Datasets - Formula 1 ML Prediction System

Data Overview

The F1 ML Prediction System is built on 7 years of historical Formula 1 data (2018-2024) collected via the FastF1 API.

Lap Times

139,135 lapsComplete lap-by-lap telemetry data

Race Results

2,537 resultsFinishing positions and points

Pit Stops

4,512 stopsPit stop timing and duration

Weather Records

127 recordsRace day weather conditions

Data Sources

Primary Data Source: FastF1 API

FastF1 is the official Python library for accessing Formula 1 timing data, telemetry, and session information.

Documentation: https://docs.fastf1.dev/
Coverage: 2018-present (live updates)
Data Quality: Official FIA timing data

The data collection script uses FastF1 to gather:

Race results and qualifying positions
Lap-by-lap timing data
Pit stop information
Weather conditions per session
Driver and team metadata

Collection Script: source/src/data/f1_data_collector.py

Dataset Files

All data files are stored in CSV format for easy processing and analysis.

Raw Data Files

Location: data/raw/

race_results.csv
lap_times.csv
pit_stops.csv
weather.csv

2,537 records - Race finishing positions and championship pointsColumns:

Year, Round, RaceName, Driver, DriverCode, Team, 
GridPosition, Position, Points, Status, Laps

Sample Data:

Year,Round,RaceName,Driver,DriverCode,Team,GridPosition,Position,Points,Status,Laps
2024,1,Bahrain Grand Prix,Max Verstappen,VER,Red Bull Racing,1,1,25,Finished,57
2024,1,Bahrain Grand Prix,Sergio Perez,PER,Red Bull Racing,2,2,18,Finished,57

Key Statistics:

Races: 127 (7 years × ~18 races/year)
Drivers: 40+ unique drivers
Teams: 12 teams across the period
DNF Rate: ~12% of all results

139,135 records - Lap-by-lap timing dataColumns:

Year, Round, RaceName, Driver, Lap, LapTime, 
Sector1Time, Sector2Time, Sector3Time, Position

Sample Data:

Year,Round,RaceName,Driver,Lap,LapTime,Sector1Time,Sector2Time,Sector3Time,Position
2024,1,Bahrain,VER,1,92.456,28.123,31.456,32.877,1
2024,1,Bahrain,VER,2,90.234,27.891,31.123,31.220,1

Key Statistics:

Average laps per race: ~1,095 laps (55 laps × 20 drivers)
Fastest lap ever: ~1:14 (Saudi Arabia)
Slowest lap: ~2:30+ (Monaco in rain)

4,512 records - Pit stop timing and durationColumns:

Year, Round, RaceName, Driver, StopNumber, 
Lap, PitInTime, PitOutTime, Duration

Sample Data:

Year,Round,RaceName,Driver,StopNumber,Lap,PitInTime,PitOutTime,Duration
2024,1,Bahrain,VER,1,18,0:28:45.123,0:29:08.456,23.333

Key Statistics:

Average stops per race: ~35-40 stops
Fastest pit stop: ~1.82 seconds (Red Bull record)
Average pit stop: ~2.5-3.0 seconds
Pit lane time loss: ~20-25 seconds total

127 records - Weather conditions per raceColumns:

Year, Round, RaceName, AirTemp, TrackTemp, 
Humidity, Rainfall, WindSpeed, Pressure

Sample Data:

Year,Round,RaceName,AirTemp,TrackTemp,Humidity,Rainfall,WindSpeed,Pressure
2024,1,Bahrain,28.5,42.3,35,0,12.3,1012

Key Statistics:

Dry races: 85% (~108 races)
Wet/damp races: 15% (~19 races)
Hottest race: ~45°C track temp (Middle East)
Coldest race: ~8°C air temp (European autumn)

Processed Data Files

Location: data/processed/ Engineered features ready for machine learning:

race_features.csv
tire_features.csv
pit_features.csv

Main feature dataset for winner prediction modelCreated by: source/feature_engineering.pyFeatures: 21 columns including:

Driver historical stats (wins, podiums, avg position)
Team performance metrics
Circuit-specific experience
Weather conditions
Grid position advantages

Records: Same as race_results (2,537) but with engineered features

Data Collection Process

Setup FastF1 Cache

Configure local cache directory for faster data access:

import fastf1
fastf1.Cache.enable_cache('data/f1_cache/')

Location: data/f1_cache/ (automatically created)

Fetch Race Sessions

Collect data for each race from 2018-2024:

for year in range(2018, 2025):
    for round in range(1, 25):  # Max 24 races per season
        session = fastf1.get_session(year, round, 'R')
        session.load()

Duration: 2-4 hours depending on internet speed

Extract & Transform

Process raw session data into structured CSV files:

