SocostarsFootball intelligence

About Socostars

Football intelligence for people who want to know where the number came from.

What it is Methodology Recorded vs estimated Sources News as an input Public API Editorial & monetisation Corrections Contact

1. What Socostars is

Socostars is a football intelligence workspace. It answers four questions in one place, for every fixture it can see across the leagues it indexes:

  • What is happening? Fixtures, kickoff times in Indian Standard Time, live scores while a match is in play, and results once it has finished.
  • What does the model expect? Win, draw and loss probabilities, a most-likely scoreline, expected goals and a confidence figure.
  • Why does it expect that? The nine inputs the model actually reads, each showing both clubs' recorded values and which side the input favours.
  • What is recorded around it? Form, home and away splits, head-to-head history, registered squads, season statistics, and headlines from named publishers.

It is a statistics and information product. It is not a betting product, it publishes no tips, no odds and no staking advice, and no figure on this site is a guarantee of an outcome.

Every number is labelled

Each figure says whether a source recorded it or the model estimated it. Nothing is presented without saying which it is.

Missing stays missing

When a source publishes nothing for a value, the page says so. It is never filled in with an estimate dressed up as a record.

The gaps are shown

A fixture with thin coverage opens with a coverage strip naming exactly which sources loaded and which did not.

Sources are named

Every card and every row credits the provider it came from, and news always links to the publisher's own article.

2. How the model works

The model turns recorded results and recorded season statistics into a probability for each outcome. It reads these inputs, and only these inputs:

Strength

Recorded attacking and defensive output per game, and the club's recorded position and points per game in its competition.

Form

Results over the club's recent matches, weighted by how recent they are and by the strength of the opponent faced.

Venue

The home club's record at home and the away club's record away, rather than a single combined figure.

Context

Rest days between fixtures, the recorded head-to-head between the two clubs, and published club news where it reports a squad availability change.

What does not move any number. Confirmed lineups, in-play statistics, possession, referee, weather, crowd, transfer gossip, and anything a model could not have known before kickoff. These are stated on every match page next to the reasoning, so it is always clear what the forecast did and did not use.

What is published, and what is not. The site publishes a probability distribution (home, draw, away), a most-likely scoreline, expected goals for each side, and a confidence figure. It publishes an accuracy figure only after a walk-forward backtest has been run against finished matches; until then, the accuracy room states that no accuracy figure exists yet rather than publishing a flattering one.

Live forecasts. When a match is in play and a source has published the score, the model re-conditions its probabilities on the current score and the time remaining, and the page shows both the pre-kickoff forecast and the live one so the movement is visible.

3. Recorded data versus model estimates

FigureStatusWhere it comes from
Fixtures, kickoff times, results, standings, registered squads, scorer tableRecordedfootball-data.org, and API-Football for competitions it does not cover
Live score and live minuteRecordedThe provider's own in-play status, shown only while a provider says the match is in play
Match statistics and expected goalsRecordedfootball-data.co.uk's free results, statistics and expected-goals archive, imported at build time
Form, home/away splits, head-to-headRecorded, then summarisedComputed from the recorded results above; the page shows the matches it counted
Win / draw / loss probabilities, most-likely score, expected goals, confidenceEstimateSocostars' own model
Strength index and ratingsEstimate, on a stated scaleComputed against the pool of clubs that actually have recorded results loaded, with the formula printed beside it
Club news headlinesRecorded, published verbatimThe publisher's own feed, linked back to the article. Never rewritten by Socostars

A rating or probability with no underlying data reads "not computable". It never reads zero, because zero is a claim.

4. Sources and licences

SourceWhat Socostars reads from it
football-data.org (v4)Fixtures, kickoff times, results, standings, each club's registered squad, and the recorded scorer table. Requests are made by Socostars' own server under its own token and answered from cache.
API-Football (v3.football.api-sports.io)Additional coverage — registered squads, availability records and head-to-head — for competitions football-data.org does not carry. Also server-side only.
football-data.co.ukThe free results, match-statistics and expected-goals CSV archive. Imported at build time. No odds columns are imported or displayed.
Wikipedia / WikimediaPublished current-squad lists, used to fill registration gaps, under CC BY-SA. Requested through Socostars' own server route.
openfootball football.jsonOpen league and fixture data used as an additional check on coverage.
ESPN public scoreboardsLive scoreboards for competitions outside the primary provider's plan, read in the browser.
TheSportsDBAdditional club and competition metadata.
BBC Sport, Sky Sports, The Guardian, ESPN, Google News indexHeadlines, published descriptions, timestamps and links, read through Socostars' own news route. Articles are never copied or republished.

These providers retain all rights in their data. Socostars claims no ownership of it and presents it for personal, informational use, with attribution beside it. Club names, badges and competition marks belong to their respective owners and are used for identification only.

5. News as a model input

Club news is shown in the dashboard news flash, in the news centre and on each club's own page, always verbatim and always linking to the publisher. Where published reporting states a squad availability change before kickoff, the model may apply a small, bounded adjustment to that club's forecast — currently a ceiling of ±7% — and the match page discloses that an adjustment was applied.

That adjustment has not yet been measured against results. Until a backtest has run, the site says so on the match page rather than implying the adjustment improves anything.

6. Public API

The site is served by a small set of read-only JSON endpoints. They exist to make the pages work, and they are documented here because they are also useful on their own:

EndpointReturns
GET /api/liveLive and recent matches with the provider's in-play status, minute and published score. Optional codes parameter restricts the competition sweep.
GET /api/fdFixtures, results, standings and season records for the covered competitions.
GET /api/rostersThe registered-squad file generated at build time.
GET /api/newsHeadlines for the general feed or a specific club, each with its publisher and a link back to the article.
GET /api/squadsCurrent registered squads resolved per club.

Responses are cached at the edge. Access is rate-limited by the platform; there is no key to request and no SLA. If you build on these endpoints, please identify your client and cache your side rather than polling.

7. Editorial and monetisation policy

These are binding on this site and will stay on this page:

  • No betting content: no odds, no tips, no staking plans, no affiliate links to bookmakers, and no promotion of gambling to anyone under 18.
  • Statistical estimates are labelled as estimates wherever they appear. No page presents an estimate as a recorded fact.
  • No invented data. Where coverage is missing, the page names the gap instead of filling it.
  • Advertising, if enabled, is separated from editorial content and never presented as a recommendation. The privacy notice is updated before any advertising, analytics or newsletter tooling loads, and any consent requirement is satisfied first.
  • Headlines are never rewritten, reworded or summarised by Socostars. Each row shows the publisher's own words and links to the article.

8. Corrections

Football data goes wrong in predictable ways: a kickoff time moves, a player transfers, a fantasy feed disagrees with the league's own registration, a competition spelling differs between providers. Socostars handles this by naming its source beside every figure, refreshing continuously, and showing a disagreement rather than hiding it.

If you find something wrong, tell us the fixture, the club or the player, the value you see, and the value you believe is correct with a public source. Corrections are applied at the source of the error, not patched in the display layer, so the fix holds everywhere the figure appears.

9. Contact

Operator of Socostars:

  • General and data corrections: hello@socostars.com
  • Privacy and legal: hello@socostars.com
  • Site: socostars.com

Please include the page address and a screenshot where you can — it is the fastest way to get a data error fixed.

Reading this as the site operator? The contact addresses above and the operator identity are the two things that must be real before this site is submitted to an advertising network or a search console as a business. Everything else on this page describes behaviour that is already implemented.
← Back to Socostars Privacy notice Terms of use Data & estimates disclaimer