{"id":27175,"date":"2005-02-25T13:23:50","date_gmt":"2005-02-25T19:23:50","guid":{"rendered":"http:\/\/www.uscho.com\/2005\/02\/25\/between-the-lines-feb-25-2005\/"},"modified":"2010-08-17T19:56:10","modified_gmt":"2010-08-18T00:56:10","slug":"between-the-lines-feb-25-2005","status":"publish","type":"post","link":"https:\/\/www.uscho.com\/2005\/02\/25\/between-the-lines-feb-25-2005\/","title":{"rendered":"Between the Lines: Feb. 25, 2005"},"content":{"rendered":"
The surest sign that there’s six more weeks of winter is not when the groundhog comes out of the ground and sees his shadow, it’s when the Pairwise ignorant start coming out of the woodwork to complain about the selection process.<\/p>\n
We all must admit that things are far better than they used to be. Specifically, at least it’s not the coaches themselves who are complaining anymore. It seems as though the hockey community at large is completely up to speed and on board with the Pairwise Rankings. It’s a beautiful thing.<\/p>\n
But there are still those out there who have no idea how it works.<\/p>\n
There is a lot of room to criticize the Pairwise Rankings and the selection process. Believe me, I’ve done it plenty. But if you criticize it, you should do so from a position of knowledge, and debate the right points.<\/p>\n
The Pairwise is still a good system. It’s objective, which doesn’t necessarily make it better than a subjective committee-picked system, but it does make it transparent — everyone knows, generally, what they have to do to make the tournament and where they stand. So that makes it better right off the bat.<\/p>\n
And, for example, the rankings of Cornell and Minnesota — two that are regularly criticized — are completely appropriate. Cornell is highly ranked (6) because of the best winning percentage in the country, with a mediocre strength of schedule. Minnesota is there (4) despite playing mediocre for over a month, but with the best strength of schedule in the country.<\/p>\n
However, this year, when you look at the Pairwise, you do see some things that stand out — such as Wisconsin being so low at around 11 (with the RPI “good-win” bonus factored in) and Dartmouth around 14 and in line for a spot. And these anomalies point out some of the flaws in the Pairwise that make it less than perfect system.<\/p>\n
The whole idea of Pairwise Rankings and other ranking systems is to adjust a team’s record for their strength of schedule, since everyone doesn’t play the same slate. The question is, how effectively does it do this?<\/p>\n
The Pairwise has RPI (Ratings Percentage Index) as a component, which factors in the records of opponents and opponent’s opponents. The Pairwise further takes into account records against good teams (Teams Under Consideration), and record against common opponents. It compares every team against every other, and ranks the teams by how many of those comparisons are won.<\/p>\n
The problems are multi-faceted, but lie most specifically in two areas.<\/p>\n
Teams Under Consideration are defined as any team with an RPI of .500 or better. As a result, the TUC criteria is won or lost depending on which opponents qualify as TUCs. The problem here is that not all TUCs are created equal. Some are barely a TUC, some are Top 5 teams. Also, the cutoff is arbitrary. If a team has a .500 RPI, then games against that team count in your TUC criteria. If that team has a .4999 RPI, games don’t count. When you consider RPI is arbitrary to begin with, it compounds the problem.<\/p>\n
The problem with RPI is legendary. Essentially, the RPI was concocted without any basis in meaningful mathematics. The fact that it does an OK job is more a case of never-ending tinkering than anything real.<\/p>\n
As a result, the Pairwise is prone to anomolies. Again, this usually comes out in the wash in the end, but this year, the Wisconsin and Dartmouth issues stand out. And if you look closely at the criteria between those two teams in particular — which Dartmouth wins — you’ll see why.<\/p>\n
If two teams didn’t play that season head-to-head, you’re basically down to three criteria. Wisconsin will beat Dartmouth on the RPI. But on common opponents, the difference comes down to Wisconsin tying Yale, while Dartmouth beat them twice. That criteria goes to Dartmouth. Not much to go on. Then the TUC criteria: Dartmouth barely wins. Why? Because Dartmouth’s strength of schedule within the TUC criteria is not as strong as Wisconsin’s. Plus, Wisconsin loses some wins from its record vs. TUC criteria every time Alaska-Anchorage bounces out as a TUC. If UAA pops back in, Wisconsin shoots upward.<\/p>\n
You can see how this arbitrary TUC level is wreaking havoc with with Wisconsin, and explains why they keep bouncing up and down.<\/p>\n
The point is that Dartmouth benefits from some PWR quirks while Wisconsin gets hurt.<\/p>\n