Race results from session.results
Lap times from session.laps
Pit stops from session.laps.pick_pit_stops()
Weather from session.weather_data

Save to CSV

Export all data to CSV format:

race_df.to_csv('data/raw/race_results.csv', index=False)
laps_df.to_csv('data/raw/lap_times.csv', index=False)
pits_df.to_csv('data/raw/pit_stops.csv', index=False)
weather_df.to_csv('data/raw/weather.csv', index=False)

Data Collection Notes:

First-time collection takes 2-4 hours
Requires stable internet connection
Some older races may have incomplete data
FastF1 API occasionally rate-limits requests

Data Statistics

Dataset Completeness

Race Results - 100% Complete

All 127 races from 2018-2024 have complete results:

Total Results: 2,537
Finished Races: 2,233 (88%)
DNFs: 304 (12%)
Missing Data: 0%

Every race has full driver positions, points, and status information.

Lap Times - 98% Complete

139,135 laps recorded with minor gaps:

Complete Laps: 136,357 (98%)
Missing Sector Times: 2,778 (2%)
Reasons: Data transmission issues, red flags

Gaps filled using interpolation or removed from training data.

Pit Stops - 95% Complete

4,512 pit stops with some duration data missing:

Complete Records: 4,286 (95%)
Missing Duration: 226 (5%)
Reasons: Penalty stops, drive-through penalties

Missing durations estimated using race average (~23s).

Weather - 90% Complete

127 race weather records with some sensor gaps:

Complete Weather: 114 races (90%)
Partial Data: 13 races (10%)
Missing Fields: Usually wind speed or pressure

Missing weather data filled with circuit historical averages.

Top Drivers (2018-2024)

Max Verstappen

52 wins | 78 podiumsDominant 2021-2024 era

Lewis Hamilton

36 wins | 71 podiumsMercedes dynasty 2018-2021

Charles Leclerc

6 wins | 32 podiumsFerrari’s lead driver

Top Teams (2018-2024)

Red Bull Racing
Mercedes
Ferrari

Wins: 78
Podiums: 156
Championships: 3 (2021, 2022, 2023)
Average Position: 2.1

Data Quality & Validation

Validation Checks
Known Issues
Data Cleaning

Automated data quality checks:✅ No duplicate records - Each driver × race combination is unique✅ Valid ranges - Lap times between 0-300 seconds✅ Position consistency - Finishing positions 1-20 per race✅ Points validation - Points match FIA rules (25-18-15-12-10-8-6-4-2-1)✅ Team continuity - Team names normalized across seasons

Tire Compound Statistics

Soft Compound
Medium Compound
Hard Compound

Usage: 35% of stintsDegradation: +0.08 sec/lap averageOptimal Stint: 15-20 lapsBest Circuits: Monaco, Singapore, Hungary

Updating the Dataset

Recollect Recent Data

Update with latest 2024/2025 races:

python source/recollect_data.py

This incremental update is much faster than full collection.

Re-run Feature Engineering

Regenerate processed features:

python source/feature_engineering.py

Retrain Models

Update ML models with new data:

python train_all_models.py

API Endpoints for Data

The Flask app serves data via REST API:

# Get driver comparison stats
GET /api/compare?driver1=VER&driver2=HAM

# Simulate race with conditions
GET /api/simulate_race?weather=DRY

# Get 2026 predictions
GET /api/season_2026

# Run lap-by-lap race
GET /api/lap_race?weather=DRY&circuit=STANDARD

Server: source/src/app.py or source/app.py

Additional Resources

FastF1 Documentation

Official API documentation and examples

FIA Official Data

Formula 1 official timing and results

Data Collector Script

source/src/data/f1_data_collector.py

Feature Engineering

source/feature_engineering.py

Additional Resources

​Data Overview

Lap Times

Race Results

Pit Stops

Weather Records

​Data Sources

​Primary Data Source: FastF1 API

​Dataset Files

​Raw Data Files

​Processed Data Files

​Data Collection Process

​Data Statistics

​Dataset Completeness

​Top Drivers (2018-2024)

Max Verstappen

Lewis Hamilton

Charles Leclerc

​Top Teams (2018-2024)

​Data Quality & Validation

​Tire Compound Statistics

​Updating the Dataset

​API Endpoints for Data

​Additional Resources

FastF1 Documentation

FIA Official Data

Data Collector Script

Feature Engineering

Build docs developers (and LLMs) love

Data Overview

Data Sources

Primary Data Source: FastF1 API

Dataset Files

Raw Data Files

Processed Data Files

Data Collection Process

Data Statistics

Dataset Completeness

Top Drivers (2018-2024)

Top Teams (2018-2024)

Data Quality & Validation

Tire Compound Statistics

Updating the Dataset

API Endpoints for Data

Additional Resources