Bye Bye RPI? New Formula Could Soon Replace Unpopular Index

LATE SUNDAY NIGHT, the 18-team NCAA Division I men’s lacrosse tournament bracket will be revealed. Inevitably, when the seeding and selections are explained, the NCAA’s sorting tool of choice — the Ratings Percentage Index, or RPI — will be invoked.

The RPI ranking for teams. Strength of schedule based on each school’s top 10 opponents’ RPI. The performance of teams against the top five or top 10 or top 20 of … yup, you guessed it, the RPI.

Nothing new here. Same as it ever was.

But now for the twist: There’s a chance it’s the last time anyone at the college game’s highest level will have to fret about those three letters again, barring an unexpected leap by Rensselaer Polytechnic Institute from D-III.

Over the last six years, Virginia coach Lars Tiffany has led a push to discover and develop a different way — a better way — to go about the selection process. A fortuitously timed overhaul of the NCAA governance structure has created the opportunity for it to happen.

“It’s going that path,” Tiffany said last month. “It’s a little scary. There’s a lot of excitement. This thing could be in place a year from now.”

It feels like an overdue change. The RPI was first utilized in men’s basketball in 1981, and not shockingly for a metric of that vintage, it isn’t exactly at the cutting edge of modern analytics. Its composition is 25 percent of a team’s winning percentage, 50 percent of its opponents’ winning percentage and 25 percent of its opponents’ opponents’ winning percentage.

By this time a decade ago, the RPI had become so unpopular in basketball circles that a sport with roughly twice as many games per team as lacrosse looked to retool its selection system. The NET (the rather drably dubbed NCAA Evaluation Tool) debuted in 2018-19.

It wasn’t long afterward that Tiffany started pondering what it would take to create a new selection process in men’s lacrosse as well, one that might not be so easily gamed by simply loading up on quality opponents (thus boosting the RPI’s biggest variable) and aiming to finish at .500 or better to become tournament-eligible.

“I started asking questions. Why are we still doing it?” he said. “If this was designed for sports that played more games than ours, then why we are the last ones to hold onto it?”

THE IMPETUS TO TIFFANY was the case of 2019 High Point, which beat two teams ranked in the top four of the RPI (No. 2 Virginia and No. 4 Duke) and went on to finish 13-3. But the Panthers lost to Richmond in the Southern Conference title game, and more notably, dropped games to No. 41 Jacksonville and No. 58 St. John's and finished 20th in the RPI.

Then-Brown athletic director Jack Hayes, the men’s lacrosse committee chair that year, noted the two high-profile victories hours after the bracket was unveiled, but also acknowledged the Panthers’ losses and mid-pack strength of schedule. “When you look at a resume that looks very different like that,” Hayes said that night, “It’s hard to move them forward.”

It did not sit well with Tiffany, whose team had lost to High Point that February and went on to win a national title.

“They’re left home with the justification being ‘Well, they had some bad losses,’” Tiffany said. “It just never satiated me, the explanation. I just was like, ‘Is there a better way? Is there a better mousetrap?’ I formed a committee that wasn’t being asked to be formed and found some people who are interested in math.”

Tiffany reserved a hotel room at the IMLCA convention in Baltimore that December and convened a discussion. This committee, with a membership that shifted over time, would meet over Zoom as many as 18 times in a calendar year, batting ideas around and trying to come up with a model that both was fair and took into account more than the antiquated RPI.

A pivot point of the work came when Notre Dame assistant Chris Wojcik, then and now a committee member, reached out to Larry Feldman. A college lacrosse player at Penn in the 1960s, Feldman’s career also included time as a software architect at Intel and featured an extensive background in mathematics and engineering.

More pertinently, he founded LaxPower and ran the site from 1997 to 2019.

“In today’s day and age, I’d call him an AI assistant,” Tiffany said. “Just type it in and there was an answer five minutes later. Somebody was like, ‘Yeah, I can run the model.’ Can you tweak the home field advantage? ‘Sure.’ Hey, what would it look like if we gave the power ratings a little more juice and give this a little less juice? Now we had a numbers guy. That transformed everything.”

It was the perfect match. Now retired, Feldman had the time — thousands of hours over the last few years, in fact — to massage various formulae to see what might work best. And there was no questioning his passion for the sport.

But there was a bit of an uh-oh moment for Feldman, who emphasized he is a mathematical consultant and does not speak for the committee.

“I had spent 20-25 years with LaxPower evaluating the RPI and the other methods and predicting who would get in the tournament,” Feldman said. “I have no trouble writing algorithms. However, sometimes you bite off more than you can chew, and I remember one night I went to sleep and I woke up in the middle of the night, and I said, ‘Well, now that you’re a part of the committee, you’re going to have to come up with a method that’s better than other methods.’”

FELDMAN MIGHT HAVE FELT that urgency from the start, but realistically, the window for change really opened last summer when the NCAA reduced its committee structure, in the process introducing committees that receive greater authority to create rules for their specific sports. That includes the ability to determine playing rules and how championships are administered.

It also meant fewer layers to get through, especially after coaches agreed in December to recommend the adoption of a system called Powerwise. The men’s lacrosse oversight committee is considering implementing the proposal this summer.

