Michigan is Getting Screwed by the RPI
By Jason Lisk
Michigan is getting screwed by the vagaries of the RPI, when it comes to their chances of getting an at-large bid.
That generally matters. RPI is the underlying language that has been used. That doesn’t mean the committee goes strictly by it, but it certainly frames the outlook. Here’s a quick recap of how many at-larges have been selected with the following RPI rankings over the last four tournaments:
61 or worse: 0
56 to 60: 2
51 to 55: 9
46 to 50: 9
41 to 45: 17
36 to 40: 15
31 to 35: 16
Being inside the top 45 almost certainly means selection, particularly if in a power conference and if you have some quality wins. Being in the 46 to 55 range is the true bubble for power conference teams. Above that, it takes a lot of quality wins to overcome a bad RPI.
Michigan’s RPI is currently 66.
They have three good wins over teams that will be seeded fairly well in the tournament (Maryland, Texas, Purdue) and that gives them a fighting chance, but if the RPI was better they would be considered a far more likely candidate.
I’ve railed on the RPI in the past, and proposed that the committee switch to something called Wins Above Bubble (WAB) as a tool. It only looks at wins and losses and not margin of victory, but it answers a simple question: how many wins would a typical bubble team have playing this exact schedule?
It has several advantages: it’s simpler, it’s intuitive, and it can’t be gamed. That’s a very important point. Here’s the most current version as compiled by Seth Burn, which now uses the average by combining several underlying rating systems.
If you compare where there are differences, the WAB results seem to jive better and seem more intuitive, whereas there are often numerous mental gymnastics to adjust the RPI rankings. Oregon one of the best 4 teams, or the 10th best team (as WAB has them)? Michigan State and Indiana at 15 and 20? WAB has them both much higher. Gonzaga, not one of the 50 best teams in the country, or should be seeded like a 9 seed, like WAB would? Mid-majors like Akron and South Dakota State, among the 40 most accomplished teams, or down in the 60’s and 70’s?
So let’s turn to Michigan. Michigan is 66th in RPI, but 43rd in the Wins Above Bubble measure, at +0.43 wins more than expected. Basically, we would expect a bubble team to have about 19.6 wins against Michigan’s schedule, so a 20-11 record should be very much in the thick of the discussion for an at-large, though no lock.
Let’s contrast that with VCU, who’s almost a mirror image at 40th in the RPI, but 65th in WAB, at -0.66, and Akron, who is 34th in the RPI and 66th in WAB.
According to RPI, those two teams have better accomplishments, by going 22-9 and 24-7 respectively against their schedules.
What’s more impressive, going 20-11 against Michigan’s schedule or 22-9 against VCU’s and 24-7 against Akron’s? Here’s a comparison of their opponents, both home and road, against different levels of opponents.
Michigan has played 12 top teams, compared to 3 total for the other two, yet their schedule is not viewed as much more difficult by the RPI. Why? Because those sub-225 games are pulling Michigan down, because those teams were not just sub-225, but really bad.
Rutgers is a problem for the Big Ten. How big? It’s hard to say exactly, but just removing two wins by Michigan State by 30+ over Rutgers would improve the Spartans’ RPI from 15th to 11th. That’s right, simply playing those games against Rutgers is costly. That makes no sense. Multiply that over the course of 13 opponents in the Big Ten, and the difference adds up.
Then, you look at the non-conference. Michigan played Xavier, NC State, Texas, Connecticut, and SMU. That’s more top quality games then most programs played. But, from December 12 to December 23, they also played Delaware State, Northern Kentucky, Youngstown State, and Bryant. They won them by an average of 40 points. Each of those teams is at 275 or below in the RPI. Those teams are collectively 28-90 against Division I teams, and play against other low level teams (that whole opponent’s opponent’s record thing).
That’s killing Michigan.
How much?
If you just took out two of those games, and replace them with home games against mid-level MAC teams like Eastern Michigan and Toledo, the RPI goes from 66 to 55. With just that change alone. Heck, even if they lost one of those games (and they would be heavy favorites at home), the RPI actually goes up slightly. That defies logic.
And getting to a 55 RPI, with literally nothing else changing about the tough games they played, means Michigan now has about a 40-50% chance of selection right now.
If you replaced two of those games with Eastern Michigan and RPI darling Akron, then it gets even better. Michigan shoots all the way up to 43 in the RPI!
From merely playing a game they most likely still win, and which has little to do with measuring how good they are. Michigan, if the RPI was 43, with the exact same profile, would be a near-lock for the tournament with those 3 big wins.
And that’s the problem with the RPI, when you peel back the onion. A couple of really bad non-conference opponents versus nearly-automatic-wins-versus-slightly-better-teams can make the difference between near-lock and virtually no chance.
Teams schedule non-D1 teams because they don’t count. (If Oregon State had played the worst D-1 team instead of Northwest Christian, for example, their RPI goes from 30 to 35.) Some teams are really good at gaming the system–the Atlantic 10 is constantly placing higher in the RPI than they do in other predictive systems like Pomeroy.
A system shouldn’t have the tail wagging the dog. It shouldn’t be able to be gamed by manipulators, with the mere existence of a matchup causing a ranking to rise or fall, before a ball is ever bounced.
Michigan is just the latest example. Going 20-11 against their schedule is more impressive than VCU at 22-9, or Oregon State at 18-11 (47th in WAB, 30th in RPI), or St. Bonaventure going 22-7 against theirs (49th in WAB, 27th in RPI).
We can do better. It’s 2016.