SSILVERMINECollege basketball stats
LIVE BOARD
2026–27 SEASON
Research / The public recordEdition Sep 15, 2026
A prediction needs a memory.

Put the forecast
on the record.

Follow what the model said before the game, what happened next, and how it compares with observed betting lines.

Read the selection rules ↓
Registered forecasts61,722

Football + men’s basketball


Original estimates retained.
Changed results recorded.
Missing evidence explained.

Loading the scorecard…

The protocol / first-eligible-registration-v1

Same games. Honest clocks.

How forecasts enter the record

Each model/game estimate is saved once with its generation time, first local registration time, model cutoff and scheduled start. Later runs preserve the original estimate. The scorecard chooses the first eligible registration per game, before looking at its result.

A confirmed start is required. A changed start time or participant identity excludes that version; a new eligible registration can qualify later. Excluded forecasts remain searchable. All games are included in their sport’s prospective evaluation, including games with no betting line.

These results are separate from the retrospective tests in the football and basketball model notebooks.

How market comparisons qualify

The quote must be captured after forecast registration and before the confirmed start, with a provider update time no later than capture. Quotes older than 24 hours when captured are excluded. Participant identities and scheduled start must match exactly.

We use the last qualifying observation per provider, bookmaker and market. It is not a verified closing line. Matched-game errors are reported separately for each bookmaker; they are not compared with model errors from a different set of games.

Home spread difference = projected home margin + home spread. Total difference = projected total − observed total. Moneyline probabilities remove the two-way overround by normalizing inverse decimal prices.

Evidence still has limits.

Registration times are local pipeline observations, not independently notarized publication times.

Changed participants or start times, unconfirmed start times and late registrations are excluded.

Quotes older than 24 hours when captured are excluded; last observed quotes may still be stale today.

Settlements use latest source finals, including overtime; source corrections can revise reported scores.

Source finals are not official bookmaker settlements. Direction results are hypothetical, without execution, odds or fees.

Live odds collection uses a separately configured licensed feed. The existing SportsDataverse betting archive lacks the timestamps needed for this scorecard and is excluded. Zero observed comparisons means unavailable evidence, not a zero model edge.

Feed documentation: The Odds API v4 · Data-use terms.

Browse the historical market archive →