Changelog

Unreleased

Added

SOS rank column on /rankings and /team/[slug]: (2·WP2 + WP3) / 3, ranked across D1 (1 = toughest). Exposed as sos_rank in /api/rankings.
Series view on /scoreboard groups games into multi-game series and shows leader, status, and per-game scores at a glance.

Changed

Scenarios is forward-only. The flip-past-result toggle is gone.
Tiered revalidation: live pages (/, /scoreboard, /api/rankings) revalidate every 5 min; /methodology hourly; /scenarios and /path every 30 min; the rest every 10 min.
Centralized site URL via web/lib/site-config.ts. NEXT_PUBLIC_SITE_URL overrides the default Vercel domain for metadataBase, robots.txt, and sitemap.xml.

Removed

/whatif → /scenarios redirect (route was renamed months ago).
Past-result flip mode in scenarios.

Pre-Unreleased

Changed

Game-data source moved from Warren Nolan team-schedule pages to the NCAA contests GraphQL endpoint (sdataprod.ncaa.com). One HTTP call per date instead of ~300 per-team page scrapes; full coverage across all 29 D1 conferences. The Edge Function is now single-shot rather than chunked.
Game identifier moved from ESPN event_id to NCAA contest_id. Box-score links go to ncaa.com/game/<id>. Schema migration 0008_games_ncaa_contest_id.sql adds the column and drops espn_event_id.
Edge function rewritten as a single-shot pipeline. The chunked phase RPCs (start_scrape / append_chunk / finalize_scrape) stay in migrations in case we need them again.
Python scraper: new cbb_hub.ncaa_contests module and scrape-ncaa CLI subcommand. Seeds team IDs from the existing season JSON to keep Supabase foreign keys stable.

Removed

ESPN scoreboard ingest (src/cbb_hub/espn.py, supabase/functions/_shared/espn.ts) and its tests.

Added

/scoreboard. Yesterday's finals, today's slate, tomorrow's matchups, with date navigation. Ranked games sort to the top.
/standings. Conference standings by league: conference record, overall record, canonical rank.
Year-round homepage. The hero copy and quick links shift between preseason, regular season, conference week, postseason, and offseason based on the snapshot itself. Adds a wire strip for yesterday, today, and on-deck.
lib/season-phase.ts. Derives the calendar moment from the snapshot (first/last/next game) so copy stays correct as the season rolls.
Brand: Field of 64. Wordmark, mark, type system, color tokens, OG and favicon assets as Next.js routes.
Design system: Container, PageHeader, Card, Badge, Button, Stat, TeamCell primitives.
Canonical rank anchored to the NCAA's published RPI. Each scrape persists published_ranks alongside teams and games; every page orders by that rank rather than locally-computed base RPI.
Live accuracy panel on /methodology: top-25 match count, exact-rank matches, mean and median absolute rank delta vs the published index.
lib/season-state.ts is a single React.cache()-backed view over season JSON, RPI computation, ranks, and records.
Public JSON API returns rank source, computed rank, and a parity summary alongside the rankings.

Changed

Methodology page replaced its static "23/25 overlap" claim with a live accuracy panel.
Scraper signature: scrape_season returns (teams, games, published_ranks); save_season accepts a published_ranks keyword.
Pages no longer call loadSeason / calculateRPI / rankByRPI directly; they consume getSeasonState() from lib/season-state.ts.

Removed

Per-page boilerplate from create-next-app.

0.1.0 (2026-05-18)

Added

Python RPI engine with hand-verified unit tests covering venue weighting, self-exclusion in WP2, and D1-only filtering.
TypeScript port of the engine, parity-tested against Python (1e-9 agreement across all 308 teams).
Warren Nolan scraper with rate-limited fetcher, on-disk HTML cache, and parser tests against pinned fixtures.
Season JSON I/O and CLI: scrape, rank, validate.
Next.js (App Router, Turbopack) frontend: top 25, full rankings, bracket projection, what-if simulator.
GitHub Actions workflow that scrapes every 30 minutes during the season and commits refreshed data when it changes.
Repo hygiene: LICENSE, CONTRIBUTING, CODE_OF_CONDUCT, SECURITY, issue and PR templates, Dependabot, CodeQL.
Sitewide metadata, sitemap, robots.txt, PWA manifest, skip link.

