New Mexico is at third nationally, behind only Duke and Miami, in ESPN’s version of the RPI, coming off a loss over the weekend at UNLV. I wrote about New Mexico as one of the RPI darlings to beware of last week, and Carl Bialik of the Wall Street Journal picked it up and did a broader comparison across several years of teams ranked differently by Pomeroy and RPI at the start of February.
Pittsburgh and New Mexico are reverse mirror images of each other when it comes to the differing rankings. Pittsburgh is 30th in the RPI, but 4th in Pomeroy’s rankings, while New Mexico is 3rd in the RPI and 33rd in Pomeroy as of this morning. These two teams provide an interesting contrast, and are a textbook example of why some of us do not like the committee’s use of the RPI as a tool. Yes, yes, I know that the response is it is not the only factor, but is just a tool. It still provides the underlying language, and is why New Mexico is projected as a #3 seed by Lunardi, while Pittsburgh is a #5 seed (and was as low as #9 when I was projecting them as a #3 seed a month ago).
New Mexico is 20-4, and Pittsburgh is 20-5, so the records are virtually identical. So, which of these two teams has played a more difficult schedule so far? Going by RPI, the answer is clearly New Mexico, hence the difference. Since the components are win-loss record, opponent’s record, and opponent’s opponent’s record, we know that New Mexico must rate much higher (since their records are so similar).
The question of what schedule is most difficult depends on your perspective. The reason that Pittsburgh rates lower in the RPI is because of the non-conference schedules, where New Mexico is rated as the fifth toughest, while Pittsburgh checks in at 273rd. Quite a difference.
How much of it matters, though? The difference is driven by the weakest parts of the schedule. Pittsburgh’s six worst non-conference opponents were dreadful, winning only 23% of games to date. Pittsburgh destroyed them by 31 points on average. Those games neither proved or disproved anything about Pittsburgh.
New Mexico’s six worst non-conference opponents were respectable, winning 50% of their games. New Mexico likewise went 6-0, winning by 11 points on average. A really good NCAA at-large team would have gone 6-0 against the bottom of Pitt’s schedule, and still probably 6-0 against the bottom of New Mexico’s schedule, with more like a 95% chance of winning those games rather than 99%. Pitt played the dregs of Division I, while New Mexico played decent small conference teams, but teams that would still be the bottom of any major conference. Winning those games proves very little about either team and how they measure against other tournament teams, but it is the reason New Mexico is rated 3rd best, over any Big Ten team, while Pittsburgh is at 30th.
If we are asking from the perspective of one of the worst teams in D-I, then New Mexico has a far more difficult schedule. There are fewer opportunities for wins for bad teams, and it doesn’t matter much whether it is Michigan or UNLV at the top. A bottom feeder isn’t beating either.
Flip the perspective, though, and the opposite is true. New Mexico might still have the more difficult schedule from the view of a bubble type team (in other words, those games against decent small conference teams would have led to a few more losses). From the perspective of an elite tournament at-large candidate, though, I think it’s arguable as to which has played the more difficult schedule. Here are their respective performances against teams currently projected either in Joe Lunardi’s most recent bracket, listed as one of the first eight out (bubble), or on probation (Connecticut, who played both).
So look at those schedules. Which one is more difficult? Pittsburgh has faced teams that are projected as a 1, 2, 3, 4, and 5 seed. New Mexico has not. I think Colorado State is under-projected by Lunardi right now (probably at the expense of New Mexico), but even setting that aside, Pitt has played the weakest teams, but also the strongest. A typical 4 seed would have a losing record against those top 5 games. A typical 4 seed would have a winning record against the top 5 games on New Mexico’s schedule. At best you could say it is a wash, because of the presence of Michigan, Louisville, and Syracuse, more than offsetting all those New Mexico games against decent teams that a great team should generally win at a high rate.
This is what is maddening to me about the RPI. Pittsburgh has a profile every bit as good against the teams that should really matter, even setting aside margin of victory and that they are +7.2 against their schedule of tourney teams, while New Mexico is barely positive. The complaints about margin of victory and critiques about people not caring about wins and losses is misplaced, for those RPI defenders.
I don’t even think a committee should use something like Pomeroy or Sagarin straight up. Wins and losses should count, but we should look at it from wins added/lost compared to tourney at-large teams win expectancy against the same schedule. The committee should stop using RPI as a tool, and instead use a wins model that projects a wins above or below average compared to a median tourney at-large team (roughly a 6 seed or 7 seed). This would provide an objective view of the “bad loss” and weird profile issues. Take a bad loss where the median at-large candidate would have won 90% of the time? You offset it by winning a game on the road with a 10% chance against a top 10 team, or by going 2-0 in toss-up games against other at-large candidates.
Looking at the two profiles, I think they are pretty similar . . . right now just from a W/L perspective against NCAA and NIT type teams. Predictively, Pittsburgh has a better chance to improve it, which is why I’ll still have them projected higher in my next bracket projections despite that “weak” RPI schedule.
[photo via USA Today Sports Images]