Find out why some hosts find no stream for a game #66

Closed
opened 2026-10-31 11:33:55 -04:00 by cmoriarty · 2 comments
Owner

What happens

To check a link, the app opens the host's player page in headless Chromium and watches the page's network requests for a stream playlist (.m3u8). If none appears within the wait, the link is refused as "no stream found" (resolver/sportsurge.py, extract_stream).

On Chelsea vs. Man Utd (2026-10-31), givemestreams.cfd, buff-streams.buzz and resports.cfd answered "no stream found" on every check from 12:04 to full time. Yet these hosts play other games:

  • givemestreams.cfd and resports.cfd played Lakers vs. Warriors on 2026-10-30;
  • buff-streams.buzz played UCF vs. Baylor.

Across all checks from 2026-10-26 to 2026-10-31, 137 of 595 links tried (23%) yielded a stream.

The question

Either the host doesn't carry the game (the hubs list placeholders, or links for another game), or our reader misses a stream that's there. A miss could be a player that only requests the playlist after a click, a further frame, a stream URL our match doesn't recognise, or a load longer than our wait.

The work

  1. Look on live games: open a sample of links that answered "no stream found", the way the app does, and see what each page holds.
  2. Fix the reader for any pattern it misses, with a test from the saved page.
  3. Record what was found in docs/stream-sources.md, dated.

If the hosts simply don't carry those games, nothing changes in the reader; #64's demotion stops them taking every read's first tries.

## What happens To check a link, the app opens the host's player page in headless Chromium and watches the page's network requests for a stream playlist (`.m3u8`). If none appears within the wait, the link is refused as "no stream found" (`resolver/sportsurge.py`, `extract_stream`). On Chelsea vs. Man Utd (2026-10-31), givemestreams.cfd, buff-streams.buzz and resports.cfd answered "no stream found" on every check from 12:04 to full time. Yet these hosts play other games: - givemestreams.cfd and resports.cfd played Lakers vs. Warriors on 2026-10-30; - buff-streams.buzz played UCF vs. Baylor. Across all checks from 2026-10-26 to 2026-10-31, 137 of 595 links tried (23%) yielded a stream. ## The question Either the host doesn't carry the game (the hubs list placeholders, or links for another game), or our reader misses a stream that's there. A miss could be a player that only requests the playlist after a click, a further frame, a stream URL our match doesn't recognise, or a load longer than our wait. ## The work 1. **Look on live games:** open a sample of links that answered "no stream found", the way the app does, and see what each page holds. 2. **Fix the reader** for any pattern it misses, with a test from the saved page. 3. **Record** what was found in `docs/stream-sources.md`, dated. If the hosts simply don't carry those games, nothing changes in the reader; #64's demotion stops them taking every read's first tries.
Author
Owner

The live look, 2026-10-31 (15 player pages on NC State vs. California, Man City vs. Brighton and Bournemouth vs. Leeds). Mostly reader misses, not hosts without the game.

1. The main miss: the embedcenter.st hls_co player

  • The shared player. givemestreams.cfd, buff-streams.buzz, resports.cfd, s2watch.cyou, iframely.lol, ihdstreams.top, powerstream /center/ and others all embed the same player network: embedcenter.st/embed/chN.php.
  • Two modes. Each channel page uses one of two players:
    • hls.php: a normal .m3u8, which we read.
    • hls_co.php: the playlist is JSON (edgestream3.pro/sh/access/<token>/cdn/hls/<id>.json, keys schema_version, segments, playlist). Its segments are 4–5 MB image/png files on p16-common-sign.tiktokcdn.com.
  • The disguise. Inside the PNG's pixel bytes are TIKTIKPX, a 4-byte length and gzip, which unpacks to clean MPEG-TS. One segment decoded to H.264 1280×720 at 29.97 fps with AAC, 10 s.
  • Why we miss it. No .m3u8 is ever requested. The .m3u8 URL written in the page script answers 403.
  • It explains Chelsea. One channel in hls_co mode knocks out every front end at once. On Chelsea, givemestreams, buff-streams, resports, s2watch and iframely all failed while topsurge, iohit, shd247 and powerstream /tv/ verified. NC State showed the same pattern today.
  • Probably also some "segment is a PNG" decoy refusals (see the stream-decoys notes).

Reading it means:

  1. taking the JSON as the stream, with Referer/Origin https://embedcenter.st/;
  2. building an m3u8 from it, and renewing its roughly 3 h token with &grant=1;
  3. unpacking each PNG to TS before the decoy check.

2. "Sandbox detected" players

The embedsports.me, sportspatrika.com and taifood-blog.asia players refuse to play. They probably see "HeadlessChrome" in navigator.userAgentData, which contradicts our User-Agent. This is not tested; it's the operator's call whether to change it.

