tif1

A fast, canonical Formula 1 data library fetched from TracingInsights data https://github.com/TracingInsights/2026 (2018-current).

Data is available ~30 minutes after the session ends. tif1's files are built from the same F1 timing data that fastf1 reads, so they appear a few minutes after that data reaches fastf1: tif1 has more data and does more processing to enrich it before making it available.

Why use tif1 instead of fastf1?

First, a note of respect: fastf1 is a great library and does a lot for the F1 data community — I've used it myself. tif1 started as a personal project with a different approach in a few areas. Here's how the two differ:

1. Fetch only what you need — no full-session downloads

fastf1 is session-oriented: you call session.load() and, by default, it downloads the whole session (all drivers, every lap, every telemetry point) before the data is usable. Telemetry comes as the session's full car-data and position streams for every driver, so even one lap of telemetry means downloading and storing all of it.

tif1 is lazy and fine-grained. A Session object costs almost nothing until you touch a property, and you can pull exactly the one lap of telemetry you need in seconds:

import tif1

session = tif1.get_session(2024, "Monaco Grand Prix", "Race")

# Fetch only Verstappen's lap 19 telemetry: the driver list, the session's lap
# table and that one lap's telemetry file, nothing else.
telemetry = session.get_driver("VER").get_lap(19).telemetry
print(telemetry[["Time", "Speed", "Throttle"]].head())

You pay for exactly the bytes you use: less waiting, less hard-disk space, less bandwidth. That efficiency also makes tif1 the more energy- and environmentally-friendly choice for large or repeated analysis workloads.

2. No rate-limited API in the loop

fastf1 pulls historical and current data from the Ergast-compatible jolpica-f1 API and the unofficial F1 live-timing API. jolpica-f1 rate-limits unauthenticated access to:

• Burst limit: 4 requests per second
• Sustained limit: 500 requests per hour

and answers excessive traffic with HTTP 429 Too Many Requests ("Request was throttled"). Those limits are scheduled to decrease further as token-based access rolls out.

Apps that loop over many sessions in a single run — a whole season, an entire grid, backtesting, live dashboards — can approach these ceilings quickly. It's a real consideration in production, and one of the reasons I built tif1 for my own app at tracinginsights.com/analysis.

tif1 does not call a rate-limited API. It reads public static files from free, global CDNs (jsDelivr primary, Hugging Face bucket fallback, StaticDelivr backup) with automatic failover, retries, and a circuit breaker — no API keys, no tokens, no request quotas to track in your code.

Sources: jolpica-f1 Terms of Use and the jolpica-f1 Rate Limits guide.

3. Batteries included — 22 one-call chart functions

If you enjoy hand-building charts, fastf1 gives you all the pieces to do so. tif1 additionally ships 22 ready-made chart functions that load the data, run the analysis, and plot in a single call:

import tif1
import matplotlib.pyplot as plt

# Top speeds by team at the 2023 Italian Grand Prix — one call.
fig, ax = tif1.plot_top_speeds(2023, "Italian Grand Prix", "Q")
plt.show()

# Alonso's lap times at the 2023 Azerbaijan Grand Prix — one call.
fig, ax = tif1.plot_driver_laptimes(2023, "Azerbaijan", "R", drivers=["ALO"])
plt.show()

# Or save straight to a PNG, no interactive session needed.
tif1.plot_telemetry_comparison(
    2024, "Monaco", "Q", drivers=["VER", "LEC"], save_path="monaco_ver_lec.png"
)

From track maps, speed traces, and telemetry comparisons to tire degradation, qualifying grids, GG diagrams, and lap-time heatmaps, there is a ready-made analysis for almost every F1 question. See the Charts tutorial section and the Charts API reference for the full list.

4. Works from anywhere — no IP restrictions

The official F1 live-timing endpoints that fastf1 uses have been reported to block VPNs, cloud/data-center IPs, and some hosting providers, which can make fastf1 tricky to run on servers. tif1 avoids that class of problem by serving data from static CDN files.

tif1 serves its data from free global CDNs (jsDelivr hosting the TracingInsights data repos, with Hugging Face buckets as the fallback and StaticDelivr as the last-resort backup), so it works from home, office, cloud, CI runners and notebooks alike, with no proxies.

