About
RPX brings the schedule, circuit, standings, recent results, historical context, and original analysis into one fast race-weekend surface.
What is live
RPX is a static site deployed on Cloudflare Pages. Current schedules, results, and standings are fetched in the browser from Jolpica. Python builders produce the committed historical dataset and circuit statistics. Six circuits are rendered from open geometry with Three.js.
There is no production streaming pipeline, lakehouse, prediction model, AI agent, or multi-cloud platform behind the site today. Earlier architecture documents in the repository explore those possibilities; they are planning material, not shipped functionality.
Shipped architecture
Race schedule, completed-session results, and championship standings are requested client-side. Error states remain visible when upstream data is unavailable.
One table covers every World Championship Grand Prix from 1950 through the current season, including winners, starting position, circuit metadata, and pole derivation.
The first published investigation measures how often pole position becomes a victory and explains the pre-2003 grid-position caveat.
Custom HTML/CSS/JavaScript keeps the companion light; Three.js provides the circuit explorer without requiring an application backend.
Data boundaries
Results and schedules come from the Jolpica F1 API. Circuit geometry comes from the open f1-circuits project. Before 2003, the historical table uses starting-grid position one as a pole proxy because complete qualifying classifications are unavailable; the dataset records that distinction.
The 2026 season is in progress. Counts and current standings reflect the most recent generated artifact or successful API response, not an immutable final season.
Roadmap
First, validate every committed data artifact and improve freshness, provenance, and upstream-failure handling. Next, expand the useful race companion and track coverage. Only then will RPX attempt one narrow predictive question with a naive baseline, time-aware evaluation, calibration, and published failure analysis.
Streaming, lakehouse, human-data, agent, safety, and multi-cloud documents remain research unless corresponding code and evidence are shipped.