Changelog

Unreleased

Added

SOS rank column on /rankings and /team/[slug]: (2·WP2 + WP3) / 3, ranked across D1 (1 = toughest). Exposed as sos_rank in /api/rankings.
Series view on /scoreboard groups games into multi-game series and shows leader, status, and per-game scores at a glance.

Changed

Scenarios is forward-only. The flip-past-result toggle is gone.
Tiered revalidation: live pages (/, /scoreboard, /api/rankings) revalidate every 5 min; /methodology hourly; /scenarios and /path every 30 min; the rest every 10 min.
Centralized site URL via web/lib/site-config.ts. NEXT_PUBLIC_SITE_URL overrides the default Vercel domain for metadataBase, robots.txt, and sitemap.xml.

Removed

/whatif → /scenarios redirect (route was renamed months ago).
Past-result flip mode in scenarios.

Pre-Unreleased

Changed

Game-data source moved from Warren Nolan team-schedule pages to the NCAA contests GraphQL endpoint (sdataprod.ncaa.com). One HTTP call per date instead of ~300 per-team page scrapes; full coverage across all 29 D1 conferences. The Edge Function is now single-shot rather than chunked.
Game identifier moved from ESPN event_id to NCAA contest_id. Box-score links go to ncaa.com/game/<id>. Schema migration 0008_games_ncaa_contest_id.sql adds the column and drops espn_event_id.
Edge function rewritten as a single-shot pipeline. The chunked phase RPCs (start_scrape / append_chunk / finalize_scrape) stay in migrations in case we need them again.
Python scraper: new cbb_hub.ncaa_contests module and scrape-ncaa CLI subcommand. Seeds team IDs from the existing season JSON to keep Supabase foreign keys stable.

Removed

ESPN scoreboard ingest (src/cbb_hub/espn.py, supabase/functions/_shared/espn.ts) and its tests.

Added

/scoreboard. Yesterday's finals, today's slate, tomorrow's matchups, with date navigation. Ranked games sort to the top.
/standings. Conference standings by league: conference record, overall record, canonical rank.
Year-round homepage. The hero copy and quick links shift between preseason, regular season, conference week, postseason, and offseason based on the snapshot itself. Adds a wire strip for yesterday, today, and on-deck.
lib/season-phase.ts. Derives the calendar moment from the snapshot (first/last/next game) so copy stays correct as the season rolls.
Brand: Field of 64. Wordmark, mark, type system, color tokens, OG and favicon assets as Next.js routes.
Design system: Container, PageHeader, Card, Badge, Button, Stat, TeamCell primitives.
Canonical rank anchored to the NCAA's published RPI. Each scrape persists published_ranks alongside teams and games; every page orders by that rank rather than locally-computed base RPI.
Live accuracy panel on /methodology: top-25 match count, exact-rank matches, mean and median absolute rank delta vs the published index.
lib/season-state.ts is a single React.cache()-backed view over season JSON, RPI computation, ranks, and records.
Public JSON API returns rank source, computed rank, and a parity summary alongside the rankings.

Changed

Methodology page replaced its static "23/25 overlap" claim with a live accuracy panel.
Scraper signature: scrape_season returns (teams, games, published_ranks); save_season accepts a published_ranks keyword.
Pages no longer call loadSeason / calculateRPI / rankByRPI directly; they consume getSeasonState() from lib/season-state.ts.

Removed

Per-page boilerplate from create-next-app.

0.1.0 (2026-05-18)

Added

Python RPI engine with hand-verified unit tests covering venue weighting, self-exclusion in WP2, and D1-only filtering.
TypeScript port of the engine, parity-tested against Python (1e-9 agreement across all 308 teams).
Warren Nolan scraper with rate-limited fetcher, on-disk HTML cache, and parser tests against pinned fixtures.
Season JSON I/O and CLI: scrape, rank, validate.
Next.js (App Router, Turbopack) frontend: top 25, full rankings, bracket projection, what-if simulator.
GitHub Actions workflow that scrapes every 30 minutes during the season and commits refreshed data when it changes.
Repo hygiene: LICENSE, CONTRIBUTING, CODE_OF_CONDUCT, SECURITY, issue and PR templates, Dependabot, CodeQL.
Sitewide metadata, sitemap, robots.txt, PWA manifest, skip link.

