How our formula works
Two halves. 40% is our own formula, described below. 60% is the consensus of the established preseason rankings: KenPom, Torvik and ESPN. A team ranked k-th by the consensus gets the score our k-th best team has, so both halves sit on one scale.
Our formula uses only box-score data, and it’s built the same way the major ones are. 65% of the score is a power rating: how many more points a team scores than it allows per possession, the strength of the competition it does it against, and whether it actually won. The other 35% are style tilts toward defense, efficient offense, offensive rebounding and an experienced roster. They move teams within the power rating’s frame rather than overturn it. And there’s room for outliers: a team truly elite at one thing can crack the Top 25 on that alone.
| Component | What it measures | Weight |
|---|
| Net efficiency | Points scored minus allowed per 100 possessions: 50% from the projected 2026-27 roster, 50% from how the team actually played last season | 32% |
| Competition | Strength of the conference a team plays in | 23% |
| Winning | Last season’s win percentage. Programs that win tend to keep winning | 10% |
| Defensive efficiency | Points allowed per 100 possessions | 10% |
| Offensive efficiency | Points scored per 100 possessions | 9% |
| Offensive rebounding | Share of its own misses a team gets back | 7% |
| Roster experience | Last season’s minutes back with the same team, plus 50% credit for minutes transfers played elsewhere in Division I | 4% |
| Turnover margin | Turnovers forced minus committed, per 100 possessions | 3% |
| Three-point shooting | Three-point percentage | 2% |
Outliers: the identity bonus
Three kinds of team get lifted beyond their overall numbers, and they’re marked in the table. Elite offense and Elite defense are teams whose projected scoring, or stopping, is far above the rest of Division I. Wins the margins is the methodical team that out-rebounds and out-possesses opponents: offensive rebounding and turnover margin together, far above the field. A team qualifies when it’s more than 2.5 standard deviations above average at either, which is roughly the top 1% of the country, and each step past that adds to its score. Only its stronger identity counts.
- Start from last season’s box scores. Every Division I team’s 2025-26 totals: possessions, points, offensive and defensive rebounds, turnovers and three-pointers, for the team and its opponents.
- Adjust every stat for competition. Forcing 18 turnovers a game against weak ball-handlers isn’t the same as doing it against good ones. Each stat is corrected by how a team’s opponents did in that stat against everyone else, recalculated until the numbers settle. Then competition counts a second time, as its own 23% component: the average strength of the team’s conference. That’s what keeps a mid-major with gaudy numbers from outranking a power team that put up good ones against far better opponents.
- Follow the players to their 2026-27 teams. Every returning player and transfer brings his own offensive rating, offensive rebounding, turnovers and three-point shooting, weighted by how many minutes he played and corrected for the schedule his old team faced. Small samples are pulled toward average: 300 minutes of average play are blended into each player, and 200 average three-point attempts into each roster.
- Fill the unknown minutes with the program. Freshmen, international signings and players new to Division I have no college numbers. Their minutes are credited at 80% of how far above average the program was last season. Good programs replace talent with talent.
- Carry defense with the program. Box scores can’t pin defense or forced turnovers on individual players, and defense is the most coach-driven part of the game. So a team keeps 65% of last season’s defensive edge for its coach and system, plus up to 30% more depending on how many minutes return.
- Switch to real numbers as games are played. Once the season starts, each team’s official 2026–27 totals from NCAA.com are put through the same formula and blended with the projection, and the outside rankings fade out at the same pace: after 10 games they count half, and more after that. NCAA.com doesn’t publish schedules, so these in-season numbers are adjusted for the strength of a team’s conference rather than its exact opponents.
- Score and rank. Each component becomes a standard score across all Division I teams (how many standard deviations above or below average), multiplied by its weight and added up. A score of 0 is a typical team. The top 25 are ranked and the next 10 are honorable mentions. All of this reruns every morning.
- Our picks. We can place a team by hand when we see something the numbers don’t. Those teams are tagged “Our pick” with the spot the formula alone gave them, and everyone else keeps the formula’s order around them.
What it can’t see
Freshmen, including five-stars, count only through their program’s track record, so teams rebuilt around a top recruiting class can sit lower here than in the polls. Coaching changes aren’t modeled. Players confirmed out for the season, such as UConn’s Solo Ball and St. John’s Donnie Freeman, are removed. Because the formula leans on hustle stats, a team that wins with size and half-court defense but doesn’t force turnovers can rank a little lower here than in most polls. That’s deliberate.
Where the numbers come from
The preseason projection uses 2025-26 team and player box-score totals and 2026-27 rosters from CollegeBasketballData, which compiles official box scores. In-season numbers come straight from NCAA.com’s official team stats; checked against last season, they agree with the box scores on every stat used. We checked them against the conferences’ own published stats: turnover margins and three-point totals match the ACC and Big East figures. The roster source hasn’t listed newcomers for many teams yet, so incoming transfers for the best-known teams were added from ESPN’s preseason breakdown, using roster facts only. Last rebuilt Oct 27, 2026.