Racing decisions, backed by the numbers behind them.
Racecards and runner odds pulled into D1, de-vigged with an n-way model, and scored by edge so value is explicit. Built around CLV-first validation from day one.
Owner-only during build phase · not public.
Every runner shows its work.
No black-box confidence score — the board shows market odds, de-vigged fair probability, and edge per runner.
Racing data spine first
Racecards, runners, odds and results are persisted in D1 before any model features are trusted.
N-way de-vig model
Races are field markets, not 1X2. The fair-probability engine handles full runner books and overround explicitly.
CLV-first validation
Closing price capture is treated as the primary signal so model quality is measurable quickly, not after months of variance.
Rule 4 and non-runners
Settlement paths are designed around racing-specific events from the start so deductions and withdrawals are modelled, not patched later.
Operational controls
Admin can run sync on demand, inspect counts, and verify the data spine before feature expansion.
Built on proven estate patterns
Auth, feedback, motion and deployment patterns are inherited from Hilton Industries to keep the build focused on racing differences.
How it works
Sync
Pull racecards, runners and odds into D1 on schedule or on demand from admin.
Price
Convert market odds into fair probabilities and rank edge runner-by-runner.
Validate
Capture closes and settlement outcomes so the model can be judged on real evidence.