5. Extra data — mini sectors and more

tif1 also includes a few things beyond fastf1's current scope:

• Mini-sector data — lap data carries the OpenF1 mini-sector status of every lap (ms1/ms2/ms3: one digit per mini-sector for yellow, green, purple, pit lane, and so on), when OpenF1 data can be matched to the session.
• Per-lap weather — weather conditions (AirTemp, TrackTemp, Rainfall, wind, …) merged into the lap rows.
• Derived channels — DriverAhead, DistanceToDriverAhead, and acceleration channels (AccelerationX/Y/Z) computed and included in the telemetry.
• Flexible backends — pandas and polars, with SQLite + in-memory LRU caching under the hood.
• FastF1-style schema — fastf1 column names for the lap and telemetry data (Sector1Time, SpeedST, nGear, …), so migration is often a one-line import change.

One honest trade-off: data becomes available in tif1 a few minutes later than in fastf1, because the data is enriched and processed before being published.

Installation

pip install tif1

Requires Python 3.11 or newer. Versions 3.11 through 3.15 are tested in CI.

Optional Dependencies

# For development
pip install tif1[dev]

Quick Start

import tif1

# Get the event schedule for a year (a DataFrame, one row per event)
events = tif1.get_events(2025)
print(events["EventName"].tolist())  # ['Australian Grand Prix', 'Chinese Grand Prix', ...]

# Get sessions for an event
sessions = tif1.get_sessions(2025, "Chinese Grand Prix")
print(sessions)  # ['Practice 1', 'Sprint Qualifying', 'Sprint', 'Qualifying', 'Race']

# Get a session
session = tif1.get_session(2024, "Monaco Grand Prix", "Race")

# Get all drivers as DataFrame
print(session.drivers_df)

# Get all laps
laps = session.laps
print(laps.head())

# Get specific driver
ver = session.get_driver("VER")
ver_laps = ver.laps

# Get telemetry for a specific lap
lap = ver.get_lap(19)
telemetry = lap.telemetry
print(telemetry[["Time", "Speed", "Throttle"]].head())

Features

• Fast: Multi-CDN access via jsDelivr (primary), Hugging Face buckets (fallback), and StaticDelivr (backup) with automatic failover and SQLite caching
• No need to session.load() - only the required data is fetched when necessary. You can just get the telemetry data of any specific lap within seconds.
• Canonical: Focused tif1 API surface
• Complete: Lap times, sectors, telemetry, tire compounds, and more
• Historical: Data from 2018-current
• Reliable: Automatic retry logic with circuit breaker and multi-CDN fallback
• Async: Parallel data fetching for better performance
• Optimized: SQLite cache holding compressed JSON payloads and ready-built DataFrames
• Flexible: Supports both pandas and polars backends
• Validated: Optional data validation with Pydantic models
• Configurable: .tif1rc configuration file support
• Type-Safe: Comprehensive type hints for IDE support
• Jupyter-Ready: Rich HTML display in notebooks after import tif1.jupyter

Advanced Usage

Canonical API Notes

• tif1 exposes canonical data/session APIs directly (get_session, Session, Laps, Lap, Driver).
• Event/session schedules come from the f1schedule files that fastf1 also reads (theOehrly/f1schedule, schedule_{year}.json). Past seasons are served from copies vendored in the package. For the current and future seasons, tif1 downloads the live file, caches it in the SQLite cache for schedule_ttl_hours (default 6, env TIF1_SCHEDULE_TTL_HOURS), and falls back to the last cached copy, then the vendored copy, when the download fails.

Configuration File

import tif1

# Create ~/.tif1rc with your settings
# {
#   "max_retries": 5,
#   "validate_data": true,
#   "lib": "polars",
#   "cdns": [
#     "https://cdn.jsdelivr.net/gh/TracingInsights",
#     "https://huggingface.co/buckets/tracinginsights",
#     "https://cdn.staticdelivr.com/gh/TracingInsights"
#   ],
#   "cdn_use_minification": false
# }

# Get configuration
config = tif1.get_config()
print(f"Max retries: {config.get('max_retries')}")

# Set configuration
config.set("validate_data", True)
config.save()

Circuit Breaker & CDN Fallback

import tif1

# Check circuit breaker status
cb = tif1.get_circuit_breaker()
print(f"Circuit breaker: {cb.state}")

# Check CDN sources
cdn = tif1.get_cdn_manager()
sources = cdn.get_sources()
print(f"Available CDNs: {len(sources)}")

# Reset if needed
if cb.state == "open":
    tif1.reset_circuit_breaker()

Fastest Laps & Telemetry (Optimized for Speed)

import tif1

session = tif1.get_session(2024, "Monaco Grand Prix", "Race")

