- In the late 1950’s, UNC basketball coach Frank Maguire and an assistant named Dean Smith recognized that their tools for evaluating team success were limited. They played an up-tempo game, so naturally they’d average more points than teams that played slowly. But who had the better offense? So instead of using total points or rebound margin, they began charting stats on a per-possession basis. And that’s it - that’s what “advanced” metrics are. They break down the game into one possession bits so that you can compare teams of different tempos. So when you hear broadcaster Cory Alexander lamenting all of these “new” analytics, just remember that they’ve been around since before he was born.
- Possession-based stats hit the mainstream with the publication of the ground breaking book Basketball on Paper by Dean Oliver in 2002. He’d been publishing journal articles on the subject since 1988, but his book was the first synthesis. Unfortunately (for us, not him), he was hired into the NBA in 2004 and so his work was no longer public. Luckily, Ken Pomeroy, a government Meteorologist, stepped in to fill the void. Over time it became obvious that Pomeroy was outperforming Vegas (good for you if you got in early) and if you fast forward to 2022 the Vegas line and the KenPom line are almost always within a point of each other. In other words, it’s a great predictor.
- KenPom and other predictive systems use efficiency margins to rank teams. This is how many points you score per possession, and how many you allow, run through a filter that adjusts these raw numbers based on the quality of offense or defense you face. Twenty years ago this was basketball’s version of Moneyball. Now all NBA teams have an analytics staff that are working at a much deeper level. College teams would be wise not to fall too far behind.
- The growing prevalence of possession based stats made the re-evaluate their garbage called the RPI, and so they released the NET in 2018-19 to move more to possession based stats. Why they wanted to reinvent the wheel is up for debate as there were several good systems already out there (KenPom, Bart Torvik, Sagarin), but it’s the NCAA so don’t try and interject reason.
- The key thing to understand about possession based stats is that they are predictive. The AP Poll is resume based. KenPom isn’t ranking resumes, he’s trying to figure out which teams are the best. The NET tries to be both. They call these two factors the Team Value Index (resume), and NET Efficiency (predictive metrics), but their calculations happen behind the curtain. It’s like they don’t realize that teams with good efficiency margins also tend to be teams with good resumes. So why use both? Early in the season this just leads to wacky results (like Colgate being one of the best teams in the country), and late in the season the NET basically resembles KenPom with a handful of outliers.
- Enter the 2021-22 FSU basketball team. I keep hearing “how can FSU be a bubble team if they’re leading the ACC?” Easy answer: the NET. And neither Florida State’s resume nor efficiency margins are very good. The NET uses four tiers, and in those tiers FSU is 2-3 vs quad 1, 3-0 vs quad 2, 4-2 vs quad 3, and 4-0 vs quad 4. Based on the current NET rankings where FSU is 58th, it’s clear that quad 3 losses hurt more than Quad 1 wins help. Only two of the top-32 teams have a quad 3 loss, and none have a quad 4 loss. So Houston is ranked third, and are 1-2 vs quad 1, and 2-0 vs quad 2, and 14-0 vs quads 3 and 4.
- So what are FSU’s bad losses? It was home vs Syracuse, and a “neutral” game vs South Carolina ~60 miles from their campus. Syracuse is currently 97th, and quad 3 in home games in teams ranked 76-160. Quad 3 in neutral games is 101-200 and South Carolina is 108. Who is number 200 you ask? That would be Tarleton State. A neutral site loss to Tarleton State would count the same as losing to USCe just down the road from their campus. Meanwhile, Houston’s two losses are by one at Alabama, and by two in a neutral site vs Wisconsin. Efficiency margins matter. Houston is two buckets away from being undefeated. In addition to FSU’s two bad losses, they lost by 16 at Florida, by 28 at Purdue, and by 22 at Wake Forest. A resume based system would count a loss as a loss, but a predictive system recognizes value in how much of a beatdown took place.
- FSU is 57th with a 5-3 record vs quads 1 and 2, and Houston is 3rd with a 3-2 record. Why? Because Houston would be a large favorite if the two teams played. Houston is 4th at KenPom, while FSU is 54th. There is as big a difference between Houston and FSU’s efficiency margins as there is between FSU and Pitt. And currently FSU is projected to be a 13-point favorite over Pitt at home.
- So FSU is on the bubble, like it or not. They’re currently projected (based on point spreads) to finish 13-7 in the ACC. That will probably good enough to just sneak in, even with a first game loss in the ACCT. But to guarantee dancing, FSU will need to outperform the computer projections and come in with at least 14 conference wins.
FSU hoops vs the Computers, or why FSU is a bubble team
It’s bubble talk time