The PairWise Rankings have been as familiar – and often controversial – as any topic across college hockey for the better part of three decades.
The term, first coined by three self-described “stat geeks” from the University of Minnesota (who went on to start USCHO back some 30 years ago), is going the way of the VHS tape this season as the NCAA Men’s Division I ice hockey committee changes how teams are selected for the postseason NCAA championship.
The NCAA Power Index, known simply as the NPI, will replace the PairWise Rankings as the selection process for the 2026 championship next March.
It is a move that has been a long time coming, one that coaches have called on to change for a number of years in hopes of finding some sort of ranking system that fairly and evenly quantifies teams to assemble the NCAA tournament field.
So what is different and how much of a difference will this new system make?
When you come down to it, things may seem similar.
USCHO recently spoke with Tim Danehy, known better in the college hockey universe as the person who once operated the well-liked website collegehockeystats.net for almost two decades. He has often been called up to put his highly-talented mathematical brain to use in a consulting role to the NCAA and ice hockey, particularly when it comes to how teams are selected for the NCAA tournament.
Danehy does acknowledge that the NPI might not produce a different 16-team field come March than the PairWise might. But he says one of the most important characteristics of the NPI is its ability to eliminate some statistical anomalies that hampered the PairWise.
As an example, Danehy points to last season and the comparison of Ohio State and Princeton. The Buckeyes finished the season ranked 10th in the PairWise. The team’s Ratings Percentage Index, or RPI (which measure your success against your opponents, taking into consideration your strength of schedule) was .5594, also ranked 10th.
Princeton finished the season 43rd in the PairWise, with an RPI of .4830. But the Tigers won twice at home against Ohio State in late November, both by the score of 3-1.
Now let’s look at the criteria of the PairWise Ranking to better understand how Princeton and Ohio State compare to one another.
There are three criteria that the PairWise has used for a number of years:
- RPI
- Common opponents
- Head-to-head
One point is awarded for each criteria won, understanding that each head-to-head victory earns a criteria point. If there is a tie in total criteria won, the team with the best RPI wins the comparison.
Ohio State wins the RPI criteria soundly, but because Ohio State and Princeton had no common opponents, those two losses in November made it impossible for the Buckeyes to win the comparison against Princeton.
“The PairWise comparison says Princeton is better,” said Danehy, noting that almost any other measurement would have Ohio State ahead. “There’s an enormous RPI gap, but [in the PairWise Ranked] that doesn’t matter.”
It is statistical anomalies that the NPI looks to address, and it does this by selectively eliminating certain games that can negatively – and also positively – impact a team’s position.
In the NPI system, games that negatively impact a team’s ranking are not simply counted against the team, but are systematically evaluated and potentially removed. Here’s how it works:
1. Initially, all games are counted in the calculation
2. The system then identifies the lowest-value wins
3. Games below a certain threshold can be removed from the calculation
4. There’s a minimum wins requirement (12 wins) to prevent teams from eliminating their entire schedule
5. When a game is removed, it’s as if the game was never played – teams are not penalized, just eliminated
6. The removal process is dynamic and iterative, meaning:
– Each game’s removal affects other teams’ calculations
– The process repeats multiple times, thus stabilizing the rankings
The goal is to prevent teams from being artificially boosted or penalized by extremely weak or strong individual games while maintaining a representative season-long performance metric.
Better said, teams get rewarded for playing stronger teams.
“We immediately solve that disconnect,” Danehy said. “If you’re ranked higher than me, you’re worth more towards strength of schedule than me. I can’t help my strength of schedule more by playing somebody that’s ranked below.”
These changes should not, and in all likelihood will not, create a radical shift in the rankings. Instead, according to Danehy, it will produce a more defensible mathematical method.
And this system has already been tested. Division III has been using the NPI across all sports (in fact, here is the link to an excellent podcast explaining NPI in depth: https://www.youtube.com/watch?v=xrKp8McryUA). And women’s Division I hockey replaced the PairWise with the NPI last season.
With the season ready to begin, we are still a long way of worrying about which teams will and won’t make the NCAA tournament. But when you Google searches for “PairWise Rankings” next March yield little, just changed the search term to “NPI.”
