-Lowell, and RPI #25 Boston University. Ignore those three comparisons and Duluth is a solid three seed in the tournament. Maine is really the only one of the three that you could maybe stretch it to make the argument that they belong under consideration for the tournament, but even that still would have gotten Duluth in.
I understand that the PWR isn't going to be perfect top to bottom with the best team winning every comparison on down to the last team winning none. If it exactly replicated the RPI every time, there would be no point to it. But there's a difference between slightly tweaking the RPI, which measures an entire body of work, and making it altogether meaningless.
Looking at each of the comparisons, I still don't have a big problem with the UMD/Maine comparison. They're fairly close in the RPI--UMD has about a .01 advantage--but Maine is a game over .500 against TUCs while UMD is three games below .500, and Maine leads in common opponents, though UMD has 8 games against common opponents to Maine's four. But the last two comparisons against Lowell and BU, the ones that keep UMD out of the tournament, are nearly identical and they are laughable. UMD has a fairly significant advantage of over .015 over both teams. But that advantage gets erased because both Lowell and BU are 7-9-3 against TUCs(.4474) while Duluth is 11-14-1(.4424). That's so incredibly close and it basically cancels Duluth big RPI advantage.
That means Duluth's tournament life is decided by their record against common opponents with those schools. Vermont is the only school that played Duluth and either of those two teams. Vermont beat UMD in the one game they played on Vermont's home ice. That zero percent winning percentage means UMD had to hope those other Hockey East teams went 0-for-the-season against Vermont--an unlikely prospect when you get multiple cracks at a team. Lowell ended the year 1-0-2 against Vermont. BU went the absolute minimum and won the comparison with an 0-2-1 record against Vermont. I'd like to think that with two more cracks at Vermont, one of which on home ice, UMD could also muster at least one tie. I doubt you could find anyone to say that there's enough data there to draw any sort of reasonable conclusion.
It's difficult to find any sort of perfect number system, but there's just so many holes in this current version and things that make no sense, and it's frustrating to watch teams get punished because of it.