As for what it does, it borrows heavily from the concepts of college hockey’s Pairwise system. Every team is compared to every other team in Division I, with a maximum rating of the total of Division I teams minus itself. (This year, there are 77 D-I teams, so the highest achievable number is 76.)

If one team owns a head-to-head advantage, it gets the point. If two teams didn’t play or split two meetings, the process moves to record against common opponents.

If neither on-field facet determines the point, it would go to a power rating system that factors in game locations and scores (with no advantages offered beyond a seven-goal margin, which Tiffany said was the average of the cap used in conference tournament tiebreakers).

Should there be a tie in the overall rankings — hypothetically, two teams at 65 points — the same criteria would be used to break that logjam and determine an order.

“It’s not about mathematics, it’s common sense,” Feldman said. “It’s about head-to-head, common opponents and your performance on the field. But after the method was developed, Chris [Wojcik] pushed the fact that 63 percent every year is based on on-field performance. I just wanted a simple method that people could understand. How good is a method if people don’t really understand it and it turns out to be a black box?”

There is one significant divergence from the hockey system, which weighs head-to-head performance equally with performance against common opponents. If two teams split in those two categories, a power rating determines the point.

In the Powerwise system, a winning head-to-head record ends the conversation for the point. For example, Towson would claim the point from Virginia this year by virtue of a 13-9 victory in Charlottesville on March 7, regardless of any strength of schedule differences creates by their respective leagues.

“Let’s say Virginia gets the common opponent point win; all right, there’s a tie, let’s go to power rankings,” Tiffany said. “Now all the sudden Virginia is in? We didn’t like that. To me that’s where you get the pitchforks and torches. And we feel adamant about it. I couldn’t imagine saying, ‘Yeah, it was a nice win Towson, but Virginia played a better schedule and has a better power rating. Sorry.’”

It is a system that Feldman said stresses simplicity, transparency and objectivity, as opposed to the annual attempt to try to divine what RPI-related factor will ultimately tip that year’s NCAA men’s lacrosse committee in one direction or another as it scrutinizes the teams contending for the last few at-large slots.

Powerwise would handle much of it. It would seed the top eight in the field, and it would be used to select the at-large teams. It wouldn’t be guaranteed to be the same, either; a year ago, Army would have claimed an at-large berth at Harvard’s expense under the Powerwise system.

“There are no other sports that rely on RPI the way we do,” Black Knights coach Joe Alberici told USA Lacrosse Magazine’s Matt DaSilva in January. “And we are the least conducive to using it.”

Army missed the NCAA tournament the last two years despite compiling a combined 23-5 record, including a 10-0 non-conference mark that included wins over traditional powers Syracuse, North Carolina and Yale. Playing in the Patriot League suppressed its RPI. Losing in the Patriot League tournament all but sealed its fate.

Alberici noted one instance last year in which Army dropped nine spots in the RPI following a league win.

“There’s a real flaw in our system,” Alberici said. “Coaches have unanimously agreed to a new system that doesn’t penalize you for winning.”

Comparing this year’s teams according to the two models, the greatest discrepancies as of Wednesday are Richmond (RPI: 4, PW: 1), Johns Hopkins (RPI: 7, PW: 10) and Tiffany's team, Virginia (RPI: 9, PW: 14).

“I tip my hat to him because Virginia is one of those places that benefits from the way it is now,” Alberici said.

The thing that would keep the committee the busiest on selection weekend wouldn’t be choosing the at-large teams and seeding the field. Instead, it would be making the first-round pairings to avoid conference rematches and account for the NCAA’s travel restrictions as best as possible.

“The lightbulb goes on in my head: ‘Oh my God, we have to convince people who are now on the selection committee that they no longer have a say,’” Tiffany said. “We are proposing a model that does the work for them. I wonder how they’re going to take that.”

THE PROPOSED FUTURE of college lacrosse postseason selection won’t feature flying cars (though, come to think of it, flying cars might make the NCAA more likely to drop the guideline limiting just two teams to first-round flights). But what would a brave new world look like?

There would be a website updated with results every day, so everyone could track how close their favorite team was to making the tournament from the start of the season. Every game would count in the strength of schedule, not just a team’s top 10 opponents.

Will there be complaining? That’s probably unavoidable, especially a system that tries to differentiate teams playing about 15 games a year. But the rules will be clearly delineated before the season even begins.

“If you came up with the perfect method, based on the work of a lot of mathematicians, the problem is there’s not sufficient data to feed this method,” Feldman said. “A method’s accuracy is dependent not only on the architecture and the algorithms used, but the data you’re feeding it. There just simply is not enough data for any method to work very efficiently to separate the bubble teams.”

Should the sport oversight committee vote to adopt the Powerwise system, one thing definitely would happen: The RPI’s days as a pinata in the men’s lacrosse would come to an end.

And after all these years, it’s probably about time to say goodbye to a metric with easy-to-exploit flaws.

“I think there will be quite a few people who say, ‘Glad that’s done,’” Tiffany said. “Twenty-five percent is our opponents’ opponents record? Why is 25 percent my record and why is 25 percent my opponents’ opponents’ record? Having said all of that, if this dinosaur goes extinct, we’ll say it had a long life and it actually provided a good service. It didn’t get it wrong that often.”