Learn
The Per-100 Rule: college basketball points per possession, on one page
College basketball points per possession is the fairest way to compare a fast team with a slow one, and the Per-100 Rule turns it into one habit: count possessions, then score every team per 100 of them. Raw points per game reward pace, so a team that runs can look like the better offense when it isn't. This sheet gives you the possession formula, the per-100 math, a worked hypothetical and a checklist you can copy into a spreadsheet.
This is the sheet from the video. The video makes one point, and this page gives you the tools to use it.
The rule from the video, in order:
- Score every team per 100 possessions. Raw points per game mix up how good an offense is with how fast it plays.
- The hypothetical. 84 points on 70 possessions is 120 per 100. 90 points on 80 possessions is about 112.
- The aha. The 90-point team looks better, but per possession the 84-point team was about 7.5 points better.
Everything below goes one level deeper on each of those.
Who this is for
You build or check a college basketball model, or you just want to read a box score without being fooled by pace. You need a spreadsheet and the basic box score for each game: field goal attempts, offensive rebounds, turnovers, free throw attempts and points. No paid data is required to start.
The field is big. Division I has 365 men's basketball teams in 32 conferences, and they play at very different speeds. That is exactly why the best-known public ratings are tempo-free. Ken Pomeroy's ratings, which cover all 365 Division I programs, are built on offensive and defensive efficiency and tempo, and they explain games possession by possession instead of game by game.
Step 1, count the possessions
A possession ends when a team shoots and doesn't get the ball back, turns it over, or goes to the line. You can estimate the count from the box score with one formula:
Possessions = FGA - OR + TO + 0.475 x FTA
KenPom calls this the standard computation and explains each part. Offensive rebounds are subtracted because they continue a possession instead of starting a new one. The 0.475 on free throw attempts is the one uncertain piece, since not every trip to the line ends a possession (and-ones and the first of two shots don't), and its value comes from play-by-play data. The UMHoops tempo-free glossary uses the same formula and calls it what it is: an estimate, but a good one, because it adds up the plays that can end a possession.
Two habits make the count better:
- Average both teams. Count possessions for each side of the game and average the two. In a single game the teams should be within about one possession of each other, so a big gap means a data error.
- Use the same formula everywhere. A model that mixes formulas across seasons or sources will drift for reasons that have nothing to do with basketball.
Step 2, score per 100 possessions
Once you have possessions, efficiency is simple:
Offensive efficiency = points / possessions x 100
The same glossary defines it as points per 100 possessions. Defensive efficiency is the same math on the points a team allows. Tempo is possessions per 40 minutes, which you keep as its own column because it tells you how many chances a game will have, not how good either team is.
Step 3, run the worked example
This is the hypothetical from the video. The teams are made up, and the numbers are chosen to make the arithmetic easy to check.
| Points | Possessions | Points per 100 | |
|---|---|---|---|
| Demo Team A (hypothetical) | 84 | 70 | 120.0 |
| Demo Team B (hypothetical) | 90 | 80 | 112.5 |
| Gap, A minus B | -6 | -10 | +7.5 |
Read the table from left to right and the story flips. On points, Team B scored six more. On possessions, Team B also had ten more chances. Per 100 possessions, Team A was 7.5 points better, because it did more with each trip down the floor. A model that rated these two teams on points per game would have the order backwards.
Step 4, put it in a sheet
Copy these columns into a spreadsheet, one row per team per game:
- Team, opponent, date, home or away.
- FGA, OR, TO, FTA, points for the team.
- The same five for the opponent.
- Possessions (team) = FGA - OR + TO + 0.475 x FTA.
- Possessions (opponent) with the same formula.
- Game possessions = the average of the two.
- Offense per 100 = team points / game possessions x 100.
- Defense per 100 = opponent points / game possessions x 100.
- Net per 100 = offense minus defense.
Average columns 7 to 9 across a team's games and you have a basic tempo-free rating. To predict a game, estimate how many possessions it will have from both teams' tempo, then convert each side's expected efficiency into points for that many possessions.
What to watch out for
- Opponent strength. A raw per-100 average still rewards a soft schedule. Real ratings adjust each game for the opponent, which matters a lot in a 365-team field where top and bottom teams rarely meet.
- Small samples early. In November a team has a handful of games, so one hot shooting night moves its rating a lot. Blend in a preseason estimate until the season fills in.
- Garbage time and blowouts. Late possessions in a decided game tell you less. Some modelers trim or down-weight them.
- Pace still matters for totals. Per-100 ratings fix the comparison between teams, but a game total still depends on how many possessions there will be. Keep tempo in the model, just don't confuse it with quality.
How to tell it worked
Your ratings are doing their job when fast teams stop sitting at the top just because they score a lot, and when the ratings line up roughly with the public tempo-free rankings you can check for free. After that, the real test is prices. At -110 you need to win about 52.38 percent of bets just to break even, so compare your predicted margins with the market and track how your numbers do against the closing line over a full season before trusting them with money.
For the bigger picture on building one, see the college basketball betting model page, and for how a model should be graded, the sports betting model scorecard page. The fair odds calculator turns a predicted win chance into a price you can compare with the books.
Quant+ is analytics for informational purposes. 21+. Nothing is guaranteed. If you or someone you know has a gambling problem, call 1-800-GAMBLER.
Questions people ask
What is a good points per possession number in college basketball?
It depends on the season and the opponent, so compare a team with the Division I average for that year rather than a fixed number. A team scoring well above the average per 100 possessions is efficient whatever its pace.
Why multiply free throw attempts by 0.475?
Not every free throw attempt ends a possession. The second of two shots usually does, the first doesn't, and an and-one doesn't start a new one. The 0.475 factor is an estimate of how many attempts end a possession, derived from play-by-play data.
Is points per possession the same as offensive efficiency?
Offensive efficiency is usually points per possession multiplied by 100, so 1.20 points per possession is an efficiency of 120. It is the same idea on a friendlier scale.
Does a per-100 rating predict the winner of a game?
It predicts which team is better per trip, which is the core of a margin prediction. You still need the expected number of possessions, home court, and an opponent adjustment, and even a good prediction loses often. Nothing is guaranteed.
Sources
- NCAA Division I, Wikipedia
- Ken Pomeroy, Wikipedia
- Ratings glossary, kenpom.com
- Tempo Free Stats Glossary, UMHoops
- Moneyline odds, Wikipedia
