Welcome to the H.U.M.A.N. poll. No longer do you have to wait years hoping the Associated Press will give you one of its 60-ish coveted spots to vote in a poll ranking the top 25. Now, with a kenpom.com subscription, you can participate in the world’s first human-based 1-362 preseason poll for college basketball. Instead of the elite voting in a poll honoring the elite, everyone can vote in a poll honoring every team.
Upon clicking this link you will be given 50 match-ups (subscribers only). In each match-up, your job is to pick the team that will be ranked higher in my ratings at the end of the season (post-tournaments). After polling closes at midnight (Mountain Time) on October 6, everyone’s selections will be aggregated into a 1-362 ranking using the Bradley-Terry algorithm.
Why would you do this? Well, it’s possible that humans can produce something useful for a 1-362 preseason poll. The snag is that it isn’t feasible to ask humans to rank all 362 teams. You could try and do it, and people have. But it’s impossible for humans to organize that much information in a coherent way. And I hear it’s ridiculously time-consuming.
However, humans can build something pretty useful if they just have to focus on a few matchups, especially if we can get a lot more people than vote in the AP poll. Maybe we can produce a poll that beats a computer’s preseason poll or at least informs a better version of a computer-created poll. At least that’s my theory.
This really should beat the pants off of the AP poll, for two reasons:
1. You’re not subject to anchoring issues.
I’m definitely not going to be the guy who says we shouldn’t release any rankings until January. Rankings are fun. Rank things ten years from now. Knock yourselves out.
That said, the rush to get rankings out as soon as the last notes of One Shining Moment are played really hurts the concept of independent thought. Increased player movement in the modern game mitigates this somewhat, but if you’re waiting until October to produce a preseason ranking, a bunch of other peoples’ overconfident opinions have already poisoned your brain.
Kids, there was a time in the summer of 2022 when the consensus was that North Carolina was a top ten team. Third, fifth, seventh, these were all reasonable opinions. Then North Carolina added the best player from the 87th-best team in the country and suddenly it was consensus that UNC was the best team in the country. Almost every smart person said so! They got over ¾ of the first-place votes in the AP poll.
Kids, UNC did not even make the tournament. In fact, they finished lower on this site’s rankings than that team that ranked 87th the year before and lost their best player. I do not believe every smart person would have thought UNC would be the best team in the country had they not had access to their colleagues’ opinions. But alas, a lot of cherry-picked pieces of evidence came together to produce an artificial consensus.
To be fair, there’s still potential influence issues with the H.U.M.A.N. approach, but if you thought North Carolina should be #15 instead of #1, you wouldn’t have to worry about the public embarrassment of ranking them 15th. You can even share your ballot with the world before voting closes and it won’t matter.
The H.U.M.A.N. poll doesn’t actually care where you think a team should be ranked. It just cares about your opinion on whether they’re better than one team – the team they’re matched up with on your ballot. Bradley-Terry will take care of aggregating everyone’s opinions into a true consensus.
There’s no doubt with this method you have more freedom to vote how you please. Even if your opinion about a team is objectively terrible, it’s not going to affect the quality of the poll. And you might not get to vote on that team anyway.
2. There’s a clear target and accuracy is rewarded.
It’s not completely clear what the AP voter is trying to do. Is a voter ranking teams in order of their chance of winning the national title? Ranking teams in order of strength at the beginning of the season? Ranking teams based on their chance of getting a one-seed? There are no rules for the AP poll. It’s the Outback of college basketball polls, except unlike the blooming onion, it’s not “just right”.
There also isn’t any reward for being accurate. Percy Allen didn’t get a bonus for having UNC lower than every voter last year (#4, wow!), and the 10 people who ranked Florida State or the 11 people who ranked Wyoming (teams that went a combined 18-45) – and the one person who ranked them both – are not losing their votes because of it.
But there is a clear measure of success in the H.U.M.A.N poll: Whether your choice in each matchup finishes higher in the final kenpom rankings. At the end of the season, we will total up everyone’s successes and figure out who did the best.
If my code works correctly, your ballot should be structured so that most of the match-ups are competitive. There will be a few confidence-builders for you as well. In addition, the ballot will be slightly weighted towards better teams. You shouldn’t be getting a bunch of potential sub-300 match-ups. But this is my first time doing this so I can’t rule out some bugs.
If you finish in the top ten* of correct picks, you will get special recognition on whatever social media platform I am using at the end of the season, and a year added to your subscription. The top 3 will get to choose one item from Ken’s closet: A signed Chris Duhon Chicago Bulls’ jersey, a 1984 Topps Don Mattingly baseball card (ungraded), or a 1981 US coin proof set.
I purposely do not provide any links to information on the ballot. You are free to do as much or as little research as you want, from whatever sources you like. You don’t even have to vote on all 50 match-ups on your ballot, but you might as well try if you want a Mattingly rookie card.
In addition, if we get at least 1000 voters, for each of the top ten* finishers, I will donate $500 to the charity of your choice. I could just give you $500 but then I’d have to get lawyers involved.
Polling will close at midnight on October 6. You are free to change your votes up until then. At that point all of the picks will be locked. I will aggregate the selections using the Bradley-Terry algorithm. A list from 1-362 will be created and published. We’ll circle back at the end of the season to see how y’all did. If you did well, we’ll try again next year and include the results in the my own preseason ratings. That’s really the hope here. Humans, assemble!
*entries with the same number of correct picks will be ordered inversely by the highest ranked team in the final kenpom rankings that was picked incorrectly. For example, if two entries have the same number of correct picks, and entry A’s highest-ranked miss was the #2 team in the final rankings while entry B’s highest-ranked miss was the #27 team, entry B would be ranked ahead of entry A.