gptilt/lol-esports-matches

GPTilt: League of Legends Esports Matches

This dataset is part of the GPTilt open-source initiative, aimed at democratizing access to high-quality LoL data for research and analysis, fostering public exploration, and advancing the community's understanding of League of Legends through data science and AI. It provides a clean, canonical record of the competitive matches and games of professional League of Legends.

By using this dataset, users accept full responsibility for any consequences arising from its use. GPTilt assumes no liability for any damages that may result. Users are strongly encouraged to review the ["Uses"](#uses) section—particularly the ["Out-of-Scope Use"](#out-of-scope-use) subsection—for guidance.

Getting Started

First, install Hugging Face's datasets package:

pip install datasets

Now, you can load the dataset!

from datasets import load_dataset

# Specify just the config_name / table
dataset = load_dataset("gptilt/lol-esports-matches", name="games")

# Or include the split!
dataset = load_dataset("gptilt/lol-esports-matches", name="games", split="train")

Dataset Summary

A periodically-refreshed record of professional League of Legends matches and games — patches, champion picks/bans, teams, results, and tournament metadata. Built from Leaguepedia (lol.fandom.com) via its structured Cargo query API. It provides a clean, canonical record of competitive games for analysis and for linking broadcast/VOD commentary back to the specific game it discusses.

Dataset Structure

The data is structured into 3 tables, each exposed as a separate config:

• —`games`: One row per competitive game, keyed by game_id (Leaguepedia GameId). Includes the patch, the blue/red teams (team_blue = Team1, team_red = Team2), winner, the champion picks and bans per side (lists of champion names, role-ordered top→jungle→mid→bot→support for picks), game_number within its series, the parent match_id and tournament, datetime_utc, gamelength, the Riot platform game id (riot_game_id), and a vod link.

• —`tournaments`: One row per tournament, keyed by tournament_id (the canonical Leaguepedia page key). Includes name, league, region, split, year, tier, and date_start/date_end.

• —`matches`: One row per match — a series (e.g. Bo3/Bo5) — keyed by match_id. Includes the two teams, the series score (team1_score/team2_score), winner (1 | 2), best_of, tab, the parent tournament, and datetime_utc.

Join keys: games.match_id → matches.match_id; games.tournament and matches.tournament → tournaments.tournament_id. Team and champion names resolve to canonical entities via the companion lol-esports-entities entity_aliases index.

Each table is a single, unpartitioned snapshot (one train split per config).

Dataset Creation

Curation Rationale

The structured record of a professional game — which patch it was played on, the champions picked and banned, the two teams, and the result — is scattered across match histories and broadcasts. This dataset was created to provide a stable, canonical record of competitive games, matches (series), and tournaments, so that competitive results can be analyzed at scale and so that broadcast/VOD commentary can be linked back to the specific game it discusses (the patch and champion compositions being the anchor).

Source Data

Data Collection and Processing

The source data originates from **Leaguepedia** (the League of Legends esports wiki), accessed through its structured **Cargo** query API.

1. 1.Extraction: The ScoreboardGames, MatchSchedule, and Tournaments Cargo tables are queried in full on a weekly schedule.
2. 2.Raw Storage: Raw Cargo responses (JSON) are snapshotted to object storage, partitioned by week.
3. 3.Cleaning: Each Cargo table is deduplicated on its canonical key (GameId, MatchId, OverviewPage) and projected to a narrow, stable schema: ScoreboardGames → games (Team1 = blue side, Team2 = red side), MatchSchedule → matches (series), and Tournaments → tournaments. Champion picks/bans are split into lists of champion names (role-ordered for picks), and a source_url back to the originating wiki page is reconstructed. The clean layer is overwrite-only — each run reflects the latest snapshot of Leaguepedia, not historical state.
4. 4.Output: Each clean table is written to Parquet and published here as a separate config.

Who are the source data producers?

The underlying data is authored by the Leaguepedia community of volunteer editors, documenting competitive League of Legends matches and tournaments. The dataset curators are the contributors to the GPTilt project, who perform the automated collection, cleaning, and projection steps described above.

Bias, Risks, and Limitations

• —Coverage Bias: Leaguepedia's depth varies by region and competitive tier. Major regions and tier-1 competition are documented far more completely than minor regions and lower tiers.
• —Snapshot-Only: The clean layer is overwrite-only — it reflects the latest state of Leaguepedia at collection time, with no history retained.
• —Editorial Lag & Gaps: As a community wiki, Leaguepedia may lag behind, and older games in particular may have incomplete picks/bans or missing identifiers.

Recommendations

• —Treat the data as a best-effort snapshot of a community wiki, not an authoritative or complete record.
• —Expect missing fields for older games; guard for null picks/bans and identifiers.
• —Acknowledge the coverage bias toward major regions/tiers when reporting aggregate results.

Uses

Disclaimer

This dataset is derived from [Leaguepedia](https://lol.fandom.com/), a Fandom-hosted community wiki. GPTilt is not endorsed by Fandom, Leaguepedia, or Riot Games, and does not reflect their views or opinions. League of Legends and Riot Games are trademarks or registered trademarks of Riot Games, Inc. League of Legends © Riot Games, Inc.

License

This dataset is licensed under `cc-by-sa-3.0`. Because it is derived from Leaguepedia content, redistribution and derivative works must provide attribution to Leaguepedia and remain under compatible share-alike terms. Every row carries a source_url linking back to the originating wiki page to support attribution.

Direct Use

This dataset is intended for non-commercial research, data analysis, and tooling aimed at understanding and organizing competitive League of Legends data. Suitable uses include:

• —Competitive analysis — draft/meta trends, win rates, objective control, and results across patches and regions.
• —Grounding — linking VOD/broadcast commentary to the specific game it discusses, via patch + champion composition.
• —Educational purposes related to data science and esports analytics.

Out-of-Scope Use

This dataset is a best-effort community snapshot and must not be relied upon as an authoritative record, nor used in any way that violates Leaguepedia's licensing terms.

Changelist

Jun 2026

• —Initial release: games, tournaments, and matches, sourced from Leaguepedia (ScoreboardGames, MatchSchedule, Tournaments) via its Cargo query API.

Citation

If you wish to use this dataset in your work, we kindly ask that you cite it.

For most informal work, a simple mention of the GPTilt project and the League of Legends Esports Matches dataset will suffice. Please also attribute Leaguepedia as the upstream source.

BibTeX:

@misc{gptilt_league_of_legends_esports_matches,
  author    = { GPTilt Contributors },
  title     = { League of Legends Esports Matches },
  year      = { 2026 },
  publisher = { Hugging Face },
  journal   = { Hugging Face Hub },
  url       = { https://huggingface.co/datasets/gptilt/lol-esports-matches }
}