F1 Agentic Race Analysis · deterministic layer + Bedrock agent briefs

2024 Spanish Grand Prix

Circuit de Barcelona-Catalunya · 66 laps · Ergast-schema lap data, fits reproducible from results.json
2026 CHAMPIONSHIP PREDICTORknow more ->

Pit-wall trace

slow loopgap to leader, top six finishers; markers are pit stops
vertical drops across pit cycles are the undercut swings measured in the ledger below

The stop that priced the win

post sessioncounterfactual pit-lap model: fixed pit loss, linear per-stint deg, no traffic interaction
modeled race time Norris left on the table by stopping on L23 instead of L17
actual finishing gap to Verstappen. Under the model's assumptions, the win was live at the first stop.

Undercut ledger

slow loopmeasured gap swings across pit cycles, rival pairs within 4s at the in-lap

Rival Watcher payload: undercut_events + pit_loss

Stint Analyst payload: stint_fits + fuel_slope

Degradation anomalies

slow looprobust z vs field slope distribution; trigger threshold |z| ≥ 1.5

Deg Explainer trigger: BOT stint 2, z=4.79

Driver metrics

post sessionfuel-corrected pace vs in-stint consistency, this race

Driver Coach: NOR payload: driver_metrics + counterfactuals

Track archetypes

cross season25 circuits, 2022-2024: tyre stress vs race neutralisation

Track Historian payload: track_profiles, focus=Barcelona

Compliance gate

slow loopdeterministic rules engine; the agent explains, it never adjudicates

Compliance Guardian payload: audit record #2

Prediction audit

slow loopfrozen-clock backtest: fits from laps 1..N, scored on laps N+1 onward; no lookahead

Race Reporter

post sessionfull-payload strategy narrative

Strategy story of the race payload: full deterministic output set