Skip to main content
tif1 does not replace fastf1. The timing and telemetry data in tif1 is generated with fastf1, then processed and published to the TracingInsights data repositories. tif1 differs in how that data is delivered. It keeps the fastf1 API and column names, so most fastf1 code runs after an import change.

Comparison


1. Download only the necessary data

fastf1 loads a whole session at a time. session.load() downloads the laps, telemetry, weather, and messages of every driver. The car data and the position data each come as one file for all drivers. session.load(telemetry=False) skips them, but fastf1 has no option to load one driver or one lap. tif1 downloads per file, when a property is read. A new session downloads no data. The example below downloads three files: the driver list, the lap table, and the telemetry of lap 19.
  • The first lap of telemetry needs three small files.
  • The cache stores only the files that are used.
  • One lap of telemetry is one file. The lap table of a session is usually one file.

2. No F1TV account needed

In October 2026 Formula 1 started to require authentication on its data APIs. Since fastf1 3.9.0, most fastf1 features need an F1TV Access (or higher) subscription. fastf1 signs in to an F1 account in one of three ways: the F1_USERNAME and F1_PASSWORD environment variables, a prompt in the terminal, or a browser add-on. The login token usually expires after four days. fastf1 cannot refresh it, so a new sign-in is necessary. With the credentials in environment variables, fastf1 signs in again automatically. tif1 does not call the F1 data APIs. It downloads files that were already generated with fastf1 and published to the CDNs. No F1 account, subscription, or sign-in is necessary, and no token expires. Servers, CI jobs, and notebooks need no stored credentials. Source: fastf1 F1TV Account Authentication.

3. No rate limits on the session data

fastf1 limits

fastf1 sends every request through its own rate limiter (fastf1/req.py in fastf1 3.9.0). Requests to these hosts all count toward the same limits:
  • the F1 live-timing archive (livetiming.formula1.com and the mirror livetiming-mirror.fastf1.dev)
  • the MultiViewer API
  • jolpica-f1
  • the schedule file on GitHub
Responses from the fastf1 cache do not count (fastf1 Rate Limits and Caching). An uncached session load sends one request for each live-timing stream, plus jolpica-f1 requests for the results. With an empty cache, these workloads reach 500 requests per hour and stop with an error:
  • loops over a full season (24 weekends, several sessions each)
  • full-grid comparisons across many races
  • backtests or model training on many sessions
  • background jobs that fill a cache
This is one reason why tif1 was built for the app at tracinginsights.com/analysis.

jolpica-f1 limits

The jolpica-f1 Rate Limits guide gives these limits for clients without a token:
  • 4 requests per second
  • 500 requests per hour
Above these limits, jolpica-f1 returns HTTP 429 Too Many Requests (“Request was throttled”). The guide says that the limits will go down when token access is added. The Terms of Use say that abuse or excessive use can get a client blocked.

tif1

tif1 reads timing and telemetry from the TracingInsights data repositories (TracingInsights/{year}). It reads the Ergast tables from the TracingInsights Stats repository, one static file for each table and round. All files go through public CDNs: jsDelivr first, then Hugging Face buckets, then StaticDelivr. A season of race results is 24 files that download in parallel, not a paged API walk.
  • No API keys or accounts.
  • No request quota. max_concurrent_requests limits the parallel requests instead.
  • Retries, CDN failover, and a circuit breaker.
  • SQLite and in-memory LRU caches, so a repeated read sends no request.
In normal use tif1 does not call jolpica-f1, so the jolpica-f1 limits do not apply. tif1 calls jolpica-f1 only:
  • for a round that is not in the Stats repository yet. The repository is updated a few hours after jolpica-f1 has the data.
  • for queries that the files do not cover: season and driver lists, status counts, status= filters, queries without a season, and current/last.
These calls keep to the stricter of the fastf1 and jolpica-f1 limits:
  • At most 4 requests in any second and 200 in any hour, shared by all threads of the process. When a limit is reached, tif1 waits. It does not raise an error.
  • On HTTP 429, tif1 waits for the Retry-After time and tries again, up to 3 times.
  • Answers stay in the cache for ergast_ttl_hours, so a repeated query sends no request.
  • ergast_rate_limit_burst and ergast_rate_limit_hourly change the limits. See Configuration.
The MultiViewer API (circuit data) is served through the free Cloudflare CDN at no cost to MultiViewer, so tif1 does not rate-limit it. tif1 caches its answers: a circuit layout is downloaded once and kept, and the circuit index is kept for schedule_ttl_hours.

4. Built-in charts

fastf1 supplies the data and plotting helpers for colors and styles. The chart code is written by hand in matplotlib. tif1 also has 22 plot_*() functions. Each one loads the data, runs the analysis, and draws the chart in one call:
Top speeds by team, drawn by one call to tif1.plot_top_speeds

Chart functions

Every chart accepts year, event, session, save_path, and dpi. Most charts also accept a drivers, teams, or compounds filter. Most charts have a tutorial in the Tutorials section.

5. No IP blocking

fastf1 reads timing and telemetry from livetiming.formula1.com. The host refuses some networks. In FastF1 discussion #949, a Google Colab runtime got HTTP 403 from it, and another network got HTTP 200 for the same URL. fastf1 then tries its mirror, livetiming-mirror.fastf1.dev. The mirror does not have every session. tif1 sends plain HTTPS requests to public CDNs (jsDelivr, then Hugging Face buckets, then StaticDelivr), with no API key, cookie, or login. Any network that can reach one of these CDNs can load the data, including cloud runtimes and CI runners.

6. Extra data and features

Mini sectors

Formula 1 timing divides each of the three sectors into mini-sectors. The number of mini-sectors in a sector is different at each circuit. Mini-sector colors show where a lap was a personal best or an overall best, below the sector level. The fastf1 laps data has no mini-sector columns. The TracingInsights data pipeline adds them from OpenF1 when the session matches:
  • ms1, ms2, and ms3 hold the codes for sectors 1, 2, and 3, usually one digit for each mini-sector.
  • Each digit is a color code, for example 1 for green, 3 for purple, and 5 for the pit lane. The data reference lists all codes.

Other additions


When not to use tif1

  • Live timing. tif1 has no live timing. It is an archive library (timing and telemetry from 2018, Ergast results from 1950), and a session is available only after it ends. For live data during a session, use fastf1 (fastf1.livetiming, which signs in with an F1TV account).
  • Newest data. fastf1 reads the F1 live-timing archive directly. tif1 can load a session only after its files are published, in under a minute after fastf1 has the data.
  • fastf1 internals. Code that uses fastf1 modules outside the documented API (for example fastf1.livetiming or fastf1._api) needs fastf1. tif1 does not include them. fastf1.ergast has its counterpart in tif1.ergast.

Migration

tif1 keeps the fastf1 column names for laps and telemetry, so many scripts need only an import change. The migration guide lists the differences, such as the 0/1 DRS flag and the missing SessionTime and Date telemetry columns.
See the Migration from fastf1 guide for the steps, or start with the Quickstart.

Quickstart

Load a first session and read laps and telemetry.

Charts API

All 22 chart functions.

Migration Guide

Move an existing fastf1 project to tif1.

Tutorials

tif1 charts applied to real race analysis.
Last modified on October 30, 2026