"Wins Above Bubble" Would Be a Substantial Improvement Over the RPI

I’ve written about the RPI many times. (Spoiler alert: not a fan). I understand some of the concern about using margins of victory or other efficiency measures, though I disagree with many of them. Some of the criticism–that RPI rewards winners over efficient losers–doesn’t necessarily hold. It’s more based on the bottom of the schedule than we’d like to think–i.e. the games that don’t matter as much.

I started calculating something called “Wins Above Bubble” a couple of years ago. Seth Burn is far better at it than I am, though, and using a very similar method, has the numbers for this year.

The idea behind it is simple. It tries to measure how likely a team is to win in each game based on opponent and venue, and assigns a probability. If a bubble team would have a 70% chance of winning a game, and our actual team wins, it is +0.3 wins above bubble. Over the course of 10 games, it would be expected to go 7-3.

It uses the underlying Ken Pomeroy efficiency rankings to assess the likelihood of winning, but one thing it does not do is include margin of victory. It is a reflection of past accomplishment, just like, theoretically, the RPI. The difference is that the mere playing of a game cannot punish a team, like with the RPI. You will see teams go down by blowing out a bad opponent in the RPI. In the WAB system, if you play someone and a bubble team would have a 95% chance of winning, the win simply keeps you holding steady. Teams improve in these rankings by racking up small victories (winning a disproportionate number of 80% games) or pulling big upsets that bubble teams would usually lose given the location and opponent.

Burn also uses three different baselines: Wins Above Bubble, Wins Above Strong Team (WAST), and Wins Above Monster (WAM). They show that strength of schedule often depends on perspective. Kansas is #1 in wins above bubble, playing a very difficult schedule. From the perspective of a “monster” powerhouse team–the exact type of team that is competing for a top seed–they are 4th and Kentucky is #1.

The WAB rankings could easily supplant the RPI as a grouping tool. In fact, at least with the higher seeds, the committee does seem to instinctively simulate a combination of those rankings. Consider Notre Dame, who is currently 28th in the RPI (thanks to a cupcake non-conference), but is 22nd, 16th, and 12th respectively in WAB, WAST, and WAM baselines. I would submit to you that Notre Dame would be seeded more like a 3 to 5 seed than a 7 to 8 seed by the Committee. The counter is true of a team like SMU, who is higher than Notre Dame in the RPI, but will be seeded closer to where WAB has them (#30, 8 seed).

While the committee tends to replicate the WAB results near the top of the seed lines, and for teams that are ranked and visible, it is at the bubble where the RPI now holds its greatest sway. But it is in this range where there is actual disagreement and nuance and the potential for teams to game the system (which you can’t do with WAB).

Let’s take the cases of Alabama and Temple this year. Temple is very much in at-large consideration right now in a RPI world; Alabama is on the outside. Temple’s RPI is 43 while Alabama is in the mid-60’s.

Using WAB, though, Alabama actually rates as slightly higher.

Alabama is 13-8 against Division I teams. Temple is 16-7. A “bubble” team would have about 13.7 wins against Alabama’s schedule, and 17.2 against Temple’s.

Alabama is 1-6 against the RPI Top 50, while Temple is 1-5. You can bet that Temple’s win over Kansas will be trumpeted as a reason.

The number of wins each should have against those top opponents is vastly different. Alabama played Kentucky twice, and that is their only home game against a top 25 team. Few teams would beat Kentucky anywhere this year. Alabama has actually played at Kentucky, at Xavier, at Wichita State, and at Arkansas, along with a neutral game against Iowa State. Most of those teams will be seeded highly by the committee.

Temple has played Kansas at home, Duke at a neutral site, and at Villanova. They have also played home games against Tulsa and SMU, both losses. Those are games a bubble team would be more likely to win and Alabama’s schedule is devoid of those type of opportunities so far.

Is Temple a much better road team? They are no doubt going to receive a boost for being 6-5 in road/neutral games, while Alabama is 2-6. You can bet that would be pointed out.

Temple isn’t any better on the road, though, and the difference is who they’ve played. Temple lost to Duke, Villanova, and Cincinnati on the road by an average of 25 points. The second best road win is at Tulane. Alabama has played more tough games. Their second best road win is Arizona State on a neutral court.

All this gets summed up in Wins Above Bubble. If you want to reward Temple twice for beating Kansas, you can do that. Although that was a win that boosted them quite a bit. The problem is that the rest of the performances have not.