new-market
Add a new betting market to the analysis engine end-to-end (recommendation generation, odds extraction, settlement, i18n, docs). Use when the user wants to support a new bet type / market (e.g. corner
SKILL.md.claude/skills/new-market/SKILL.md
GitHubAdd a Betting Market
Wire a new market through the full pipeline so it is generated, priced, settled, calibrated, and translated.
Checklist (do all that apply)
- Generate it — add a row to
raw_recommendationsinrecommendations.py:generate_recommendations:(market_type, selection, probability, category, odds). Deriveprobabilityfrom existing stats/Elo signals; don't invent a constant. - Price it — add an odds-extraction pattern in
recommendations.py:_market_odds()(and_extract_market_odds) so bookmaker odds and EV get populated. The matching is fuzzy regex — verify against real API market names frompython tippmix.py events. - Settle it — add result-evaluation logic in
evaluation.py:evaluate_run()so the market can be graded win/loss/push. Handle voids (push pays, doesn't zero combos). - Score impact (optional) — if it should influence match confidence, weight it in
recommendations.py:score_match_confidence()and add the weight key toconfig.py:DEFAULT_CONFIG["weights"](+tippmix.config.jsonif user-configurable). - Translate it — add
market_label/selection/result_labelstrings totippmix_system/i18n.py. Code stays English; only output strings are Hungarian. - Calibrate it — new markets start with adjustment 0 until ≥ 5 settle; that's expected.
- Test it — add a fixture-based test under
tests/(mirrortest_combo_settlement.py/test_calibration.py). Cover generation, odds extraction, and settlement. - Document it — update the market tables in
ARCHITECTURE.mdandCLAUDE.md.
Verify
python -m py_compile tippmix_system/recommendations.py tippmix_system/evaluation.py
python -m pytest tests/ -q
python tippmix.py analyze # confirm the market appears with sane odds/EV