Stats Corner: NET vs KenPom Rankings

In 2019 after the NCAA tournament seeding was released N.C. State Athletic Director Debbie Yow release a statement “The NCAA NET calculation was introduced as the new ranking system to replace the RPI as the primary sorting tool for evaluating teams…Based on the metrics sited above that the NCAA indicated they would use to evaluate team performance, we are disappointed for our athletes, coaches and fans that our total body of work was not rewarded with selection to the NCAA Tournament.” And TCU coach Jamie Dixon said during his press conference “You look at the NET, there are six teams that had a lower NETs than us. They created a tool and talked about it and then there are six teams there that are lower than us that are in.” In the years since then, the NET system has been changed. The first ever Stats Corner article was explaining the NET system and four years later we are looking at the NET and KenPom systems and how they ranking the NCAA teams.

NET

NET is an acronym for NCAA Evaluation Tool and while the NCAA screws up a lot of things, they absolutely got this acronym correct. It is a predictive-learning model. Predictive learning is a method of machine “learning”. The machine attempts to create a model by simulating different outcomes in situations. These situations are numerous, with the computer comparing expected results with actual results and adjusting as needed. Think of it as a small child trying to solve a problem. They try something it doesn’t work, so they adjust based on what they have just observed and experienced. NET is not learning in the human sense of the word, the purpose is to use the known effects of actions and create planning operators, or in non-geek, it is more like discovering which models produce the most realistic results by comparing them to actual results. It is not artificial intelligence; the robotic overloads are not coming (yet).

When the NET was introduced, there was five factors it considered: Team value index, Net efficiency, winning percentage, adjusted win percentage, and scoring margin. The team value index considered 3 factors opponent, location, and winner. Net efficiency was team offensive efficiency minus defense efficiency. Adjusted win percentage gave added points for winning on the road and took points away for losing at home, in short where you played mattered. Winning percentage and scoring margin are self-explanatory, only note is scoring margin was capped at 10 points.

Also, there are four quadrants based on where your opponent is ranked, you will hear about Quad 1 and Quad 2 records. For example, San Diego State is ranked 21^st and is 2-3 in Quad 1, while Utah State is 24^th and is 0-1 in Quad 1 and 2-2 Quad 4. This means that while the Aztecs do have 3 loses, they are to high caliber teams and will not affect their seeding in the tournament. The Aggies however, lost their only Quad 1 game (Boise State ranked 25), and have lost 2 games they should have won in Quad 4. The divisions for the quadrants are:

Quadrant 1: Home 1-30; Neutral 1-50; Away 1-75.

Quadrant 2: Home 31-75; Neutral 51-100; Away 76-135.

Quadrant 3: Home 76-160; Neutral 101-200; Away 136-240.

Quadrant 4: Home 161-plus; Neutral 201-plus; Away 241-plus.

Wyoming is ranked 205, so a Boise State-Wyoming game would be a Q1 game for Wyoming regardless of where it was played, if it at Boise, or neutral (say MWC tournament), it would be a Q4 game for the Broncos, if it was played at Wyoming it would be a Q3 game for the Broncos.

Beginning in 2020, winning percentage, adjusted winning percentage and scoring margin were all dropped from the NET ranking. Therefore, only team value and team value index are used to ranked teams. So, the only factors which go into the rankings are efficiency (offensive points per possession minus opponents points per possession), strength of opponents played, location of game, wins and losses, and who won the game.

KenPom

Created by Ken Pomeroy, the KenPom rankings are independent of the NCAA. He has his own autonomous website. He ranks every NCAA team by adjusted efficiency margin which is adjusted offensive efficiency minus adjusted defensive efficiency. Before you say “that’s what the NET system does” two things. First, KenPom came first. Second, it is ADJUSTED efficiency. Team efficiency is how many points are scored per 100 possessions. The adjusted factor for KenPom is based on how many possessions a team has per game. St Mary’s has a team offensive efficiency of 107.0 (they would score 107 points per 100 possessions) which is 48^th in the country, however they have an adjusted team efficiency of 109.4, indicating they play at a slower pace so they, and their opponents, do not have as many possessions per game. Therefore St. Mary’s are more effective with their fewer per game possessions. If you go to KenPom’s website you will see his different metrics he uses going from left to right, we will use the highest ranked MWC, San Diego State #21, as an example.

Adjusted Efficiency Margin (AdjEM): This is the final ranking, everything else go into this. The higher the score the higher ranking. In short, it’s the offensive efficiency minus the defensive efficiency. At 21^st the Aztecs have a score of 18.39, which means they would beat an NCAA team by an average of 18 points.

Adjusted offensive efficiency (AdjO) and Adjusted defensive efficiency (AdjD) have been covered. For the Aztecs AdjO is 110.5 (43^rd) and AdjD is 92.1 (20^th).

Adjusted tempo (AdjT): The amount of possessions a team has per game. There is a formula for it: Possessions per game = Field goals attempted - offensive rebounds + turnovers + 0.475 x attempted free throws. This is used to equalize the scores between run and shoot teams with lots of possessions and teams which grind it out and play at a slower pace. San Diego State is at 69.0 which is 114^th in the country, so they place a little faster than the average.

Luck rating: This is your close game factor. Statistically, a team should go .500 in one possession games. The luck rating is the deviation between a team’s actual winning percentage and the expected from the above metrics. If teams are winning more of the close games they are considered “lucky”. The Aztecs are lucky at +.043 which is 98^th in the country. FYI, 5-10 Stanford is the unluckiest team in the country at -.191.

Strength of Schedule: This is the average AdjEM, AdjO, and AdjD of the opponents. The better the opponents the higher the SOS. San Diego State’s opponents have an AdjEM of 6.67 which is the 18^th toughest schedule in the country, as the Aztecs opponents are winning their games by an average of 7 points.

Non-conference strength of schedule (NCSOS): The conference schedule is out of the college’s control, they non-conference schedule is something the college controls. This section is a reward for putting together a more difficult non-conference schedule instead of playing cupcakes. The Aztecs had a good non-conference schedule at 22^nd with an AdjEM of 6.73. Compared to Air Force who has the 362^nd NCSOS at -10.62, meaning their non-conference opponents lost by 11 points.

While both methods use similar data, the biggest differences are

1) NET uses game location as a factor in the ranking system, while KenPom does not factor where the game was played in rankings.

2) Ken Pom uses adjusted efficiency based on how many possessions a team has per game while NET uses points per 100 possessions regardless of possession per game.

3) Scoring margin does not matter to the NET, but it does to KenPom.

4) KenPom does not factor wins and losses, it is simply a efficiency number.

5) KenPom uses strict formulas (predictive) to create ranking while NET uses a learning algorithm comparing what is expected to what happens.