theScore API: Free Endpoints from the Scores App Feed, Not theScore Bet

theScore API is the JSON feed at api.thescore.com that runs theScore sports media app: scores, schedules, box scores, play-by-play, standings and player pages. It is unofficial and undocumented. theScore publishes no developer docs for it, asks for no key and charges nothing. Developers who search for the score api usually mean this feed, which belongs to the scores app. theScore Bet, the sportsbook, is a separate product. Looking for odds APIs that carry theScore Bet lines? See theScore Bet odds API guide.

We called every endpoint on this page with plain curl on 2 November 2026 and ran the code samples as written in Python (requests) and Node (fetch). The feed behaves a lot like the ESPN API. Both are undocumented, both cover the same major leagues, and either one can change without notice. theScore's terms allow personal, non-commercial use, so read Limitations and Terms before you build on it.

One host serves the data. A second serves the images that responses point to.

Objects link to each other through api_uri. Event 2462797 carries "api_uri": "/nba/events/2462797", its box score carries /nba/box_scores/247125, and each team carries a path such as /nba/teams/18. Put the host in front of one and you have the next URL.

https://api.thescore.com/leagues returned 100 leagues, each with the slug that opens every path. These are the ones developers ask about most, with the season size from /{league}/schedule (tested November 2026).

Serie A is seri: /seria/events/current answered 404. The list also carries the KHL, IPL cricket, Formula 1, NASCAR, MMA and the World Cup. Tennis and golf use a different shape. Their events/current returns one tournament object with start_date and end_date in place of a list of games: /atp/events/current returned the Rolex Paris Masters.

Paths follow /{league}/{resource} or /{league}/{resource}/{id}. Every row below answered HTTP 200 on 2 November 2026, and the examples use the ids we tested.

Some paths that look right do not exist. /nba/events/2462799/play_by_play, /nba/teams/11/roster and /nba/odds/12853680 answered 404. League-wide /nba/injuries answered 200 with an empty list while /nba/teams/11/injuries listed five Atlanta Hawks players, and /nhl/injuries answered 404. Read injuries team by team.

https://api.thescore.com/{league}/events/{id}

An event is one game, and everything else hangs off it: the box score link, the play-by-play flag and a betting line. Our test game was event 2462797, the Toronto Raptors at the Los Angeles Lakers, a 131 to 125 Lakers win (43 KB).

Games further out can have no box score: all five of Atlanta's upcoming games had box_score set to null. The feed names no sportsbook for odd or timestamped_odds, so treat them as a reference line. For theScore Bet's own prices, see theScore Bet API.

Box score, player lines and play-by-play

/nba/box_scores/247125 returned line_scores by quarter (Lakers 36, 30, 38, 27), team_records with team totals and attendance "19,057". /nba/box_scores/247125/player_records returned 35 player lines, led by Luka Dončić with 45 points and 15 rebounds. /nba/events/2462797/play_by_play_records returned 521 plays, each with description, event_type, progress.clock_label and the score after it. Live games work too: Atlanta at Indiana (2462799) had 126 plays at 10:06 of the second quarter.

Other sports change the play fields. NFL plays add down and distance (166 plays, 1.3 MB, for event 140696). MLB gave one record per half-inning, 18 for World Series game 100780, and Premier League records carry a category such as Goal or Corner.

rpp matters most on events/current: without it, /nba/events/current returned 5 games while 9 were in progress, and ?rpp=20 returned all 9. Leave every filter off /nba/events and you get the whole season, 1,267 games in one 8.9 MB response.

For one US day, run game_date.in from 08:00 UTC to 08:00 UTC the next day. That window returned the four NBA games of Sunday 1 November. Midnight to midnight UTC returned six: four from Saturday night in the US and two from Sunday afternoon. A bad value such as game_date.in=yesterday,today returns 400.

The schedule as an index

/nfl/schedule returned current_season, 26 groups from "HOF" to "Super Bowl", and current_group, Week 8 with 14 event ids. Daily leagues group by date and tennis and golf by month. Pass a group's event_ids to id.in to get the games.

