cfb_schedules

College football models and reference tables — the game schedule, one row per game.

26 seasons (2001–2026) · 78 assets · 4.3 MB · last asset written 2026-08-27

Upstream source: CFBD + ESPN, scored through the cfbfastR model stack.

Dataset notes

_{Written by hand for this dataset and preserved verbatim.}

College football schedules (CFBD-sourced), one asset trio per season.

Seasons 2001-2026, cfb_schedules_{season}.{parquet,rds,csv.gz}.

Rebuilt 2026-08-27 to fix sportsdataverse-py#9: seasons 2001-2022 had been built when cfbfastR::cfbd_game_info() defaulted to season_type = "regular", so every bowl and playoff game was missing. The current default is "both"; all seasons are now refetched with it.

Schema note: CFBD's nested playoff bracket object is flattened to playoff_* scalar columns (competition, format, round, roundName, bracketSlot, homeSeed, awaySeed, bowlName), so all 26 seasons share one 39-column schema.

Files in this release

Format	Files	Name pattern
csv.gz	26	cfb_schedules_{season}.csv.gz
parquet	26	cfb_schedules_{season}.parquet
rds	26	cfb_schedules_{season}.rds

Direct download (no auth, no API key):

https://github.com/sportsdataverse/sportsdataverse-data/releases/download/cfb_schedules/cfb_schedules_2026.parquet

How to load it

Surface	Call
Python	from sportsdataverse.cfb import load_cfb_schedule load_cfb_schedule(seasons=[2026])
R	cfbfastR::load_cfb_schedules(seasons = 2026)
SQL	SELECT * FROM cfb.schedule WHERE season = 2026 in the sportsdataverse Postgres warehouse
HTTP	GET https://data.sportsdataverse.org/v1/cfb/schedule?season=2026 (bearer token required)
Any	pd.read_parquet("https://github.com/sportsdataverse/sportsdataverse-data/releases/download/cfb_schedules/...") — the assets are plain files on a public URL

How it is produced

1. Capture — cfbfastR-cfb-raw: the same ESPN capture that feeds espn_cfb_*.
2. Build — cfbfastR-cfb-data: python/cfb_model_build/ (ratings, FPI, recruiting, model artifacts).
3. Publish — python/cfb_model_build/cfb_model_publish/cli.py --tag <this tag> uploads the assets to this tag.

The code that writes these files:

• python/cfb_data_build/schedules_unified.py  _{cfbfastR-cfb-data}
• python/espn_cfb_01_schedules_scrape.py  _{cfbfastR-cfb-raw}
• python/scrape_cfb_schedules.py  _{cfbfastR-cfb-raw}
• R/espn_cfb_15_team_summaries_creation.R  _{cfbfastR-cfb-data}

Automation

GitHub Actions — scheduled workflows in the producing repositories:

Repo	Workflow	Name	Schedule (UTC)	Last scheduled run
cfbfastR-cfb-data	cfb_model_pipeline.yml	CFB Model Pipeline	06:00 UTC, Feb (days 5)	—
cfbfastR-cfb-data	cfb_ratings_cron.yml	CFB Ratings	13:00 UTC, Aug 13:00 UTC, Sep-Nov 13:00 UTC, Dec 13:00 UTC, Jan (days 1-25)	✅ ok · last run 23h ago
cfbfastR-cfb-data	cfb_recruiting_proj_cron.yml	CFB Recruiting Projections	12:00 UTC, Dec (days 5) 12:00 UTC, Jan-Aug (days 5)	—
cfbfastR-cfb-data	daily_cfb.yml	Update CFB Data	11:00 UTC, Aug 11:00 UTC, Sep-Nov 11:00 UTC, Dec 11:00 UTC, Jan (days 1-25)	✅ ok · last run 1d ago
cfbfastR-cfb-raw	scrape_cfb_raw.yml	Scrape CFB Raw Data	09:00 UTC, Aug 09:00 UTC, Sep-Nov 09:00 UTC, Dec 09:00 UTC, Jan (days 1-25)	✅ ok · last run 1d ago

Orchestrator — registered in sdv-orch as pipeline cfb_models (CFB Models · cfbfastR-models), seasons 2014–2026.

Stage	Repo	Script
raw.scrape	cfbfastR-models	scripts/daily_cfb_scraper.sh

The Prefect schedule for this pipeline is currently paused; the GitHub Actions workflows above are the live driver.

What depends on it

• sportsdataverse-py — sportsdataverse.cfb.load_cfb_schedule() reads this tag directly (API docs).
• cfbfastR — cfbfastR::load_cfb_schedules() reads this tag directly.
• sdv-db — registered in the sdv-db catalog and ingested into the sportsdataverse Postgres warehouse as cfb.schedule, and served by the Data API at /v1/cfb/schedule.

Related releases

cfb_crosswalk, cfb_fpi_weekly, cfb_model_artifacts, cfb_ratings, cfb_ratings_weekly, cfb_recruiting_proj, cfb_recruits, cfb_returning_production, cfb_team_info, cfb_team_summaries_weekly, cfb_team_talent

---

Part of SportsDataverse. Every release in this repository is a public, versioned data asset: the files are stable URLs you can read straight from R, Python, or anything that speaks HTTP. Issues with the data belong on the producing repository listed above; issues with a loader belong on that package's repository.