Multi-Pass Source Collection

For each race, the engine executes 10–20 targeted web searches across five passes: Tier 1–3 press outlets (Autosport, Motorsport.com, The Race, BBC Sport, Sky Sports), international press (Auto Motor und Sport, Gazzetta dello Sport, L'Equipe, Marca), Reddit megathreads and week-after discussion, YouTube/social reaction metrics, and contextual championship data. Each source is tagged with credibility tier, geographic market, temporal distance from race, and structured sentiment scores.

Dual Scoring Engine

The engine scores each race across both a 4-axis model (Drama Index, Overtake Rate, Press Score, Fan Sentiment) and an 8-sub-axis model (incident density, championship stakes, last-lap action, pass quality, pass volume, press reception, press tone, fan engagement). The 4-axis view is the default display; the 8-sub-axis model stores granular signals that can be independently reweighted.

Verdict Derivation

The weighted composite score determines the verdict threshold. FULL RACE ≥ 74, HIGHLIGHTS 52–73, RESULT ONLY < 52. The engine is authoritative — the AI's self-reported verdict is always recomputed from the composite score to prevent gaming.

Adaptive Learning Loop

Community votes create a calibration signal with three thresholds: 5+ votes to display community consensus, 10+ votes to begin influencing the AI score (blended at up to 30% weight), and a 15-point drift threshold that triggers automatic score revision. Every revision is logged with timestamp, trigger type, and composite delta — creating an audit trail showing how verdicts evolve over time.

Temporal Re-scoring

For new races, sources are weighted by temporal proximity — early hot takes carry less weight than settled press reviews. At 30, 90, and 365-day milestones, the scheduler fetches new retrospective sources (podcast deep-dives, anniversary coverage, community reflection threads) and updates the score without re-fetching already-stored categories.