The data backbone of Ruslan Magana Sports Intelligence — refreshed automatically every day. Large data never lives in GitHub — it lives here, partitioned by sport / date / layer. - Raw is append-only and immutable — full provenance back to the source API. - Bronze/Silver normalize to one canonical schema with stable IDs and de-duplication. - Gold is analytics/ML-ready: a feature store plus train/validation/test splits. - ML data is partitioned Parquet (by sport/date/layer) — never one giant CSV. Every feature is computed leakage-safely — for a given fixture, only matches with date < fixture.matchdate are used: The prediction label is the realized match outcome (home/draw/away or 2-way per…