It’s time to reveal the results from the H.U.M.A.N. poll, America’s first ranking of every D-I men’s college basketball team by the people. You can read about the method behind this exercise here.
A total of 719 voters made selections on 32,489 matchups. The resulting rankings are shown here. Subscribers can click on a team to dig down into the individual results. All of the raw votes are here.1
Congrats to Kansas on being selected for the top spot in the inaugural H.U.M.A.N. poll. That’s a banner that nobody will ever make you take down.2
One of the many nice features of the H.U.M.A.N. poll is that the results are more interpretable than the AP poll. For this purpose, two ratings are provided. The first is the Bradley-Terry rating (BT Rtg). This allows you to determine how often one team would be preferred over another by our average voter.
To illustrate this, let’s use the top two teams as an example. Kansas has a rating of 383.4 and Duke has a rating of 161.3. This does not mean that voters thought Kansas was 2.4 times better than Duke. It means that the average voter preferred Kansas over Duke 70% of the time. You get to this number by taking the share of Kansas’s rating relative to the sum of both team’s ratings: 383.4/(383.4+161.3).
In the Kansas voting summary, you’ll notice that the Kansas-Duke matchup was served up 13 times. Kansas was chosen in 10 of those, for a 77% win rate. The 70% figure comes from the additional information gained from all of the other votes. Technically, every pick has a non-zero effect on the rating of every other team.
The problem with the BT rating is that you can’t compare differences in magnitude among teams at different locations in the list unless you are very good at doing division in your head. So give a warm welcome to the linear rating (Ln Rtg), the hero we need. With the linear rating, you can do things like say #7 Michigan State finished closer to 3rd (.48) than 9th (.51).
In fact, besides the gap from first to second, the largest gap between any two spots in the rankings was from 7th to 8th. Voters separated the top seven from everyone else, which either speaks to the quality of the top seven, or that you don’t have to get very far down the list to find some serious parity.
The linear rating also allows us to derive conference ratings. Did voters properly identify the best conference in the land? Yes, yes they did.
avg rtg
1 B12 12.05
2 BE 11.70
3 B10 11.66
4 SEC 11.63
5 ACC 10.78
6 P12 10.74
7 MWC 9.47
8 WCC 9.31
9 Amer 9.06
10 A10 8.71 If we are going to criticize the humans, they seem have a brand-name bias. In the absence of doing a few minutes of research, voters seemed to rely on a team’s reputation to guide them.3
This worked in both directions. UCLA landed at 12th and they may not get a single AP voter to rank them that high. Either there’s a bunch of Berke Buyuktuncel fans out there or more likely, a blind belief that Mick Cronin can have success regardless of roster.
On the flip side, Colorado will get some top 25 consideration by the AP voters, but they could barely crack the top 50 here. And Northwestern (#62) and Stanford (#96?!) seemed to be judged more on their reputation than what’s on their roster.
As we get further down the list, the ratings figure to become less trustworthy. The computer served up fewer matchups for those teams, so there’s bound to be more noise, and people just know less about those teams. Still, I think teams are generally in the right neighborhood. Though my nomination for the worst pick is Chicago State at #349. They cracked the top 300 last year and might have a better roster.
Finally, the sole motivation for this project was Florida Atlantic, who is the most interesting pre-season story in the sport. My precious preseason AP poll is normally awesome because it’s the only time voters rank the teams they think will be the best. And while that’s a non-specific goal, we all kind of know what it means.
I suspect FAU will be ranked in the AP top ten and sadly, the rationale seems to be that they “deserve” to be there after earning a nine-seed in the NCAA tournament, getting to within one shot of an appearance in the championship game, and bringing everyone back. I mean if they were UNC, they’d be ranked #1 probably.
But do voters really think that FAU is one of the ten best teams in the country? This is where the methodology in the H.U.M.A.N. poll is pretty powerful. Instead of simply ranking FAU in the top ten, you have to decide whether you actually think they’ll be better than Michigan State, or Creighton, or North Carolina.
To be fair, some H.U.M.A.N. voters said yes to those questions - the Owls did narrowly beat out UNC in the poll - but not enough to rank them higher than 20th. Which honestly feels more fair to FAU.
Thanks to everyone who participated. I learned a few things from this inaugural effort to improve on it next season. While my computer is not yet feeling threatened by the humans on the full 1-362 list, the top 50 seems pretty reasonable. I suspect the results from this exercise will be useful in some way to inform future preseason rankings from my computer.
But most importantly, it’s a unique addition to the pre-season rankings scene. I hope you enjoy looking through the results as much as I enjoyed collecting the data. And if you disagree with something, well, you only have your fellow humans to blame.
Eventually, you’ll be able to track the success of your picks on the site.
That’s assuming you want to put up a banner for this, which I encourage.
It’s worth noting that national media types seem to have an anti-brand name bias. Whether it was Dayton last year, St. Bonaventure the year before that, or Richmond in 2020-21, there’s often an effort to push non-traditional teams into the poll who don’t last there long.