MatchIQ
Your personal sports prediction, bet builder and friends leagues app, with honest evidence labels, a sealed prediction history and a private Bet Diary.
How it fits together
pipeline/ (Python) fetches the schedule and results, publishes predictions, seals them in data/prediction_history.json, settles results and writes performance summaries. Run python -m pipeline.run (add --offline or --out <dir> --now <iso> for isolated test runs).index.html, styles.css, app.js and market_calculations.js make up the static app. It only renders what the pipeline published; it never computes predictions..github/workflows/matchiq-refresh.yml runs the pipeline every 6 hours once it is on the default branch. It can also be started by hand with "Run workflow". The frozen-model workflows (winner-forward.yml, forward.yml) are unchanged.python build_public_dist.py builds the allowlisted public-dist/ bundle that is deployed.
Data sources
| Source | Use | Terms |
|---|
| football-data.org (free tier) | Schedule, results, postponements | Optional. Needs a free token in the repository secret FOOTBALL_DATA_ORG_TOKEN; limit 10 requests/minute |
| OpenLigaDB | Schedule and results | ODbL 1.0, attribution required |
| openfootball/football.json | Schedule, results, Championship history | CC0 |
football-data.co.uk (data/matches.json) | Corners, cards, shots statistics; results | Personal/research use only; refreshed manually by update.yml, not automated |
When sources disagree on a score, a majority of two or more wins and the dissent is recorded. A tie stays RESULT_CONFLICT and is not settled.
Tests
python -m pip install -r pipeline/requirements.txt
python -m unittest discover -s tests -p "test_*.py"
for t in tests/*.js; do node "$t"; done
Research and evidence: experiments/m063_prediction_challenge/REPORT.md and experiments/m064_promoted_hierarchy/REPORT.md.
