tif1 keeps the fastf1 API shape for sessions, laps, and telemetry. This guide shows the similarities and the key differences.
Comparison at a Glance
Migration Steps
1
Update Imports
Change
import fastf1 to import tif1.2
Update get_session Calls
Full GP names are the clearest. Circuit, city, and country names such as
"Monaco" or "Spa" also resolve to the event of that season. Session names are case-insensitive, so "Race", "race", and "R" all work. In fastf1, session.laps is available only after session.load(). In tif1, session.load() is optional because tif1 loads data lazily.3
Check Filters
tif1.Laps has every pick_ method of fastf1.core.Laps, with the same names and the same rules. For example, pick_drivers upper-cases the codes, pick_quicklaps keeps the laps strictly below the threshold, and pick_fastest takes the IsPersonalBest laps. pick_driver accepts a 3-letter code or a driver number. Session adds direct accessors. session.get_fastest_laps() skips only the deleted laps, so it can pick a lap that is not a personal best.Key API Differences
1. Lazy Loading
tif1 uses lazy loading by default. Asession.load() call is not necessary. fastf1 raises DataNotLoadedError when code reads session.laps before session.load().
2. Telemetry Access
Both libraries have thelap.telemetry property and the lap.get_telemetry() method. Telemetry code needs no change. get_telemetry(frequency=...) accepts fastf1’s values: "original", or a sampling rate in Hz.
tif1
3. Driver Information
session.drivers holds driver numbers as strings in classification order, as in fastf1. session.drivers_df adds the codes, teams, and colors in one DataFrame.
session.get_driver() returns the results row of the driver, as in fastf1. The row holds Position, GridPosition, Points, Status, and the other result columns, and dnf is a property. tif1 adds laps, get_lap(), get_fastest_lap(), and get_fastest_lap_tel(). The identifier can be the abbreviation, the car number, or the full name. An unknown identifier raises DriverNotFoundError, which is a ValueError too, so a fastf1 except ValueError handler still catches it.
4. Lap Columns
tif1 laps have the 31 fastf1 columns in fastf1’s order, and some extra columns:- On pandas, an
indexcolumn comes first.drvandteamrepeat the driver and the team from the source file. - Every lap row carries a weather sample that the source data matched to the lap (
WeatherTime,AirTemp,TrackTemp,Rainfall, and others). That match is misaligned for many laps.laps.get_weather_data()does not read these columns: it picks the sample of each lap fromsession.weather_datawith the fastf1 rule. ItsTimecolumn is the session time of the sample. Its rows keep the laps index, where fastf1 uses the index of the weather samples. LapTimeSecondsholds the lap time as float seconds. It comes after the weather columns, near the end of the table.ms1,ms2, andms3hold mini-sector color codes for each lap.
Driver, Team, Compound, and TrackStatus are categorical columns, and Deleted is a nullable boolean. Some pandas operations treat categoricals differently from strings.
5. Telemetry Differences
Telemetry frames have the fastf1 columns in fastf1’s order. tif1 addsAccelerationX, AccelerationY, AccelerationZ, Driver, and LapNumber at the end. Some values differ from fastf1:
DRSis a 0/1 flag. fastf1 keeps the raw values, where 10, 12, and 14 mean open. The fastf1 filtertel["DRS"] >= 10never matches in tif1. Usetel["DRS"] == 1.RPM,Speed, andThrottleare floats, because the source values are interpolated.nGearandDRSare nullable integers (Int64).Sourceis"merged", andStatusis always"OnTrack". The source data is already the merge of car data and position data, and it has no on-track channel. The edge samples ofinterpolate_edgesand the gaps thatfill_missing()fills have theSource"interpolation", as in fastf1.Timestarts at zero on a single lap. The telemetry of several laps counts from the start of the first lap, as in fastf1. The sample that two laps share at their boundary appears once.SessionTimeholds the session time andDatethe wall-clock time of each sample.DriverAheadholds driver numbers.DistanceToDriverAheadis in meters. When nobody is ahead, it is NaN on pandas and null on polars.session.car_dataandsession.pos_dataare dicts keyed by driver number, as in fastf1. Both hold the same merged frame. ItsTimeis the session clock, equal toSessionTime, as in fastf1. The first read of a key fetches the telemetry of that driver.
tif1.utils.delta_time(ref_lap, comp_lap) uses the fastf1 algorithm. It returns (delta, ref_tel, comp_tel). A positive delta means the comparison lap is slower at that point. The delta is an estimate from the distance samples, so it can differ from the real lap time gap.
6. Session State
fastf1 reads the track status, the session status, and the session information from live-timing streams. Those streams need a Formula 1 account with an active F1TV subscription. tif1 reads the same values from thesession_info.json file of each session (2018 onward), so no account is necessary. The shapes, the status codes, and the messages are the fastf1 ones.
Without
session_info.json, tif1 derives these values from the race control messages, the laps, and the schedule. The derived values are an approximation of the live streams.
laps.pick_track_status() filters on the TrackStatus codes of each lap, as in fastf1. It does not read session.track_status.
7. No Live Timing
tif1 reads historical files only.session.load(livedata=...) raises NotImplementedError for any value other than None. fastf1.livetiming, fastf1.set_auth_method(), and F1AuthError have no tif1 equivalent.
8. Polars Backend
Withlib="polars", the session returns polars classes with the fastf1 methods: PolarsLaps, PolarsLap, PolarsTelemetry, PolarsSessionResults, and PolarsDriver. Time columns are Duration("ns"), and LapStartDate and Date are Datetime("ns"). A polars filter(), select(), sort(), or with_columns() returns a plain polars.DataFrame. The df.f1 namespace gives the tif1 methods back.
df.f1 namespace.
9. Async Code
session.laps already fetches the lap data in parallel. In code that already runs an event loop, await laps_async():
10. Ergast Data
fastf1.ergast code works with an import change. The answers come from the TracingInsights Stats repository (one file per table and round, no rate limit). jolpica-f1 is asked only when that repository cannot answer. By default a Stats-backed query returns every row instead of fastf1’s first 30; pass limit to page.
session.results also carries the official classification (grid, points, status, total race time, Q1 to Q3), so the GridPosition and Points columns of races and sprints are filled. Refer to Ergast.
Feature Parity
Archive Only
tif1 focuses on historical data: timing and telemetry from 2018, Ergast results, lap times and standings from 1950. It does not support Live Timing.Track Maps
session.get_circuit_info() returns the corners, the marshal lights, the marshal sectors, and the rotation. For seasons with an old corners.json, the marshal data comes from the MultiViewer API, as in fastf1. Six track-map charts, such as plot_track_speed_map and plot_gear_shifts, draw the layout.Weather Data
session.weather holds the raw samples. laps.get_weather_data() returns the sample of each lap from them, with the fastf1 rule. Every lap row also carries weather columns from the source data (AirTemp, TrackTemp, Rainfall, …), but that match is misaligned for many laps.Categorical Optimization
On the pandas backend,
Driver, Team, Compound, and TrackStatus are categorical columns. Some pandas operations treat categoricals differently from strings.Ready to start? Read the Quickstart guide.
Related Pages
Why tif1?
The full fastf1 comparison — rate limits, granularity, and more.
Quickstart
Get started quickly.
API Reference
Core API details.
FastF1 Compat
Compatibility layer.
Best Practices
tif1 patterns.