Methodology

1. Source observations remain separate

Every fetched or imported result is stored as an observation with its source, fetch time, parser version and a cryptographic fingerprint. The original payload may be retained for diagnosis and audit.

2. Canonical results are reconciled

Observations for the same game and draw date are compared. Two matching independent observations can mark a draw verified. Multiple conflicting fingerprints mark the draw as a conflict until reviewed.

3. Numbers are stored as text

A four-digit result such as 0042 is stored as 0042, not as the integer 42. This prevents leading-zero corruption.

4. Statistics are descriptive

Frequency, first appearance, last appearance and prize-tier counts describe the records in the database. They do not predict future results and are not presented as recommendations.

5. Corrections remain auditable

Administrative changes and imports are logged. A source observation is not silently overwritten merely because a newer source disagrees.