Tennis point-by-point dataset, free for academic research
The dataset is a match index of 173,571 completed tennis matches across ATP, WTA, Challenger and ITF (1 January 2023 to 13 August 2026). 98.2% of them (170,527 matches) carry a point-by-point score sequence. Access is free for non-commercial academic research and teaching. Data is delivered as bulk CSV snapshots with a seven-day embargo after match completion.
Request access (research@livetennisapi.com) Public sample on Zenodo October 2026 sample on Hugging Face
What the data is
- Match index. 173,571 completed matches with tournament, tier, surface, round, best-of, scheduled time (UTC) and stable player IDs. 23,683 distinct players appear in the window.
- Live-observed point layer. 22,031 matches / 3,184,514 point states captured live by our own collection with real UTC timestamps. Live observation began 12 October 2025 and covers every tier down to ITF, where there has been almost no public point-level data.
- Reconstructed point layer. Earlier matches (all of 2023 to 2025) carry each match's recorded point sequence, expanded into score states after the match. They have no capture timestamps. This layer is not in the public files. Availability under the programme is decided per request.
- Player metadata. 32,678 players: name, country, date of birth, and handedness.
The tennis dataset as CSV files
The public sample and the research snapshots use the same three CSV files, gzip compressed, with a header row. Column definitions ship with the files.
| File | One row is | Columns |
|---|---|---|
matches.csv | a completed match | match_id, player1_id, player2_id,
tournament, tournament_key, tier_key,
surface, indoor, best_of, round,
scheduled_time_utc, event_status, is_qualifying |
players.csv | a player or doubles team | player_id, name, tour, country,
birthday, hand |
points_sample_<month>.csv | a score state we observed, in capture order | match_id, sets_p1, sets_p2, games_p1,
games_p2, points_p1, points_p2, server,
is_tiebreak, timestamp_utc |
Rows are score states, not strokes. A tape can start after the first point and can end before the last one. Check the first and last rows of a match before you model it.
Coverage, honestly
Numbers measured 2026-08-22 on the production database. Coverage is stated per provenance class because the difference matters for modelling.
| Tier | Completed matches | With point sequence | Live-observed |
|---|---|---|---|
| ITF women (singles) | 37,379 | 37,132 | 4,843 |
| Challenger men (singles) | 36,851 | 36,710 | 4,305 |
| ITF men (singles) | 35,642 | 35,377 | 4,896 |
| ATP (singles) | 16,476 | 16,393 | 1,785 |
| WTA (singles) | 15,950 | 15,860 | 1,643 |
| Challenger men (doubles) | 9,999 | 9,880 | 846 |
| Challenger women (singles) | 5,659 | 5,644 | 640 |
| ATP (doubles) | 5,116 | 5,071 | 326 |
| WTA (doubles) | 4,545 | 4,498 | 291 |
Known limitations
- The live-observed layer starts 12 October 2025 (roughly 10 to 13% of matches per tier so far). Earlier point sequences exist only as post-match reconstructions without capture timestamps.
- An observed tape can begin mid-match (collection attached late) or end before the final point (the recorded scoreboard stops early). Every match declares its own coverage. Do not assume a tape runs 0-0 to handshake.
- Serve speed, serve direction and shot-level data are not in the dataset.
- This dataset's point layer starts January 2023. A separate reconstructed point-by-point corpus for 2013-2022 exists as a commercial product (97,901 ATP, WTA and Challenger matches, no capture timestamps on any row). It is not part of this academic release. If your research needs those years, say so when you write and we will tell you what is possible. Nothing point-level exists before 2013, here or there.
Four published point-level findings, re-run on this dataset with the code and a CC BY results file: tennis findings, re-tested.
The tennis research reading list gathers journal papers on these questions. Each entry says whether the public sample can re-run it, and one BibTeX file holds them all.
Teaching a data science or statistics course? The capstone brief builds a project on this public sample, with four milestones, a rubric and starter code.
Results before 2023
This dataset starts in 2023. For earlier seasons, our results archive covers 1968 to 2022.
Getting access
- Email research@livetennisapi.com from an institutional address (students: include ORCID or a named supervisor).
- One paragraph: what you're researching and what data you need.
- Approval target: 3 business days. Annual renewal with a one-line progress note.
Conditions: non-commercial research and teaching only · no redistribution of raw data (published aggregates and findings are fine) · citation required · tell us when you publish. If you need a field or a cut we don't publish, ask. We decide case by case, and we'll tell you honestly whether the corpus holds it.
How to cite
Live Tennis API (2026). Tennis Point-by-Point Dataset — ATP, WTA, Challenger, ITF. Zenodo. 10.5281/zenodo.22048730
@dataset{livetennisapi_2026_tennis_pbp,
author = {{Live Tennis API}},
title = {Tennis Point-by-Point Dataset --- ATP, WTA, Challenger, ITF},
year = 2026,
publisher = {Zenodo},
doi = {10.5281/zenodo.22048730},
url = {https://doi.org/10.5281/zenodo.22048730}
}
Every correction to the public sample is listed with its version in our corrections log.
Questions researchers ask
Where can I get tennis point-by-point data for research?
Live Tennis API provides point-by-point tennis data free for non-commercial academic research. The match index holds 173,571 completed matches (1 January 2023 to 13 August 2026) across ATP, WTA, Challenger and ITF. 98.2% of them carry a point-level score sequence. A public sample is on Zenodo (DOI 10.5281/zenodo.22048730). The full dataset is granted on application to research@livetennisapi.com.
Is the data free for academic research?
Yes. Non-commercial research and teaching use is free. Conditions: no redistribution of the raw data (published aggregates and findings are fine), citation required, and tell us when you publish. Commercial use converts to commercial terms. The dataset covers 173,571 completed matches (1 January 2023 to 13 August 2026) across ATP, WTA, Challenger and ITF. 170,527 of them (98.2%) carry point-level score states, each with a coverage flag. Apply on this page with your institution and project, and access is granted by email.
Does the dataset include ITF and Challenger matches?
Yes, that is its main contribution. The index holds 37,379 ITF women's and 35,642 ITF men's singles matches plus 36,851 Challenger men's singles matches. The live-observed point layer covers the lower tiers, where there has been almost no public point-level data.
What is the difference between observed and reconstructed points?
Observed rows were captured live by our own collection with real UTC timestamps: 22,031 matches and 3,184,514 point states. Live observation began 12 October 2025. Reconstructed rows carry the recorded point sequence of a finished match, expanded into score states after it ended, with no capture timestamps. Public files contain observed rows only. Reconstructed data is decided per request.
How do I cite the dataset?
Cite the Zenodo record: Live Tennis API (2026). Tennis Point-by-Point Dataset — ATP, WTA, Challenger, ITF. Zenodo. DOI 10.5281/zenodo.22048730. BibTeX is on this page and on the Zenodo record. This DOI always resolves to the newest version. To name the exact snapshot you used, cite that version's own DOI from the record. Please also send us the published paper or preprint so we can list it.
Does it include serve speed or shot-level data?
No. Serve speed, serve direction and shot direction are not in the dataset. What it holds is the point-by-point score state: who served, the score before and after each point, break points, tiebreaks. A coverage flag per match says whether the sequence was observed live or reconstructed. Shot-level charting data is a separate part of the API, on its Ultra plan.