Group labels follow each game's UTC date. At 01:11 UTC on 3 November (Monday evening in the US), current_group was "Nov 2" and held two games from the night before, while that evening's games sat in "Nov 3". Trust game_date and status over the label.

Games on a date in Python

import requests

# NBA games of Sunday 1 November 2026, US time (08:00 UTC to 08:00 UTC)
url = ('https://api.thescore.com/nba/events'
       '?game_date.in=2026-11-01T08:00:00Z,2026-11-02T08:00:00Z')
for game in requests.get(url, timeout=10).json():
    score = (game.get('box_score') or {}).get('score', {})
    print(game['away_team']['full_name'], score.get('away', {}).get('score'), 'at',
          game['home_team']['full_name'], score.get('home', {}).get('score'),
          game['status'])

It printed four lines, the last one "Toronto Raptors 125 at Los Angeles Lakers 131 final". The box_score guard covers dates far ahead, where games have no box score yet.

Live games and one event with cURL

# Every NBA game theScore treats as current, up to 20
curl 'https://api.thescore.com/nba/events/current?rpp=20'

# One finished game, with its line and box score link
curl 'https://api.thescore.com/nba/events/2462797'

Play-by-play in JavaScript

// Run as an .mjs file in Node 18+
const url = 'https://api.thescore.com/nba/events/2462797/play_by_play_records';
const plays = await fetch(url).then(r => r.json());
const last = plays[plays.length - 1];
console.log(plays.length, 'plays');
console.log(last.progress.clock_label, last.description);
console.log(`Final: ${last.away_score_after} to ${last.home_score_after}`);

It printed 521 plays and the last one, a Jamison Battle dunk at 0:01 of the fourth. Responses carry access-control-allow-origin: *, the header a browser needs before it lets a page read a cross-origin fetch.

404 Not Found

A bad league slug, a bad id and an unknown path all return 404 with this body (from /seria/events/current):

{"api_uri":"/error/events/404","error":{"request":"https://api.thescore.com/seria/events/current","code":404,"reason":"Not Found"}}

The same shape came back for /xyz/events/current, /nba/events/999999999, /nba/events/abc, /nba/teams/99999 and /nba/players/0.

400 Bad Request

/nba/events?game_date.in=yesterday,today returned 400. The filter wants two ISO 8601 times such as 2026-11-01T08:00:00Z.

200 with an empty list

We saw three causes. No games in the window: /mlb/events/current was empty after the World Series, while /mlb/schedule still listed the season. A bad id.in value: /nba/events?id.in=abc returned [] in place of an error. And the league-wide injuries path, covered above.

No 403 and no 429

None of our 128 calls to api.thescore.com was blocked or throttled. Each one sent the client's default headers (curl, Python requests or Node fetch), with no key, cookie or browser header.

The licence covers a personal script that follows your own teams. A public app, a paid product or a bulk archive falls outside it, and section 7 also rules out heavy load and commercial use. The terms are at thescore.com/pages/terms, which redirects to theScore's help centre.

• No docs, no promise: theScore can change or close any path on this page without notice.
• robots.txt says no crawlers: api.thescore.com/robots.txt disallows every path for every user agent.
• No published rate limit: our calls ran at about one a second with no 429 and no X-RateLimit header. Cache-Control said max-age=10 on live games, 300 on a finished game and 7200 on past dates, so polling a live game faster than every 10 seconds gains nothing.
• Licensed text inside: the recap_data on our test game was a wire-service story, one more reason not to republish the feed.

A product built on this feed needs a licence from theScore. For production apps with an SLA, compare licensed sports data providers. For odds, browse the directory.

Browse API Providers →

We link no client library for this feed. A GitHub search on 2 November 2026 turned up one maintained Python wrapper, and it sends a copied Chrome User-Agent plus Origin and Referer headers on every call. You do not need them: every call in our tests answered 200 with the default headers of curl, Python requests and Node fetch. Section 7 of theScore's terms also bars forged headers that disguise where a request comes from, so a plain client is the one to write.

For the same leagues from a second undocumented source, see the ESPN API guide. For open historical data, see free sports datasets and free sports APIs.