Skip to main content
Bronze is a faithful copy of what FastF1 and Jolpica return including their rough edges. This page catalogues the gaps and quirks worth knowing before you query Bronze directly, so you can tell “the data really is like this” from “something broke.”

​
Coverage

Generated directly from ingestion/verify_bronze.py --markdown against the files actually on disk never hand-typed, so it can’t drift from reality. Regenerate it yourself with make docs-coverage.

​
Known issues

A FastF1 v3.8.3 change to how it represents the Time column means session_time_s doesn’t populate for any 2024 race in telemetry or race-control output. The fields it’s normally used to join are still derivable: use lap_number for telemetry joins, and Lap for race-control joins. Laps and weather data for 2024 are unaffected.
FastF1 raises an internal timing-accuracy flag for several drivers at the 2024 Las Vegas Grand Prix. Impact is low the laps themselves are present and usable and the root cause is still under investigation upstream.
FastF1 raises when asked to resolve a pre-season testing event by round number, since testing isn’t numbered like a points-paying round. Ingestion already excludes round zero from its season loop, so this surfaces only as a benign log line, never a failed write.
Red-flagged and short qualifying sessions are common, and expected the data-quality engine warns but never blocks a qualifying write on row count alone.

​
Phantom columns you might expect but don’t exist

Earlier, hand-maintained docs for this pipeline named several columns that were never actually present in the Parquet output. They’re listed here so a search for one of these names lands on its real equivalent instead of a dead end.
Weather there is no cumulative rainfall_mm; rain is a boolean flag (rainfall_flag) only. wind_direction_deg and wind_speed_kmh don’t exist either the real columns are wind_direction (degrees, as a plain integer) and wind_speed_ms (metres per second, not km/h).Telemetry there is no brake_pct; FastF1 exposes no brake-pressure channel, only a boolean brake. drs as a boolean doesn’t exist the real column is DRS, a numeric enum where only 10, 12, and 14 mean the flap is open. gear doesn’t exist either the real column is nGear.Race Control there is no code column; it was never ingested. The lowercase lap, time, and flag don’t exist the real columns are Title-Case: Lap, Time, Flag.
TrackStatus on the laps dataset is ingested as an opaque VARCHAR digit string, passed through exactly as FastF1 returns it. Ingestion applies no decoding interpreting the digits into safety-car, VSC, or flag states is a transform-layer concern, not a Bronze one.