(Author’s note: This article will get a little nerdy. My trade is in market research, where I use a ton of applied statistics, so I enjoy this kind of stuff. But rest assured, advanced analytics may not impact your enjoyment of college basketball at all, depending on your point of view. The goal of this article is to explain one popular advanced analytical approach, KenPom, at a beginner’s level, to help readers better understand the metrics and avoid confusion in OTB’s game previews.)
Utilizing advanced statistics from the analytics site KenPom is something Aaron does quite frequently during the season, including every game preview. It’s the brainchild of Ken Pomeroy, who now writes for The Athletic. His site has been in operation for the better part of two decades, but it’s only been the past few years that his advanced statistics have become more mainstream in regard to measuring the quality of college basketball teams.
If there is just one thing you take away from this article, it’s that there is absolutely nothing subjective about the KenPom rankings. It strictly measures efficiency of all 353 Division I teams and helps predict how each team will fare game to game based on those advanced statistics when matched against opponents. If you want subjectivity in your college basketball analysis, there’s plenty of that elsewhere!
That KenPom rankings are purely measured from an efficiency perspective explains why Penn State, who finished 14-18 last season, also finished with a final KenPom ranking of 43 (which is very good). Importantly, Penn State had the 10th highest strength of schedule in the country last year; they faced a gauntlet both in- and out-of-conference, and though they lost more than they won, from a pure efficiency standpoint they were a better basketball team than most NIT teams and NCAA bubble teams in 2018-2019.
It also explains the current top ten ranking on KenPom as of December 9, 2019:
While several of the teams in the KenPom top ten currently have zero or one loss, you can also see Michigan State and Purdue (two three-loss teams) in the top ten. To KenPom’s statistical model, it matters who a team plays and how efficiently it plays them (offensively and defensively) more than it matters who wins.
(Nerdy side-note, feel free to disregard: I’m speculating here, because of course the KenPom algorithm is proprietary, but I’m guessing KenPom uses a logistic regression model similar to what Nate Silver uses at FiveThirtyEight to predict election results based on polling data. It updates multiple times a day, after games go final, and it simply uses box score data to update its results. From a statistical perspective, it’s a pretty remarkable thing…)
A Deeper Dive into KenPom
Next, we will explain some of the measurements that KenPom has created and in an effort to relate that to better interpreting Rutgers basketball from an analytical perspective.
There is a ton of data for every team, but most of that is only available if you have a subscription to the site, which I do. I will focus instead on the main page that anyone can view at www.kenpom.com (the screenshot above is from that main page).
What we will look at here is everything that you will see on the front page first and then get into some of the other odds and ends stuff in a later piece.
If you visit www.kenpom.com and keep it open in a separate tab while reading this article, it’d be helpful. We are going to explain each column of the rankings, left to right, as follows:
· Strength of Schedule:
· NCSOS AdjEM
Something to keep in mind as we go down this list: many of the metrics we’re discussing are “per 100 possessions”. On average, a college basketball team gets 60-80 possessions in a game. So if you were to look at the KenPom numbers deeply and do the math on something like points scored (AdjO), you might think “wow, these numbers are ridiculously high” – it’s because you can’t think of them on a per-game basis. It’s more like “per one-and-a-third games.”
Adjusted Efficiency Margin (AdjEM): This is how KenPom determines the overall ranking of teams. The more positive, the better. This takes the offensive efficiency minus the defensive efficiency to determine by how many points a team would outscore the “average” Division I program by. As of the morning of December 8, 2019, Rutgers has a AdjEM value of +10.01, meaning if Rutgers were to play the hypothetical average Division I team, they’d be expected to win by 10 points.
Adjusted offensive efficiency (AdjO): This is the amount of points a team would score per 100 possessions, or trips down the floor with the basketball, against an average Division I opponent.
Adjusted defensive efficiency (AdjD): This is the amount of points a team would allow per 100 possessions, against an average Division I opponent.
Adjusted tempo (AdjT): We’re going to spend a little bit of extra time on this one.
