R.I.P. to the R.P.I.: Selection Committee Breaks Out New Math

The Kansas Jayhawks entered the postseason with their wings clipped. Having lost two stars midseason — one to an N.C.A.A. suspension, one to unspecified “personal matters” — they are poised to sustain double-digit losses for only the second time in 19 seasons.

Yet according to the Rating Percentage Index statistic, which typically saturates bracketological prognostications ahead of the N.C.A.A. tournament’s Selection Sunday reveal, Kansas was the No. 1-ranked team in the country on Thursday — higher than Kentucky, which beat Kansas handily in January; higher than Virginia, which has zero losses to teams not named Duke; higher even than Duke and its basketball messiah, Zion Williamson.

So should you start penciling Kansas into one of the four No. 1 seeds before the N.C.A.A. tournament bracket is released Sunday night?

Not so fast. While you can still find websites that calculate R.P.I., officially the statistic is no more in the men’s game. The N.C.A.A., which created it nearly four decades ago, disowned it in the statistic’s most prominent sport last year.

“We as a committee have decided the R.P.I. is kind of yesterday’s news,” this season’s selection committee chairman, Stanford Athletic Director Bernard Muir, said last month.

The R.P.I.’s replacement as the first-among-equals rating, the N.C.A.A. Evaluation Tool, puts Kansas all the way down at No. 20. It is a shockingly large divergence that much more closely matches the No. 4 seed Kansas is generally expected to receive after the committee considers not only NET but other advanced stats, winning percentages, individual game results and even the eyeball test.

As Kansas shows, we are not in R.P.I.-land anymore. And it is difficult to find close observers of college basketball who do not think this is a good thing.

“By and large, eyeballing it, it has been better than the R.P.I. at measuring good versus bad teams,” the ESPN tournament analyst Joe Lunardi said of the NET.

More subtle but at least as important, moving from R.P.I. to NET — even the acronym seems an improvement! — may have heralded a shift in the selection committee’s philosophical underpinnings. (The selection committee for the Division I women’s tournament will still use the R.P.I.)

Every year, there is a long, loud debate when it comes to selecting the at-large bids — the N.C.A.A. teams that do not receive automatic tournament slots by virtue of winning their conference tournament championships. The fundamental question is: In sifting the candidates, should the committee pick on the basis of who is “most deserving,” which is to say the teams that had the better seasons to that point, or simply pick the “best,” which is to say the teams that gave other indications of overall quality?

R.P.I., which essentially measured only won-loss record and strength of schedule, was a tool for “most deserving.” NET, which combines those R.P.I.-like inputs with efficiency ratings and average margins of victory, is an argument for “best.”

Joel Sokol, a Georgia Tech professor of engineering who compiles a college basketball rating known as L.R.M.C., noted that NET has more closely tracked other advanced ratings, including his own and KenPom’s main rating, which measures offensive and defensive efficiency per possession, adjusted for opponent strength.

“It’s a lot better than the R.P.I., a lot more reflective of how good the teams are,” Sokol said.

Stanford’s Muir, who has served for several years on the selection committee that picks and then seeds the final N.C.A.A. men’s field of 68 teams, this week called NET “contemporary.”

“We’re quite pleased with how the new metric is working,” he said last month, adding, “There’s a thousand possessions that occur over the course of a year. Coming down to one possession is not going to adjust your NET that significantly.”

By contrast, the R.P.I really could be affected by a single possession. It was a good-faith attempt to do two things: situate teams’ records in the context of their strength of schedule, and discourage teams from running up the score. But this meant that a single bad (if close) loss or good (if close) victory could disproportionately sway a team’s rating.

And the R.P.I.’s myopic focus encouraged teams simply to schedule good opponents, or have the good fortune, shared only by major-conference teams, of facing many good opponents in league play. Worse, by not accounting for home-court advantage — in a sport in which the home team wins nearly two in three games — R.P.I. boosted teams that could afford to schedule more nonconference home games, which, again, tended to be major-conference teams.

“Depending upon the amount of money and support you have and your arena, you can basically isolate yourself from the rest of the world or at least control who you play and when you play ’em and where you play ’em,” said Doug Fullerton, who served on the selection committee when he was Big Sky Conference commissioner. “It’s a have and have-not situation.

Meanwhile, discouraging blowouts meant ignoring margin of victory, which sports analysts for decades have generally said is a better predictor of future winning percentage.

“When you played a good team, you didn’t get a bonus for winning by 40 than winning by 1,” Sokol said. He noted that Kansas has an unlikely 6-1 record in one-possession and overtime games this season, further explaining its aberrantly high R.P.I.

By contrast, NET factors game location into its Team Value Index, and the N.C.A.A. also sorts wins into quadrants based on whether they came at home, on the road or at a neutral court. And NET accounts for margin of victory, although in a nod to sportsmanship, the stat is capped at 10 points.

“I think that’s an honest attempt to try to capture what a point spread would tell you without creating an incentive for someone to run the score up,” Fullerton said.

To see the kinds of teams NET favors over R.P.I., and vice versa, consider two bubble squads: North Carolina State and Arizona State.

Despite playing in the competitive Atlantic Coast Conference, the Wolfpack have a dreadful R.P.I. of 100 thanks to a schedule rated 140th-most difficult. That is partly the result of a conference slate featuring only four games against the A.C.C.’s top three teams — Duke, Virginia and North Carolina — and a nonconference schedule that began with five home court gimmes. Yet with close road losses to the future low seeds Wisconsin and Florida State as well as wins over the likely tournament teams Syracuse and Auburn, North Carolina State’s NET is an eminently respectable 32nd.

On the other hand, Arizona State is a classic good-R.P.I. team, having defeated Kansas and much of a middling — but R.P.I.-boosted — Pacific-12 Conference schedule. On the other hand, there is much evidence that the Sun Devils are not that good. In the R.P.I., they are No. 37; in NET, they are No. 67. (They are No. 62 in KenPom’s ranking and No. 61 in Sokol’s L.R.M.C.)

Both teams are expected to claim one of the final at-large spots on Sunday. But if North Carolina State gets in while Arizona State is left out, it will be the surest sign yet that the Age of R.P.I. is over.

Source: Read Full Article