# Get fastest lap per driver
fastest_by_driver = session.get_fastest_laps(by_driver=True)
# Pandas backend: LapTime is Timedelta, LapTimeSeconds is numeric seconds
print(fastest_by_driver[["Driver", "LapTime", "LapTimeSeconds"]].head())

# Get overall fastest lap
overall_fastest = session.get_fastest_laps(by_driver=False)

# Get driver's fastest lap
ver = session.get_driver("VER")
ver_fastest = ver.get_fastest_lap()

# Get telemetry from overall fastest lap
fastest_tel = session.get_fastest_lap_tel()

# Get telemetry from each driver's fastest lap (parallel fetching!)
fastest_tels = session.get_fastest_laps_tels(by_driver=True)

# Get telemetry for specific drivers' fastest laps
top3_tels = session.get_fastest_laps_tels(by_driver=True, drivers=["VER", "HAM", "LEC"])

# Get specific driver's fastest lap telemetry
ver_fastest_tel = ver.get_fastest_lap_tel()

Logging

import tif1
import logging

# Enable debug logging
tif1.setup_logging(logging.DEBUG)

session = tif1.get_session(2024, "Monaco Grand Prix", "Race")

Cache Management

import tif1

# Get cache instance (SQLite)
cache = tif1.get_cache()
print(f"Cache location: {cache.cache_dir}")

# Clear cache
cache.clear()

# Disable caching for a session
session = tif1.get_session(2024, "Monaco Grand Prix", "Race", enable_cache=False)

Error Handling

import tif1

try:
    # Monaco had no Sprint: a session that is not on the event's schedule raises ValueError.
    # An event name that matches no event raises tif1.EventNotFoundError, also a ValueError.
    session = tif1.get_session(2024, "Monaco Grand Prix", "Sprint")
    laps = session.laps
except ValueError as e:
    print(f"Invalid event or session: {e}")
except tif1.DataNotFoundError:
    print("Data not available")
except tif1.NetworkError:
    print("Network error occurred")
except tif1.InvalidDataError:
    print("Data is corrupted")

Data Available

• Lap times and sectors (S1, S2, S3)
  • LapTime is canonicalized as Timedelta on pandas backends
  • LapTimeSeconds is provided as a numeric helper for sorting/comparisons
• Tire compounds and stint information
• Telemetry: speed, throttle, brake, RPM, gear, DRS
• Position data (X, Y, Z coordinates)
• Acceleration data (X, Y, Z axes)
• Distance and relative distance
• Driver ahead and distance to the driver ahead
• Driver and LapNumber on every telemetry row

Data Reference

tif1 serves every session from the TracingInsights GitHub telemetry data repositories (per-year repos like {year}, 2018-current). Each session is stored as a set of raw JSON files. The fields of each file are documented field-by-field in DATA_REFERENCE.md:

File	Description
{driver}/laptimes.json	Per-driver lap timing: lap/sector times, session times, speed traps (vi1/vi2/vfl/vst), tire compound/life/stint, OpenF1 mini-sectors (ms1/ms2/ms3), qualifying segment (qs), position, track status, personal best, pit in/out times, and data-quality flags. Also enriched with per-lap weather when it can be matched.
session_laptimes.json	The same lap fields for every driver of the session in one file (rows in alphabetical driver order).
{driver}/{lap}_tel.json	Per-lap telemetry, about 7-8 samples per second (car and position samples merged): RPM, speed, gear, throttle, brake, DRS, distance, relative distance, X/Y/Z position, driver-ahead info (DriverAhead, DistanceToDriverAhead), and derived acceleration (acc_x/acc_y/acc_z).
weather.json	Environmental conditions recorded ~once per minute: air/track temperature, humidity, pressure, rainfall, wind direction, wind speed.
rcm.json	Race control messages: time, category, message, status, flag, scope (Track/Sector/Driver), affected track sector, driver number, lap.
drivers.json	Static driver metadata: 3-letter code, team, car number, first/last name, team color (hex), headshot URL.
corners.json	Circuit corner geometry: corner number, X/Y coordinates, angle, distance from start/finish, circuit rotation.

The raw JSON files use short abbreviated keys (e.g. s1, vi1, wWS). tif1 maps them to fastf1-compatible PascalCase columns (Sector1Time, SpeedI1, WindSpeed) in its DataFrames — see the Data Schema Reference for the canonical DataFrame layout.

