Make the forecast
earn your trust.

Held-out transition 2026 improved on the prior-net baseline by 0.50 points per 100 possessions in this edition. The result is one chronological season comparison, not a guarantee of future accuracy.

Roster releases are source snapshots without a verified pre-season publication clock.

The challenger is research-only and does not replace the primary forecast or enter the prospective ledger.

Only two historical roster transitions are available for fitting; the held-out evaluation is one season and is not a guarantee of future performance.

Roster listings do not establish eligibility, availability, transfer reason, injury status or depth-chart role.

Publisher Box BPM is unavailable for some source IDs; rows without prior BPM coverage are withheld from the challenger rather than imputed.

The scenario changes the primary margin by the learned team-strength delta but does not recalibrate win probability or uncertainty.

Does workload continuity travel across more seasons?

NCAA roster and player-season releases provide a separate workload-only challenger. It uses unique season name mappings to the ESPN schedule and never enters the production Box BPM scenario.

NCAA and ESPN team IDs are different namespaces; only unique season name mappings enter this replay.

The NCAA roster release has no publisher Box BPM, so this is a workload-only historical challenger and its scores are not comparable to the production feature set.

Roster listings are source snapshots and do not establish eligibility, availability, injury status or depth-chart role.

Sorted by the challenger's predicted net efficiency after the listed-roster inputs. This is a research scenario, not a national ranking, a depth chart or an availability claim. BPM is shown only when the source athlete IDs have an attributed publisher value.

Roster continuity uses publisher athlete IDs and prior recorded minutes. An absent source row is not treated as a departure, and a recruiting announcement is not treated as eligibility. Use these counts to decide where staff verification is still needed before turning historical production into a lineup assumption.

These are retrospective preseason predictions using source records as currently published. They do not prove an edge over sportsbooks. Final-score evaluation includes overtime, while predictions use regulation pace.

The model notebook currently has 583 held-out cold-start games for interval calibration; the public slate reports 50 exploratory estimates alongside 1,579 primary forecasts.