premier-league-data — English Premier League Results & Betting Odds Dataset (Python)

Open, pip-installable English Premier League results — with every bookmaker's odds kept intact. Every season from 1993-94 to the present (12,700+ matches), cleaned into two tidy CSV/Parquet tables and bundled so they load in one line of Python — no download or scraping required.

Most public "PL results" datasets throw away the betting odds. This one keeps them all — including opening and closing prices — which is what makes closing-line and market-efficiency analysis possible.

Data source: football-data.co.uk. This project only fetches, cleans, and reshapes their freely published files. See Data source & licensing.

Contents

• Install
• Quickstart
• The data
• Regenerate from source
• Data source & licensing

Install

pip install git+https://github.com/AnishKhetani/premier-league-data

Quickstart

import premier_league_data as pl

results = pl.load_results()                       # 12,700+ matches, all seasons
table   = results[results.season == "2023-24"]    # slice a single season
odds    = pl.load_results_with_odds("2023-24")     # same rows + every bookmaker's odds
print(pl.available_seasons()[:3])                  # ['1993-94', '1994-95', '1995-96']
print(odds[["home_team", "away_team", "bet365_1x2_home_close"]].head())

See examples/quickstart.ipynb for rebuilding a season's final league table, head-to-head records, and the chart above.

The data

Two tables, each shipped as CSV and Parquet in data/processed/ and bundled inside the package (so load_* works with no download).

results — core match results

Loaded with load_results(). One row per match, standardized across every era.

column	type	description
match_id	str	stable unique id, {code}-{home_slug}-{away_slug}
season	str	e.g. 2023-24
season_code	str	football-data code, e.g. 2324
date	datetime	match date, resolved to full year
home_team	str	home side
away_team	str	away side
fthg	Int64	full-time home goals
ftag	Int64	full-time away goals
ftr	str	full-time result: H / D / A
hthg	Int64	half-time home goals (NA before 1995-96)
htag	Int64	half-time away goals (NA before 1995-96)
htr	str	half-time result (NA before 1995-96)
referee	str	referee (NA before ~2000-01)

results_with_odds — results + all bookmaker odds

Loaded with load_results_with_odds(). The match-identity keys (match_id, season, date, home_team, away_team) plus 183 odds columns.

column	type	description
match_id	str	join key back to results
season, date, home_team, away_team	–	match identity
{bookmaker}_1x2_home	float	home win odds (decimal)
{bookmaker}_1x2_draw	float	draw odds
{bookmaker}_1x2_away	float	away win odds
{bookmaker}_1x2_home_close	float	closing home odds (_close = closing line)
{bookmaker}_over25	float	over 2.5 goals
{bookmaker}_under25	float	under 2.5 goals
{bookmaker}_ah_home / _ah_away	float	Asian-handicap prices
ah_line	float	market Asian-handicap line

Odds are sparse — a column is NaN for seasons where that bookmaker/market wasn't quoted. Odds columns follow football-data's {book}[C]{market} grammar, renamed {bookmaker}_{market}[_close]. Bookmakers & aggregators covered:

bet365 · pinnacle · william­hill · bwin · interwetten · ladbrokes · vcbet · betvictor · coral · betmgm · betfair (exchange & sportsbook) · gamebookers · stan james · sportingbet · blue square · 1xbet · market_max / market_avg (and legacy betbrain_max / betbrain_avg)

The three Betbrain price-count columns (Bb1X2, BbOU, BbAH) keep their original names. Full details in SPEC.md.

Regenerate from source

The data is reproducible end-to-end — nothing here is hand-edited.

pip install -e ".[dev]"           # dev install with fetch/plot deps
python scripts/fetch_data.py      # download all seasons -> data/raw/ (gitignored)
python scripts/build_dataset.py   # clean + combine -> data/processed/ (CSV + Parquet)

The current season refreshes automatically: a GitHub Actions workflow re-runs the pipeline on Monday mornings during the season and opens a PR when the data changes.

Data source & licensing

All data comes from football-data.co.uk, which publishes free historical football results and odds. This project does not create the underlying data — it only fetches, cleans, and reshapes it. If you use this data, please credit football-data.co.uk.

• Pipeline code: MIT licensed — see LICENSE.
• Underlying data: this project claims no license or ownership over it. football-data.co.uk states no explicit redistribution terms, so review their site before redistributing. See DATA_LICENSE.md.