Generally, rules changes in college hockey take place in even year summers, which made it somewhat of a surprise this week when it was mentioned at the AHCA coaching convention in Naples, FL, with NCAA President Charlie Baker, Northeastern AD Jim Madigan, and Hockey East Commissioner Steve Metcalf among the attendees, that the Pairwise was going away. The Pairwise has long been held as the gold standard in college hockey, essentially the highest publicity NCAA team sport that selects teams for its tournament based solely on mathematical record rather than by committee vibes. So what is the replacement NCAA Percentage Index (NPI) and how will this impact things going forward?
First thing is first, this isn’t a shock. The NCAA D-III Championships Committee recommended using NPI to select participants in team sports last spring, the NCAA began implementing it last fall, and hockey is no exception to the rule, with their mens’s and women’s tournaments both selected using NPI this past spring. The Women’s D-I tournament has already been using NPI as a part of their Pairwise in place of RPI after a trial period, in addition to other sports and divisions who have already started using it. So this has been coming down the pipeline for a while now, and it had already been considered for men’s hockey in previous offseasons, but ultimately pushed back until now. Proponents of NPI believe that RPI was an arbitrary way of ranking teams while NPI depends on what happened on the ice.
The NPI is, at a base level, incredibly similar to RPI and to the old system. There are many sports where the addition of an objective measure is a shock to the system or where the objective measure is a departure from the prior measures. This is not one of them. Tim Danehy, inventor of the Pairwise, longtime member of the college hockey community, and the godfather of the beloved and defunct College Hockey Stats website, worked with the NCAA to create NPI. And the end result is a system that bears a lot of similarities to the one he had already created. The factors that go into NPI are:
- Winning percentage. In RPI, this value was worth 25% of a teams calculation. In default NPI (we’ll circle back to what default NPI means) it’s also worth 25%.
- Strength of scheule. In RPI, strength of schedule was a bane of contention. The remaining 75% of a team’s record was based on 24% strength of opponents and 51% strength of those opponents’ opponents. In default NPI, this is simplified. Strength of schedule is just the NPI rating of the teams you played against, averaged together. And that’s it, worth all 75%, replacing both schedule components of RPI.
- Home/Road and Overtime Weight. You know and love this from Pairwise, and it’s not going anywhere. Currently, a home loss/road win is counted as 1.2 games in the Pairwise, and a home win/road loss is 0.8. Any result that occurs in the 3v3 portion of a game is counted as 2/3 of a win and 1/3 of a loss, while a shootout is a tie. These factors all will still exist in NPI, but the exact values might change, as they have many times in the past.
- Quality Win Bonus. This was once again a part of the Pairwise already and is coming back, but in a new way. Previously, a team received bonus RPI for beating a strong opponent, based on team RPI rank. A win over the #1 team in the country received the maximum credit, while beating a team in 20th was worth 1/20 of that credit. Now, the credit will be based on the team’s NPI and how far it is above the baseline for a good team, not their rank. So beating the best team will still give the best credit, while beating any team with a “good” NPI will still give some extra credit. This also allows the system to be more granular, it’s based on your strength, not your rank. There could be 40 “strong” teams if the results demand it, not a hard lock arbitrarily at 20/64 teams.
- Minimum Countable Wins. This one is new and… you might consider it exciting? It’s really not. You might be familiar with the RPI concept of dropping bad wins. Essentially, if NU plays Stonehill, you expect them to win that game. The math also expects them to win that game. It expects it so much that even if NU does win the game, Stonehill might bring down their strength of schedule so much that they leave the game with a lower RPI than if they hadn’t played at all. That concept still exists with NPI, but it also happens in reverse (Stonehill could lose 20-0 to NU and leave the game with a higher NPI than before.) So to fix that, they just throw out any result that causes either of those things to happen. But at the end of the season, we want to make sure that every team has enough samples going into the math, so if a team has too many wins not counting, some might get added back into the math. This will almost definitely never come up, if they even choose to bother using it at all, and if it does someone from Mankato, Minnesota will be mad online about it.
And that’s it. It sounds a lot like the old RPI and Pairwise because it is. The hockey model has been refined a bit and is coming out of our little corner of the NCAA and is spreading to the entire organization. One day, when NPI is just a standard way of doing things, college hockey will be the reason why. Pretty cool. Now, let’s talk about the implications.
I said we’d get back to default NPI, and we are. The reason why I called it default NPI is because the sport committee has the power to change the weight of every factor when they implement it. They think strength of schedule should be worth an extra 5%? They can do it. OT wins should be 100%? Sure thing. A quality win should actually be worth double? You bet. The world is their oyster, and reports are that the coaches were presented with a number of different NPI calculations of the 2024-2025 season results, each based on different values being given to each factor. We still don’t know yet what values the committee may decide to change before the season. But at a base level, if they use the same factors that existed in the Pairwise last year, most teams fall in the same spot or within 1 spot of where they were in Pairwise. Personally, I would bet that most things will stay with the same weight they had before, except home/road weight may get a bit closer to baseline. The 20% adjustment has always seemed pretty high. They’ll also have to decide on values for the new Quality Win Bonus and Minimum Countable Win calculations (which they could just remove by setting the minimum to 0), but there are default values from Danehy to guide them.
Besides RPI, the other thing going away is the “extra” components of the Pairwise, head to head and record vs. common opponents. This isn’t much of a change. These things still matter, they just matter in that they factor directly into NPI and don’t need to be counted again. These were honestly archaic parts of the Pairwise that served a purpose 20+ years ago but almost never really mattered or changed any team rankings in a meaningful way in recent memory. They mostly served to be a thorn that occasionally caused a result that was one spot different than a straight RPI ranking, and even that one spot change happening was pretty rare.
So last but not least, what does this mean for NU? It depends on the factors that the committee ultimately goes with, but at the default values, it looks like a slight disadvantage for NU compared to the current system. The simplification of strength of schedule makes record matter just a little more, so the teams at the top of “bad” conferences like Bentley and Minnesota State will get a bump while teams that finish mid-pack in Hockey East, the Big Ten, or the NCHC will get knocked down a bit. But again, these are relatively minor differences that may result in a one spot change and don’t seem to matter… until suddenly at the end of the season someone is the first team out instead of the last team in by a decimal place.
If you want to see a few examples of the math, head on over to BC Interruption, where Grant Salzano broke it down as well and did some base calculations with the factors.