I’m Andrew Michael H, and every Sunday here at BTP I’ll be doing a deep dive into the numbers. Every Sunday, that is, except this one. There just haven’t been enough games for there to be meaningful statistics yet. So instead, I’ll give you a brief overview of the kinds of numbers I’ll be discussing.

KenPom Is King

There are five advanced college basketball metrics that are worth paying attention to: KenPom, T-Rank, Sagarin, ESPN’s BPI, and the new NET (NCAA Evaluation Tool), which replaced the RPI as the metric that the NCAA uses to organize teams when the Selection Committee evaluates teams.

Of these five rankings, I think ESPN’s BPI is probably the best of the bunch when it comes to predicting the results of games (mainly because it’s capable of factoring in suspensions/injuries), but only KenPom and T-Rank provide any kind of in-depth detail, and it’s the detail where the real insight comes from. Since KenPom’s site is so user-friendly and because it has more cache in the collective college basketball consciousness, at least 80% of what I discuss will come from there. T-Rank has some interesting simulations that you can run, so later in the season I’ll probably hit that a little harder.

Disclaimer: Most of the KenPom stuff I’ll be using is behind a paywall. Rather than relying on me to bring the most important information to you, you should subscribe to KenPom.com. It’s the best $20 a basketball fan can spend. To be honest, I don’t even check CBS or ESPN for scores anymore. I use KenPom’s FanMatch every morning. It gives you a rundown of all the previous night’s games, and any upsets are red so they’re easy to spot.

Efficiency, Efficiency, Efficiency

I think the tempo-free revolution has finally penetrated the consciousness of college basketball fans as a whole, but in case it hasn’t, here’s a brief summary.

• It used to be the only stats we had were per-game stats. “Team X scores 100 points per game.” “Team Y only gives up 45 points per game.” Team X must be great at offense; Team Y must be great at defense.
• But no. Team X just plays a lot faster than Team Y. They score more points, but they give up more points, too. Per-game stats don’t give the full picture because some games have 55 possessions and some games have 75 possessions.
• What you want to look at, therefore, are per-possession numbers.
• Last year, the NCAA Men’s D-I average points per possession was 1.05.
• Good offensive teams score about 1.10 points per possession or more.
• Good defensive teams give up about 0.98 points per possession or fewer.
• You have to adjust the raw points-per-possession numbers to account for the teams that you’ve played. Scoring 1.07 points per possession against a Big Ten schedule is far more impressive than doing the same thing in the MEAC.

Offensive and defensive efficiencies are the primary underlying numbers that drive everything else, but we’ll look at turnovers, rebounds, individual player statistics, anything that jumps out as interesting or unusual or highly explanatory of the results we’re seeing. But those stats are only important in terms of how they reflect offensive and defensive efficiency. At the end of the day, the game of basketball is simple: you’re trying to score as much as possible and give up as few points as possible.

“It’s Early”

Unfortunately, it takes awhile for any of the advanced statistics to become reliable. The more data you have, the more accurate your predictions become. The less data you have, the more things can be skewed by freaky results. And early on, we don’t have much data. Plus there are teams that were expected to be good but actually suck, and vice versa. Water eventually finds its level, but not right away. So what should you pay attention to early?

A system like KenPom is built on top of Bayesian predictions. “Bayesian” is a fancy statistical term that means you start with a particular underlying assumption and then update that assumption as you observe actual results. (The latter part of that all has a solid underlying mathematical basis. The first part, picking the starting underlying assumption, is where things are more art than science and where most of the arguments around Bayesian statistics come from.)

That means that as the year goes on, KemPom relies on the starting assumption less and less. It takes until January for the starting assumption to go away completely, but that doesn’t mean the numbers are garbage until then. Instead, you should pay attention to what direction your team’s ranking is moving.

If a team starts out at No. 55 and they pull a few upsets or blowout wins and rise up to No. 35, it’s probably a safe bet that they’re actually somewhat better than No. 35 and will keep moving up. How much better? That’s difficult to say. But if a fan of a rival team tries to tell you their team is better because they’re sitting at No. 27, the argument that you want to use is “KenPom’s starting algorithm underrated our team. We’re better than our ranking, and it will bear out as we play more games.” And it wouldn’t be wrong to call his girlfriend ugly, either.

Now if you’re sitting at No. 55 and after a month you find yourself at No. 52, well that’s a sign that you’re probably in the right spot overall. Sorry. Enjoy life on the bubble. That doesn’t mean your team can’t pull together and go on a run, but they will have to actually get better to do so. You aren’t getting “screwed by the computers”; you’re getting screwed by your coach and your shitty players. You should probably fire your coach. And the AD that hired him, too, while you’re at it. Don’t settle for less.

One other good rule of thumb to tell if your team is overrated or underrated by KenPom early in the season is to look at the Vegas lines on your games. Late in the season, unless there’s been a major injury or something, Vegas lines are typically within a couple of points of the KenPom prediction. But if Vegas has your team as a 14-point favorite in a December game whereas KenPom only predicts you’ll win by 6, that’s evidence that your team is probably better than its KenPom ranking. (Or that your opponent is worse than their KenPom ranking. Or possibly that your fanbase is full of insane people who overestimate the strength of your team and Vegas is looking to take their money.)

Conclusion

College basketball is awesome. Apart from rooting for my alma mater (Boiler up!), I’m mainly here for the memes and the math. I can’t compete with Twitter and Reddit when it comes to memes, but maybe I’ll have something original to say when it comes to math.

Oh, and one more thing. This column posts on Sunday mornings, which means it needs to be written by Saturday night at the latest. KenPom updates continuously as game results come in. I’m not staying up until 2 AM for the last of the west coast games to finish, so the numbers you get from me here will always be a little out of date. They shouldn’t be off by much, but that’s just one more reason you really should buy your own KenPom subscription.