New Pairwise System Worked Well
by Adam Wodon/Managing Editor
A few days removed from the NCAA tournament selectionsm it's interesting to note how different it would've been under last year's Pairwise formula.
This year, the Ice Hockey Committee removed the "Record vs. Teams Under Consideration" criteria from the selection process. It was replaced by the "Quality Win Bonus," a different way of handling "good wins." The bonus gave teams extra percentage points on its RPI based on which teams they defeated in the Top 20, on a sliding scale.
Also added to the Pairwise this year was a home-road weighting system.
Looking at it, it's obvious to see that the teams which beat more top teams, and the teams that did better on the road than at home, are the teams that benefitted most. This stood to reason logically, and certainly bears itself out in the numbers.
The order of teams based on last year's Pairwise:
St. Cloud State
The only team that failed to make the NCAAs that would've made it otherwise is Cornell; which is a bit ironic given that, much of the push to include a home-road weighting was from ECAC teams that have a hard time scheduling non-conference games at their arenas. Cornell's issue wasn't that, however; it had very few "good wins" and its QWB was low compared to other teams.
Michigan still would've been out, and Vermont and Cornell would've swapped.
The biggest drops are from Wisconsin and Notre Dame. In Wisconsin's case, it won many more home games than road. In Notre Dame's case, it benefitted this year from a huge QWB. Providence, North Dakota and Lowell would've been better off.
The biggest issue, however, would've been the myriad of bracket problems the Committee would've faced with this order of teams. This year's bracket, just by luck, basically, broke down in such a way that no tinkering was necessary.
But nevermind the comparison to last year's Pairwise. The most interesting part, however, of the new Pairwise is how closely it matched KRACH.
KRACH is the rating system that is considered the most "mathematically pure," and has long been endorsed by CHN as a replacement for the RPI. KRACH is usually a much better indicator of a team's relative strength. The Pairwise has always been a patchwork of ideas. The two often match somewhat, but not that closely.
With the changes to the Pairwise, however, the ratings — however roundabout the algorithm may have been to come to that point — are almost directly in line with each other.
This means that fans and teams can feel very comfortable about the new Pairwise. No system with such a small sample size as 35-or-so games will ever be good enough to be perfectly precise. But this is as good as we can ask for.