Changelog

Unreleased

Added

SOS rank column on /rankings and /team/[slug]: (2·WP2 + WP3) / 3, ranked across D1 (1 = toughest). Exposed as sos_rank in /api/rankings.
Series view on /scoreboard groups games into multi-game series and shows leader, status, and per-game scores at a glance.

Changed

Scenarios is forward-only. The flip-past-result toggle is gone.
Tiered revalidation: live pages (/, /scoreboard, /api/rankings) revalidate every 5 min; /methodology hourly; /scenarios and /path every 30 min; the rest every 10 min.
Centralized site URL via web/lib/site-config.ts. NEXT_PUBLIC_SITE_URL overrides the default Vercel domain for metadataBase, robots.txt, and sitemap.xml.

Removed

/whatif → /scenarios redirect (route was renamed months ago).
Past-result flip mode in scenarios.

Pre-Unreleased

Changed

Game-data source moved from Warren Nolan team-schedule pages to the NCAA contests GraphQL endpoint (sdataprod.ncaa.com). One HTTP call per date instead of ~300 per-team page scrapes; full coverage across all 29 D1 conferences. The Edge Function is now single-shot rather than chunked.
Game identifier moved from ESPN event_id to NCAA contest_id. Box-score links go to ncaa.com/game/<id>. Schema migration 0008_games_ncaa_contest_id.sql adds the column and drops espn_event_id.
Edge function rewritten as a single-shot pipeline. The chunked phase RPCs (start_scrape / append_chunk / finalize_scrape) stay in migrations in case we need them again.
Python scraper: new cbb_hub.ncaa_contests module and scrape-ncaa CLI subcommand. Seeds team IDs from the existing season JSON to keep Supabase foreign keys stable.

Removed

ESPN scoreboard ingest (src/cbb_hub/espn.py, supabase/functions/_shared/espn.ts) and its tests.

Added

/scoreboard. Yesterday's finals, today's slate, tomorrow's matchups, with date navigation. Ranked games sort to the top.
/standings. Conference standings by league: conference record, overall record, canonical rank.
Year-round homepage. The hero copy and quick links shift between preseason, regular season, conference week, postseason, and offseason based on the snapshot itself. Adds a wire strip for yesterday, today, and on-deck.
lib/season-phase.ts. Derives the calendar moment from the snapshot (first/last/next game) so copy stays correct as the season rolls.
Brand: Field of 64. Wordmark, mark, type system, color tokens, OG and favicon assets as Next.js routes.
Design system: Container, PageHeader, Card, Badge, Button, Stat, TeamCell primitives.
Canonical rank anchored to the NCAA's published RPI. Each scrape persists published_ranks alongside teams and games; every page orders by that rank rather than locally-computed base RPI.
Live accuracy panel on /methodology: top-25 match count, exact-rank matches, mean and median absolute rank delta vs the published index.
lib/season-state.ts is a single React.cache()-backed view over season JSON, RPI computation, ranks, and records.
Public JSON API returns rank source, computed rank, and a parity summary alongside the rankings.

Changed

Methodology page replaced its static "23/25 overlap" claim with a live accuracy panel.
Scraper signature: scrape_season returns (teams, games, published_ranks); save_season accepts a published_ranks keyword.
Pages no longer call loadSeason / calculateRPI / rankByRPI directly; they consume getSeasonState() from lib/season-state.ts.

Removed

Per-page boilerplate from create-next-app.

0.1.0 (2026-05-18)

Added

Python RPI engine with hand-verified unit tests covering venue weighting, self-exclusion in WP2, and D1-only filtering.
TypeScript port of the engine, parity-tested against Python (1e-9 agreement across all 308 teams).
Warren Nolan scraper with rate-limited fetcher, on-disk HTML cache, and parser tests against pinned fixtures.
Season JSON I/O and CLI: scrape, rank, validate.
Next.js (App Router, Turbopack) frontend: top 25, full rankings, bracket projection, what-if simulator.
GitHub Actions workflow that scrapes every 30 minutes during the season and commits refreshed data when it changes.
Repo hygiene: LICENSE, CONTRIBUTING, CODE_OF_CONDUCT, SECURITY, issue and PR templates, Dependabot, CodeQL.
Sitewide metadata, sitemap, robots.txt, PWA manifest, skip link.

