README.md

metadata

license: cc0-1.0
task_categories:
  - tabular-classification
language:
  - en
tags:
  - sports
  - wrestling
  - wwe
  - aew
  - wcw
  - ecw
  - match-data
  - relational
pretty_name: Ringside Analytics — Pro Wrestling Match Archive
size_categories:
  - 100K<n<1M
configs:
  - config_name: matches
    data_files: matches.parquet
  - config_name: match_participants
    data_files: match_participants.parquet
  - config_name: wrestlers
    data_files: wrestlers.parquet
  - config_name: events
    data_files: events.parquet
  - config_name: promotions
    data_files: promotions.parquet
  - config_name: wrestler_aliases
    data_files: wrestler_aliases.parquet
  - config_name: titles
    data_files: titles.parquet
  - config_name: title_reigns
    data_files: title_reigns.parquet
  - config_name: alignment_turns
    data_files: alignment_turns.parquet

Ringside Analytics — Pro Wrestling Match Archive

A relational snapshot of professional wrestling history from 1980 to the present: 292K matches, 611K wrestler-match participations, 35K events, and 12.8K wrestlers across WWE, AEW, WCW, ECW, NXT, TNA, and others. Sourced from public Cagematch.net scrapes and the alexdiresta profightdb dump, normalized into a Postgres schema, and exported as parquet files that preserve the relational structure (one file per table, joinable by id).

This is the source-of-truth companion to the trained model at theodorerubin/ringside-wrestling-archive-match-winner. If you want to train your own model, reshape the features, or just explore 40+ years of booking patterns — start here.

Files

File	Rows	Description
matches.parquet	292,780	One row per match. Type, stipulation, duration, title match flag, Cagematch rating.
match_participants.parquet	611,515	One row per wrestler-per-match. result is the label for outcome prediction.
wrestlers.parquet	12,814	Ring name, real name, gender, debut date, status.
wrestler_aliases.parquet	13,230	Alternate ring names with active-period bounds.
events.parquet	35,064	Event name, date, venue, city, country, event type.
promotions.parquet	6	WWE, AEW, WCW, ECW, NXT, TNA with founding / defunct dates.
titles.parquet	121	Championship belts per promotion.
title_reigns.parquet	1,753	Reign start/end + number of defenses.
alignment_turns.parquet	631	Face / heel / tweener transitions per wrestler.
manifest.json	—	Export manifest: row counts, columns, UTC timestamp.

Schema (join keys)

promotions.id ─┬─< wrestlers.primary_promotion_id
               ├─< events.promotion_id
               ├─< titles.promotion_id
               └─< wrestler_aliases.promotion_id

wrestlers.id ──┬─< match_participants.wrestler_id
               ├─< wrestler_aliases.wrestler_id
               ├─< title_reigns.wrestler_id
               └─< alignment_turns.wrestler_id

events.id ─────┬─< matches.event_id
               └─< alignment_turns.event_id  (nullable)

matches.id ────── match_participants.match_id

titles.id ─────── title_reigns.title_id

Starter queries

import pandas as pd

matches = pd.read_parquet("matches.parquet")
participants = pd.read_parquet("match_participants.parquet")
wrestlers = pd.read_parquet("wrestlers.parquet")

# Every match The Rock has wrestled, with opponents
rock_id = wrestlers.query("ring_name == 'The Rock'")["id"].iloc[0]
rock_matches = participants[participants["wrestler_id"] == rock_id]

-- If you load these into DuckDB:
SELECT w.ring_name, COUNT(*) AS wins
FROM match_participants mp
JOIN wrestlers w ON w.id = mp.wrestler_id
WHERE mp.result = 'win'
GROUP BY 1
ORDER BY 2 DESC
LIMIT 20;

Provenance

• Cagematch.net (public HTML scrape, non-commercial use): the bulk of match-level data for 1990-present.
• alexdiresta/all-wwe-and-wwf-matches Kaggle dataset (profightdb dump): cross-validation + pre-1990 coverage.
• Normalization + dedup: entity resolution on wrestler names, match-type classification into a fixed ENUM, and natural-key deduplication to collapse records across sources.

The ETL code and scraper are open source at tedrubin80/wrastlingfirst.

Caveats

• Kayfabe, not athletics. Pro wrestling is scripted. A result field records who was booked to win, not who would win an athletic contest.
• Temporal coverage is uneven. 2000-present is well-covered; 1980s are thinner, especially for regional/territory promotions.
• Gender imbalance. Women's division sample size is smaller — expect wider confidence intervals for any women's-division model.
• Ratings are crowd-sourced (Cagematch user ratings). They're a proxy for match quality as perceived by Internet wrestling fans — biased toward work-rate and away from entertainment/story.

License

Released under CC0 1.0 (public domain dedication). Attribution is appreciated but not required. Note that the underlying sources (Cagematch.net, profightdb) have their own terms; this archive is a derivative work made available for research and entertainment.

Citation

@dataset{ringside_analytics_2026,
  author = {Rubin, Theodore},
  title  = {Ringside Analytics: Pro Wrestling Match Archive (1980--present)},
  year   = {2026},
  url    = {https://www.kaggle.com/datasets/theodorerubin/ringside-wrestling-archive}
}