3. Smaller reader issues, fixed in this batch

  • A reason cut short. "no stream found; the player page is behind a bot challenge…" is cut at the ;, so bot-challenged pages count as "no stream".
  • goto(wait_until="load"). A player that never finishes loading becomes "source page unreachable".
  • Page-replacing ads. An ad that navigates the whole page away can end the read.
  • &amp; in data-href is not unescaped.

Every stream that did appear did so within 4 s. Nothing was missed for want of a click or a longer wait.

**The live look, 2026-10-31 (15 player pages on NC State vs. California, Man City vs. Brighton and Bournemouth vs. Leeds).** Mostly reader misses, not hosts without the game. ## 1. The main miss: the embedcenter.st `hls_co` player - **The shared player.** givemestreams.cfd, buff-streams.buzz, resports.cfd, s2watch.cyou, iframely.lol, ihdstreams.top, powerstream `/center/` and others all embed the same player network: `embedcenter.st/embed/chN.php`. - **Two modes.** Each channel page uses one of two players: - `hls.php`: a normal `.m3u8`, which we read. - `hls_co.php`: the playlist is JSON (`edgestream3.pro/sh/access/<token>/cdn/hls/<id>.json`, keys `schema_version`, `segments`, `playlist`). Its segments are 4–5 MB `image/png` files on `p16-common-sign.tiktokcdn.com`. - **The disguise.** Inside the PNG's pixel bytes are `TIKTIKPX`, a 4-byte length and gzip, which unpacks to clean MPEG-TS. One segment decoded to H.264 1280×720 at 29.97 fps with AAC, 10 s. - **Why we miss it.** No `.m3u8` is ever requested. The `.m3u8` URL written in the page script answers 403. - **It explains Chelsea.** One channel in `hls_co` mode knocks out every front end at once. On Chelsea, givemestreams, buff-streams, resports, s2watch and iframely all failed while topsurge, iohit, shd247 and powerstream `/tv/` verified. NC State showed the same pattern today. - **Probably also some "segment is a PNG" decoy refusals** (see the stream-decoys notes). Reading it means: 1. taking the JSON as the stream, with Referer/Origin `https://embedcenter.st/`; 2. building an m3u8 from it, and renewing its roughly 3 h token with `&grant=1`; 3. unpacking each PNG to TS before the decoy check. ## 2. "Sandbox detected" players The `embedsports.me`, `sportspatrika.com` and `taifood-blog.asia` players refuse to play. They probably see `"HeadlessChrome"` in `navigator.userAgentData`, which contradicts our User-Agent. This is not tested; it's the operator's call whether to change it. ## 3. Smaller reader issues, fixed in this batch - **A reason cut short.** "no stream found; the player page is behind a bot challenge…" is cut at the `;`, so bot-challenged pages count as "no stream". - **`goto(wait_until="load")`.** A player that never finishes loading becomes "source page unreachable". - **Page-replacing ads.** An ad that navigates the whole page away can end the read. - **`&amp;` in `data-href`** is not unescaped. Every stream that did appear did so within 4 s. Nothing was missed for want of a click or a longer wait.
Author
Owner

Closing: the look is done and its fixes are deployed (PR #67, merged as 1938e97, live 2026-10-31 17:44 UTC).

Shipped (reader-misses):

  • A bot-challenged page is logged with its challenge clause and counted as a challenge on the card, not as "no stream".
  • The read polls the page's requests from the navigation's commit, so a page whose load never ends is read, not reported unreachable.
  • Match-page links are unescaped.
  • The findings are in docs/stream-sources.md under 2026-10-31.

Carried forward:

  • The disguised hls_co player is #68, the next batch ("More working links", milestone 33). Its format is pinned down: Pillow's RGB bytes, TIKTIKPX, a big-endian length, then gzip, which gives MPEG-TS. That takes 0.04 s a segment.
  • The "Sandbox detected" players are an open question for the operator, noted in milestone 33.
Closing: the look is done and its fixes are deployed (PR #67, merged as 1938e97, live 2026-10-31 17:44 UTC). **Shipped (`reader-misses`):** - A bot-challenged page is logged with its challenge clause and counted as a challenge on the card, not as "no stream". - The read polls the page's requests from the navigation's commit, so a page whose load never ends is read, not reported unreachable. - Match-page links are unescaped. - The findings are in `docs/stream-sources.md` under 2026-10-31. **Carried forward:** - **The disguised `hls_co` player** is #68, the next batch ("More working links", milestone 33). Its format is pinned down: Pillow's RGB bytes, `TIKTIKPX`, a big-endian length, then gzip, which gives MPEG-TS. That takes 0.04 s a segment. - **The "Sandbox detected" players** are an open question for the operator, noted in milestone 33.
Sign in to join this conversation.
No labels
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
cmoriarty/sports-breaker#66
No description provided.