Changelog

Unreleased

Added

SOS rank column on /rankings and /team/[slug]: (2·WP2 + WP3) / 3, ranked across D1 (1 = toughest). Exposed as sos_rank in /api/rankings.
Series view on /scoreboard groups games into multi-game series and shows leader, status, and per-game scores at a glance.

Changed

Scenarios is forward-only. The flip-past-result toggle is gone.
Tiered revalidation: live pages (/, /scoreboard, /api/rankings) revalidate every 5 min; /methodology hourly; /scenarios and /path every 30 min; the rest every 10 min.
Centralized site URL via web/lib/site-config.ts. NEXT_PUBLIC_SITE_URL overrides the default Vercel domain for metadataBase, robots.txt, and sitemap.xml.

Removed

/whatif → /scenarios redirect (route was renamed months ago).
Past-result flip mode in scenarios.

Pre-Unreleased

Changed

Game-data source moved from Warren Nolan team-schedule pages to the NCAA contests GraphQL endpoint (sdataprod.ncaa.com). One HTTP call per date instead of ~300 per-team page scrapes; full coverage across all 29 D1 conferences. The Edge Function is now single-shot rather than chunked.
Game identifier moved from ESPN event_id to NCAA contest_id. Box-score links go to ncaa.com/game/<id>. Schema migration 0008_games_ncaa_contest_id.sql adds the column and drops espn_event_id.
Edge function rewritten as a single-shot pipeline. The chunked phase RPCs (start_scrape / append_chunk / finalize_scrape) stay in migrations in case we need them again.
Python scraper: new cbb_hub.ncaa_contests module and scrape-ncaa CLI subcommand. Seeds team IDs from the existing season JSON to keep Supabase foreign keys stable.

Removed

ESPN scoreboard ingest (src/cbb_hub/espn.py, supabase/functions/_shared/espn.ts) and its tests.

Added

/scoreboard. Yesterday's finals, today's slate, tomorrow's matchups, with date navigation. Ranked games sort to the top.
/standings. Conference standings by league: conference record, overall record, canonical rank.
Year-round homepage. The hero copy and quick links shift between preseason, regular season, conference week, postseason, and offseason based on the snapshot itself. Adds a wire strip for yesterday, today, and on-deck.
lib/season-phase.ts. Derives the calendar moment from the snapshot (first/last/next game) so copy stays correct as the season rolls.
Brand: Field of 64. Wordmark, mark, type system, color tokens, OG and favicon assets as Next.js routes.
Design system: Container, PageHeader, Card, Badge, Button, Stat, TeamCell primitives.
Canonical rank anchored to the NCAA's published RPI. Each scrape persists published_ranks alongside teams and games; every page orders by that rank rather than locally-computed base RPI.
Live accuracy panel on /methodology: top-25 match count, exact-rank matches, mean and median absolute rank delta vs the published index.
lib/season-state.ts is a single React.cache()-backed view over season JSON, RPI computation, ranks, and records.
Public JSON API returns rank source, computed rank, and a parity summary alongside the rankings.

Changed

Methodology page replaced its static "23/25 overlap" claim with a live accuracy panel.
Scraper signature: scrape_season returns (teams, games, published_ranks); save_season accepts a published_ranks keyword.
Pages no longer call loadSeason / calculateRPI / rankByRPI directly; they consume getSeasonState() from lib/season-state.ts.

Removed

Per-page boilerplate from create-next-app.

0.1.0 (2026-05-18)

Added

Python RPI engine with hand-verified unit tests covering venue weighting, self-exclusion in WP2, and D1-only filtering.
TypeScript port of the engine, parity-tested against Python (1e-9 agreement across all 308 teams).
Warren Nolan scraper with rate-limited fetcher, on-disk HTML cache, and parser tests against pinned fixtures.
Season JSON I/O and CLI: scrape, rank, validate.
Next.js (App Router, Turbopack) frontend: top 25, full rankings, bracket projection, what-if simulator.
GitHub Actions workflow that scrapes every 30 minutes during the season and commits refreshed data when it changes.
Repo hygiene: LICENSE, CONTRIBUTING, CODE_OF_CONDUCT, SECURITY, issue and PR templates, Dependabot, CodeQL.
Sitewide metadata, sitemap, robots.txt, PWA manifest, skip link.