Comparing Teams Viewed Differently by RPI and Pomeroy Ratings

Every year about this time, I like to compare teams that are ranked differently in the RPI rankings, the NCAA’s primary tool, and Ken Pomeroy’s ratings. Last year, I also looked at what happened when there was a big difference between the selection committee seeding and the Pomeroy rankings, entering the tournament.

I like to do this to separate some of the myth from reality of why RPI ranks some teams higher. Several years ago, I challenged the notion that the RPI rewarded inefficient winners while Pomeroy’s ratings favored efficient losers.

Let’s check in on this year. Here are the 15 teams, among those ranked in the Top 50 in either Pomeroy, the RPI, or both, with the largest difference in their ratings, where Pomeroy’s rates them better.

And here are the 15 teams that RPI rates as better among the same group of teams.

The “Pomeroy” teams are collectively 207-66 (75.8% win percentage). The RPI darlings are 209-73 (74.1% win percentage). That’s right, the Pomeroy teams have a better win percentage this year than those most favored by RPI!

The difference, then, must be schedule. And if you speak the RPI’s own language, when talking schedule strength, you’ll see that those RPI teams are considered to have played tougher schedules. But does that hold up? You’ll notice that there is a vast difference of opinion on the Pac-12, for example, with five of the fifteen favored by RPI coming from that conference. Those teams, of course, play each other frequently and will perpetuate the cycle in the two systems.

Let’s take Butler and Texas Tech as one example, since they are basically mirror images of each other in the two rankings, and might be considered “bubble”-type teams heading toward February.

Butler (ranked higher in Pomeroy) is 13-6; Texas Tech is 12-7. Against the RPI top 50, Butler is 1-5 (win over Purdue), Texas Tech is 1-7. Against the Top 100, Butler is 4-6, Texas Tech 4-7. Those look like very similar schedules.

Yet, the RPI has Texas Tech with the 5th hardest schedule, and Butler with the 77th. Where does that difference come from? Butler has four games against teams outside of 250 in the RPI, while Texas Tech has one. But when we think of difficult schedule, what do we think of? Beating up on a not-quite-as-horrible small conference team at home instead of a horrible one, or playing tough competition?

A tourney type team would beat a team ranked 200 in the RPI at basically the same rate as teams rated lower, particularly at home. Thus, those games that are swinging how teams are ranked in the RPI have zero practical impact. To prove this, I looked at the last 8 at-large teams in and first 8 out in the Bracket Matrix projections. Those bubble teams went 24-2 against the teams ranked 180-220 in the RPI. The two losses were on the road at major conference schools (Syracuse at St. John’s; Clemson at Minnesota).

This is part of the reason that I suggest using Wins Above Bubble, by measuring how a bubble team would do against the same schedule. Winning a game that you should win 93% of the time instead of 98% of the time doesn’t tell us all that much, but it goes along way in explaining some wonky RPI numbers. By that measure, Butler is above the bubble by about 0.7 wins, while Texas Tech is slightly below the bubble (though many projections currently have them in the field). The difference is that Purdue win right now, on a neutral court.

We can also look at that Purdue team compared to Oregon. Right now, Oregon is top 4 in the RPI. Purdue is outside the Top 20. They have the same number of losses. The difference is that, again, the RPI says Oregon’s schedule is much tougher. And if you go by things like Top 50 RPI wins (which seemed the biggest factor in the mock committee room last year), then Oregon blows it away. They are 8-2 right now agains the RPI top 50. Meanwhile, Purdue has only five such games (3-2 record).

That would seem to heavily favor Oregon. Let’s look at those schedules. These are best 10 games, by opponent RPI ranking, for each team.

Which one looks tougher? Purdue played Iowa, currently ranked in the top 4, twice. Oregon’s toughest opponent by RPI is Utah. The only ranked opponent is the home game against Baylor. Valparaiso and UC-Irvine are home Top 50 wins. Butler is not a Top 50 game for Purdue. In fact, 7 of Oregon’s toughest games by RPI ranking are home games. Four of Purdue’s have been true home games.

That’s great that I did that breakdown to show how this top 50 thing can lead to some absurd results. But that takes time that gets crunched at the bubble. The “Wins Above Bubble” ranking has these teams in a virtual dead heat right now, both projected as 4/5 seeds. Oregon’s tough schedule includes a non-Division I game. That doesn’t count, so in the RPI universe, it’s better to not exist than to be a game against a low-end D-I team, which pulls Purdue’s RPI down.

I talked about this several years ago in explaining how it was the wins at the bottom of the schedule, and not bad losses or quality wins, that had Maryland out and Boise State in back in 2013. Boise State played Walla Walla and Corban, games which did not count against the RPI, and made Boise State’s schedule appear tougher. My proposal would be to lop off the bottom 3 wins off the schedule so they don’t count against you and actually hurt the RPI. It makes no sense that beating a 11-20 team rather than a 5-25 team from a small conference can be the difference on the bubble.