Session Types

• Practice 1, Practice 2, Practice 3
• Qualifying
• Sprint, Sprint Qualifying, Sprint Shootout
• Race

Development

# Install with uv
uv sync --all-extras

# Install git hooks with prek (commit + push)
uv run prek install

# Run the full local quality gate (same checks as CI lint job)
uv run prek run --all-files

# Run example
uv run python examples/async_loading.py

# Run tests (parallel by default via xdist)
uv run pytest tests/ -v

# Run linting
uv run ruff check src/ tests/

# Run type checking
uv run ty check src/tif1

# Run benchmarks (serial for stable timing)
uv run pytest -o addopts='' tests/benchmarks/ -v -m benchmark --benchmark-only --no-cov -n 0

Documentation

Full documentation available at: tif1.tracinginsights.com

Contributing

Contributions are welcome! A few ideas what you can contribute:

Improve documentation.
Add more tests.
Improve performance.
Found a bug? Fix it!
Made an article about tif1? Great! Let's add it into the README.md.
Don't have time to code? No worries! Just tell your friends and subscribers about the project. More users -> more contributors -> more cool features.

License

MIT License - see LICENSE file for details.

What's available from fastf1

Verified against docs.fastf1.dev (FastF1 API reference) and current tif1 code/tests.

Verification basis:

• FastF1 docs: session/event API, core.Session, core.Laps, core.Lap, core.Telemetry

Status key:

• supported: implemented with compatible behavior
• partial: implemented with simplified/placeholder behavior
• missing: not implemented yet

Top-level and event/session API

FastF1 API	tif1 status	Notes
get_session	supported	round/event aliases supported
get_event	supported	event lookup supported
get_event_schedule	supported	fastf1 schedule columns plus GmtOffset; testing rows (round 0) kept unless include_testing=False
Cache.enable_cache	supported	compatibility shim
set_log_level	supported	compatibility shim
plotting.setup_mpl	supported	compatibility helper module
utils module	partial	core helpers implemented; not full parity
get_testing_session	missing	planned
get_testing_event	missing	planned
get_events_remaining	missing	planned

core.Session

FastF1 API	tif1 status	Notes
load	supported	lazy-load compatible
laps	supported	main lap timing surface
get_driver	supported	driver validation present
get_circuit_info	partial	corners and rotation from corners.json; marshal_lights / marshal_sectors are empty
results	partial	classification order, status and laps; no Q1/Q2/Q3 or Time; GridPosition and Points are NaN
session_info	partial	minimal metadata dict (Year, EventName, SessionName)
session_start_time / t0_date	partial	placeholder: always NaT
session_status	partial	returns the race control messages table
track_status	partial	returns the per-lap TrackStatus column of session.laps
race_control_messages	partial	available when source data exists
weather_data	partial	passthrough compatibility

core.Laps, core.Lap, core.Telemetry

FastF1 API	tif1 status	Notes
Laps.pick_driver(s)	supported	implemented
Laps.pick_lap(s)	supported	implemented
Laps.pick_fastest	supported	implemented
Laps.get_car_data / get_pos_data	supported	both return the merged telemetry (car and position channels)
Laps.get_weather_data	partial	returns the whole session's weather table
Laps.split_qualifying_sessions	supported	splits on the QualifyingSession (Q1/Q2/Q3) column; returns three full copies when that column is missing
Lap.get_telemetry / get_car_data / get_pos_data	supported	all return the merged telemetry (car and position channels)
Telemetry.add_distance	supported	keeps an existing Distance column (drop_existing is ignored)
Telemetry.add_relative_distance	supported	keeps an existing RelativeDistance column
Telemetry.add_differential_distance	supported	implemented
Telemetry.merge_channels / resample_channels	partial	extra fastf1 keyword arguments are ignored; merge_channels merges on the nearest Time, resample_channels keeps numeric columns only
Telemetry.slice_by_lap / slice_by_time / slice_by_mask	supported	interpolate_edges is accepted but ignored
Telemetry.fill_missing	supported	implemented
Telemetry.add_driver_ahead	partial	the columns come precomputed in the data; on a frame without them it adds empty (None/NaN) columns
Telemetry.add_track_status	partial	adds a constant TrackStatus of "1" when the column is missing

What's planned to be added in tif1 from fastf1

High-priority API gaps for parity with docs.fastf1.dev:

• get_testing_session
• get_testing_event
• get_events_remaining
• deeper parity for Session metadata/classification fields

What's out of scope for tif1 from fastf1

• Ergast (fastf1.ergast)
• LiveTiming (fastf1.livetiming)