Let's talk about Memphis
What happens when the AP rankings and the KP rankings diverge?
It’s that time of year when the whimsical minds of AP voters diverge from the predictive realities of scoring margin. For instance, Memphis made its season debut in the AP top 10 on Monday despite being ranked 45th in my ratings.
Memphis appears to have achieved their ranking based on a combination of name recognition, a long winning streak, and perhaps most importantly, credit for beating teams who seemed a lot better when the Tigers played them.
I loathe the cherry-picking industry that the hot twitter accounts have hung their hat on, but allow me to briefly play that game before indulging in more serious analysis. Here are the teams ranked in the top 12 in this week’s AP poll and the number of wins they have against teams in the top 30 of my ratings, as of Tuesday:
1. Connecticut 2
2. Purdue 6
3. Kansas 5
4. North Carolina 2
5. Houston 2
6. Tennessee 2
7. Duke 2
8. Kentucky 1
9. Baylor 2
10. Memphis 0
11. Wisconsin 2
12. Arizona 5
I feel dirty for having engaged in this exercise but also I can see the appeal of using arbitrary cut-offs to make one’s point. It’s so much fun to cheat! Cherry-picking aficionados were easily able to spot that my thresholds here were carefully chosen. The 13th team in the AP poll is Auburn and they do not have a top 30 win, either, so we only did the top 12. And as of Tuesday, Memphis’s best win was against #31 Clemson, so let’s just use the top 30.
Still, this effort does show that Memphis doesn’t have a high-end quality win. Also, this is unrelated, but what are people doing with Arizona? I try to respect the AP poll and give its many fans their space. People should be allowed to like whatever strange thing they want in the privacy of their own home.
But I don’t see the utility of a poll where the team that has the second-best odds to win a national title and has a pile of quality wins to back it up is ranked 12th. What is the point of this exercise? I guess it’s to punish teams for losing to Washington State.
Enough silliness, let’s embark on some more useful work. We can look at all of the teams since the 2011 season who were ranked in the AP top ten and simultaneously not ranked in my computer’s top 30.
Memphis is the 12th team to join this club. Sadly, the previous 11 didn’t really distinguish themselves. None of them ended up being seeded better than 4th and none of them got past the 3rd round. And the group’s average kenpom rank between the time of the AP poll and the end of season was almost identical, so it’s not like AP voters have any skill at identifying who might be underrated in my system.
I will say that, accounting for seed, these teams didn’t underperform in the tournament. They did almost exactly what would be expected based on history. But given that each squad was in the top ten at some point, failing to get past the Sweet 16 is a bit disappointing.
Since I’m off the cherry-picking train, and back to being intellectually honest, I’ll reveal that on December 16, 2013, UConn was ranked tenth in the freshly-revealed AP poll while ranked 28th in my rankings. After an unremarkable regular season - UConn entered the tournament 25th in my ratings - the Huskies went on to win the national title as a seven-seed. But it’s also worth noting that UConn was 28th then and Memphis is 45th now and that’s a pretty big difference.
Hey, Memphis could be an exception, but people are reliably overconfident in thinking they can identify the exceptions. I think if one did the research, they’d find that people thought each of the cases above would be the exception at the time. Alas.
While we’re here, let’s look at another case even higher in the AP poll. Kansas is ranked 3rd this week - they even got three first-place votes - while being ranked 18th in my ratings. Here are the cases of teams that were in the top 5 of the AP poll while being outside the top 15 of my ratings:
The difference between this group and the first group is that this group gets better seeds in the tourney. Another difference is that they significantly underperform (by about a half-win per team) relative to their seed. Five of the six teams seeded first or second couldn’t get past the second round. O-ver-rat-ed. That chant sucks if you do it at a game, but you can break it out now. Silently, in your head. It’s totally appropriate.
To be fair, there is a Four Four appearance by 2017 Oregon here and an Elite Eight appearance by 2014 Michigan State to salvage some dignity. But the overall trend is pretty brutal. Like Memphis, Kansas could buck the odds and join those teams. It was a good start to bury Oklahoma State last night. And with Bill Self on the bench, you have a good reason to buy in.
But as with the cases like Memphis, I suspect most every time this happens, people can find reasons why the AP poll and the “eye test” are finding things the predictive metrics are missing.
And most every time, those people end up being wrong.