HKJC Data 101: Raw scrapes vs. cleaned datasets
Why scraped data and properly cleaned datasets are fundamentally different — encoding, gaps, duplicates.
Coming soon
Resources
Resources
Free race-day sample CSV, full dataset reference, and a tailored proposal request for your team.
Race-day sample CSV
Full-day Race Results sample from Happy Valley on 2025-07-16. Load directly in Python or R to verify schema and field completeness.
- Complete race day (9 races, ~110 rows)
- 14 core columns: Race_Key, Horse_ID, Rank, Finish_Time, Draw, Rating, Trainer, Jockey, Win_Odds, etc.
- Sourced from official public results, cross-validated against multiple sources
Tailored proposal document
Send us your data scope, use case, and budget context — we prepare a one-page PDF proposal covering coverage, delivery, and engagement terms tailored to your team.
- 24-hour response with initial recommendations
- Full schema and pricing released after mutual NDA
- Attach your existing pipeline or templates to accelerate alignment
Datasets
Dataset reference
Four categories, 30+ products, delivered per season.
-
Pre-Race
Race info · Racecards · Race readings · Predicted pacing · Analyst comments · etc.
-
Post-Race
Results · Sectional position / time / pacing · Dividends · Pool sizes · Comments · Replay video
-
General
Horse info · Veterinary · Trackwork · News · Barrier trials (info / results / comments / video)
-
Odds
Win · Place · Quinella · Quinella Place · Trio · Double — both final and time-series
Blog
Technical articles (coming soon)
Point-in-Time Pitfalls: Why your backtest is inflated
How look-ahead bias sneaks into feature pipelines and how to design a proper PIT store.
Coming soon
From CSV to Feature Store: A syndicate's full data pipeline
Building a low-cost, reproducible HK racing feature store with Polars + DuckDB.
Coming soon
Ready to talk about your data needs?
Request schemas, samples, or a tailored proposal — we reply within 24 hours.