Reference

Provisional Members

Whether a game against a provisional or reclassifying member counts toward the NCAA Power Index (NPI), and whether that member can reach the NCAA postseason, depends on the member's stage.

• NPI: Games against third-year provisional members and third-year reclassifying-into-Division III members count. Games against first- and second-year members, all members reclassifying out of Division III, and exploratory schools do not count.
• Postseason: No provisional or reclassifying member is eligible for NCAA championships in any year. Eligibility begins only after the school converts to active membership, which follows all three provisional years.

The current cycle (2026-27):

Sources: NCAA Provisional & Reclassifying Selection Criteria, NCAA D3 Provisional Member Process FAQ, d3hoops NCAA Tournament FAQ.

NPI "Settling" Criteria

NPI is computed by iteration. A team's rating is the average of its contributing game NPIs, and each game's NPI depends on the opponent's rating from the previous pass. Because every rating depends on opponents' ratings, there is no closed form: the calculation repeats over the full set of teams until the ratings stabilize. "Settling" means the iteration reaches a stable set of ratings (a fixed point), where another pass leaves every rating essentially unchanged.

The D3 Stat Lab's implementation: this site stops iterating once the average per-team change between passes falls below 0.0001, with a minimum of 10 passes and a maximum of 80. The NCAA does not publish how many iterations its official NPI runs; it most likely uses a fixed count set high enough that convergence is reached well before the final pass (for example, around 100), rather than testing for convergence directly. That difference does not change the result in practice: once the ratings have settled, further passes do not move them, so a convergence-threshold stop and a fixed high-count stop arrive at the same ratings, as long as the system settles at all (the condition below).

How readily it settles is governed by how strongly each team's rating amplifies changes in its opponents' ratings. For one team that amplification is:

α = SW + QBM × (quality wins ÷ contributing games)

• The system is guaranteed to settle when the largest α across all teams stays below 1. This is sufficient, not necessary.
• A schedule-proof bound is SW + QBM ≤ 1, equivalently QBM ≤ RW. At SW = 0.80 that caps QBM at 0.20.
• Per-team breakpoint: a team's α passes 1 once more than RW ÷ QBM of its contributing games are quality wins.

The default dials for women's basketball:

At these defaults the breakpoint is 0.30, so any team whose contributing games are more than 30% quality wins has α above 1. The system still settles because the quantity that actually governs convergence, the spectral radius of the iteration's Jacobian, stays below 1 given real schedule structure rather than because of margin in the α bound. A "contributing game" is one that feeds a team's rating: its top wins plus games at or below the team's own current rating.