It’s not enough to just take offensive and defensive efficiency metrics and spit out the numbers from there. KenPom also accounts for tempo, which is the amount of possessions that a team has per 40 minutes (over the course of one game). Remember, as mentioned above, in an average game the average team has 60-80 possessions – but that’s a wide range, and teams vary with respect to how they push (or slowly move) the ball down the court.
This is how KenPom says possessions in a game can be estimated using a box score:
Possessions per game = Field goals attempted - offensive rebounds + turnovers + 0.475 x free throws attempted
Possessions are counted for both teams and then averaged out to give us the AdjT metric.
This is something that people should look at as just a piece of the puzzle as opposed to something that is a damning stat about a team’s offensive prowess. For instance, Virginia had the 2nd-most efficient offense in the country last year despite being dead last in tempo. While teams certainly vary in the extent to which they use tempo to their advantage, efficiency is based on making the possessions that a team has count.
Luck rating (Luck): KenPom defines this as a measure of the deviation between a team’s actual winning percentage and what one would expect from its game-by-game efficiencies. Essentially, a team involved in a lot of close games should not win (or lose) all of them. Those that do will be viewed as lucky (or unlucky).
(If you’re a baseball stats nerd, think of this as a college basketball variation of the Pythagorean record, which assumes every team should go .500 in one-run games and measures luck as how reality varies from that assumption.)
Strength of Schedule: This measures the total efficiency of the opponents that a team has faced on the year. You’ll notice these three columns map the first three columns of the spreadsheet; they mean the exact same thing, but from a team’s opponents’ perspective.
Non-conference strength of schedule (NCSOS): KenPom attempts to paint a picture here of the portion of the schedule that a team’s athletic department can control, which obviously rewards a team that schedules tougher opponents as opposed to cupcakes in non-conference play. The AdjEm metric here (the final column of the spreadsheet) measures the point differential by which your opponents would defeat the average Division I school by.
Again, this doesn’t take into consideration the caliber of teams. Most non-conference schedules for Power 5 schools are fairly light with some bigger matchups. This is more a measure of how bad the worst teams you play are. They really should rename this part of it the cupcake metric.
What Can We Say About Rutgers Right Now?
At the present time, Rutgers is rated (ranked):
· AdjEM: +10.31 (74)
· AdjO: 103.6 (108)
· AdjD: 93.3 (56)
· AdjT: 69.8 (167)
· Luck: -.022 (207)
· Strength of Schedule:
o AdjEM: -4.18 (300)
o OppO: 98.1 (280)
o OppD: 102.3 (297)
· NCSOS AdjEM: -7.02 (334)
What does the above mean? It may match what you already know about the 2019-2020 Scarlet Knights so far. Rutgers is presently more efficient from a defensive perspective than offensive, though they are above average at both this season. Neither is a particular strength from a national perspective right now, though these rankings adjust multiple times per day and Rutgers’ identity may evolve as the conference slate begins.
From a tempo perspective, Rutgers is decidedly average and in line with last season. Though in Steve Pikiell’s first two years Rutgers played at a decidedly slower tempo, there’s been a conscious effort since 2018 to pick things up on offense. Rutgers will likely never be a speed-demon offense (and if you were to sort the nation’s teams on tempo, you wouldn’t want Rutgers to be one – the fastest-paced offenses in college basketball are usually run by smaller schools who can’t keep up from a size perspective).
Rutgers’ luck has been average so far, and their strength of schedule has been very weak – no surprise if you’ve been following the team closely.
KenPom can be intimidating and, to be totally honest, has limitations. Sometimes advanced stats in general do not tell the whole story. There are certain bounces of the ball or a shot going in and out that can change the complexion of an entire game. It is why this is arguably the best sport, especially when tournament time comes around. Anything can happen inside the vacuum of one game if shots are falling for one team and are not for the other (see, #16 UMBC vs. #1 Virginia as the most popular example!).
All that KenPom serves to do is create a profile of a team over an extended amount of time for background. There is more advanced stuff on the site that is a little more complicated to explain than what we went over here, but for the average fan, the above is a pretty good starting point, especially if you’re wanting to know what the full picture of a team looks like.