Multiregion Results Agent

Fifty-nine official publishers in three regions, no two alike: JSON APIs, server-rendered tables, JavaScript applications, PDF documents, bot-protected pages, and geo-restricted endpoints. There is no common feed, and no aggregator carries complete coverage. The hard part is not storing the data — it is acquiring it from all of them, reliably, politely, and sustainably.

A fetch pipeline built around compliance instead of evasion: robots.txt checks, per-host rate budgets with jitter, exponential backoff on errors, and every raw page saved before parsing so nothing is ever lost. A fleet of per-source parsers with fixture-tested coverage handles the formats; geo-aware egress through a controlled VPN with region-matched exits unlocks region-locked official pages; and a headless-browser tier covers the sites that need one.

The hard rule: sites with bot protection are never bypassed — no CAPTCHA solving, no stealth evasion. They wait for an assisted manual import instead. Roadblocks are met with a compliant acquisition policy, not tricks.

Everything lands in a local PostgreSQL archive: a searchable dashboard, per-jurisdiction pages, date-range reports with streaming Excel export, derived draw schedules, and next-draw reports — all served from one local source of truth. The web is queried once at acquisition time and never at read time.

The probability lab explores probability and statistics on real public data: probabilistic outlooks built from stored history, scored against simple baselines with walk-forward backtests, calibration checks, and fairness tests. On deep, fair games no strategy beats chance — and the tool says so plainly. That honesty is the point of the experiment.

There are no accounts and no user database: access works through unguessable secret links in three roles — administrator, full-access, and view-only demonstration. The gate fails closed, token checks are constant-time, failed entries are throttled per IP, and view-only limits are enforced on the server, not just in the UI.

The live system is available by private link. A view-only demonstration link — rotatable, server-enforced read-only — can be issued on request. If a link you were given no longer works, it has been rotated; ask for a current one.

This is an engineering prototype — a technical exercise, not a product and not a service. It carries no uptime commitment and offers no financial, statistical, or wagering guidance: nothing here places, prices, or facilitates a wager, and every probability surface reports a distribution rather than a prediction.

All content comes from official, publicly available pages and remains the property of the respective operators. This project is not affiliated with, endorsed by, or connected to any of them. Everything here is for informational, educational, and experimental purposes only: the analysis features describe the past — they do not and cannot predict future draws.