Skip to main content
tif1 provides data in a structured format consistent with fastf1. This page lists every column in the main DataFrames.
tif1 uses the fastf1 column names, which are mostly PascalCase, for example LapTime and Speed. A few source columns keep lower-case names: index, drv, team, ms1 to ms3, and the telemetry column nGear.
This page documents the DataFrame columns that tif1 provides. tif1 builds these DataFrames from raw JSON files in the TracingInsights data repos. The repos store the data under short abbreviated keys: s1, vi1, wWS. Every raw field is documented in the Data Reference.

⏱️ Lap Data (session.laps)

The laps DataFrame contains one row for every lap completed by every driver in the session. The dtype of a column depends on the backend. On pandas, time columns are timedelta64[ns]. On polars, the same columns are Duration("ns").
Compare pandas time columns with pd.Timedelta, or use LapTimeSeconds, which is a float on both backends. laps[laps["LapTime"] < 90] raises TypeError on pandas. On polars, pl.col("LapTime") < 90 compares nanoseconds and selects no lap.

Core Timing

Sectors & Speed Traps

Tires & Strategy

Status & Metadata

Per-Lap Weather

Every lap carries one weather sample that the source data matched to the lap. WeatherTime gives the session time of that sample. These columns are misaligned in the source data. In the 2025 Bahrain race, 435 of the 1,128 laps carry a sample other than the one that the fastf1 rule picks. To relate weather to laps, use laps.get_weather_data(). It picks the sample of each lap from session.weather with the fastf1 rule. The rule takes the first sample within the lap, else the last one before the lap ended.

🎯 Telemetry Data (lap.telemetry)

Samples recorded throughout a lap, about 0.13 s apart in the 2024 Monaco Race data. Time is lap-relative and starts at 0 at the lap start. The telemetry of several laps (laps.telemetry) counts Time from the start of the first lap, and session.car_data[n] holds the session time in Time, as fastf1 does. SessionTime and Date give the session time and the wall-clock date of each sample. The columns follow the fastf1 order, then the tif1 extras.

Car Metrics

Spatial Data

Driver Ahead

Lap Identity

Acceleration (G-Forces)

These are mathematically derived and smoothed from position and speed data. All three columns are float64 (pandas) or Float64 (polars) in m/s².
Longitudinal acceleration is smoothed with a 3-point moving average. Lateral and Vertical are smoothed with a 9-point moving average to reduce noise.

🌡️ Weather Data (session.weather)

Environmental conditions recorded once per minute. The table has no LapNumber column. laps.get_weather_data() returns the sample of each lap.

🚩 Race Control (session.race_control_messages)

Official messages issued during the session.

🏁 Session Results (session.results)

One row per driver, in classification order. On pandas, session.results is a SessionResults frame indexed by driver number. On polars, it is a PolarsSessionResults frame with DriverNumber as a column. Race, Sprint, and Qualifying sessions take the official Ergast classification: grid, points, status, and times.
Without an official classification (practice, the sprint qualifying formats, or ergast=False), tif1 classifies the session from the laps, as fastf1 does. Sprint Qualifying, Sprint Shootout, and Qualifying with ergast=False take the best accurate lap of each segment into Q1 to Q3 and rank by Q3, then Q2, then Q1. Other sessions keep the order that drivers.json lists when that is already a classification. Otherwise tif1 rebuilds it. Races and sprints rank by laps completed, then by position on the last lap. Practice ranks by the best lap that is not deleted. Post-session penalties are not visible in lap data. A rebuilt order can differ from the official result by a place or two. session.drivers and session.drivers_df share this order.
Last modified on October 30, 2026