Suzuka 2025 F1 Prediction Model

Built for Fans. Backed by Data. Powered by Otto.rentals

Welcome to Otto's open-source F1 prediction engine — a machine learning project designed to forecast race results, generate driver insights for Formula 1 fans.

This repo currently supports predictions for the Suzuka and Shanghai Grand Prix in 2025.

---

What’s Inside

Module	Description
suzuka_f1.py	Full-featured simulator for the Suzuka GP. Uses a Random Forest model, historical data, custom heuristics, and outputs 7 social-ready plots.
shanghai_f1.py	Predicts results for the Shanghai GP using Random Forest + team/driver embeddings. Includes fallback handling and grid simulation.
suzuka_2025_predictions.csv	Ranked table of driver predictions with podium/points probabilities.
*.png images	Visual outputs (charts, comparisons, podiums) generated by the model for social sharing.

---

Why We Built This

At Otto, we're all about helping car rental businesses go digital. But we’re also car enthusiasts, so we wanted to do something fun. We built this to:

• Use machine learning to predict Formula 1 race results
• Create cool, shareable insights for F1 fans
• Show off what Otto can do and add our own unique spin

---

Installation

pip install fastf1 pandas numpy scikit-learn matplotlib seaborn pillow

We recommend using a virtual environment (e.g. venv or conda) for clean installs.

---

How to Use

1. Predict Suzuka GP

python suzuka_f1.py

Outputs:

• suzuka_2025_race_prediction.png: Full grid predictions w/ error bars
• suzuka_2025_podium_probability.png: Top podium chances
• suzuka_2025_points_probability_social.png: Points finish visual for IG
• suzuka_2025_driver_comparison.png: Head-to-head radar chart
• suzuka_2025_grid_vs_prediction.png: Quali vs prediction
• suzuka_2025_team_performance.png: Team-level performance
• suzuka_2025_prediction_uncertainty.png: Prediction risk vs confidence
• suzuka_2025_predictions.csv: Tabular output for analysis

---

Sample Visuals

Here’s a preview of the kinds of outputs we generate:

Full Grid Prediction	Podium Probability Chart

More visuals available in the *.png files in the repo. All plots are Otto-branded and social-ready.

---

Predicted Results for Suzuka 2025

Podium Finish Prediction

Predicted Pos	Driver	Team	Grid Pos	Avg Finish	Uncertainty	Podium %	Points %
1	Lando Norris	McLaren	2	2.518	2.83014	86.2	95.5
2	Max Verstappen	Red Bull Racing	1	2.892	2.8419	85.4	95.2
3	Oscar Piastri	McLaren	3	3.768	2.86957	62.1	94.5

Full Grid Prediction

Predicted Pos	Driver	Team	Grid Pos	Avg Finish	Uncertainty	Podium %	Points %
4	Charles Leclerc	Ferrari	4	4.553	3.04573	36.3	93
5	George Russell	Mercedes	5	5.918	3.01237	11.6	92.4
6	Lewis Hamilton	Ferrari	8	7.106	3.43607	8.8	87.6
7	Andrea Kimi Antonelli	Mercedes	6	7.805	3.43843	3.4	84.7
8	Alexander Albon	Williams	9	9.461	3.41821	1.4	67.7
9	Isack Hadjar	VCARB	7	9.981	3.36035	0.5	65.2
10	Pierre Gasly	Alpine	11	11.201	3.83605	1.1	45
11	Carlos Sainz Jr.	Williams	12	12.14	3.73341	0.5	33.5
12	Fernando Alonso	Aston Martin	13	12.377	3.96921	1	32
13	Oliver Bearman	Haas F1 Team	10	12.489	3.35785	0	28.6
14	Liam Lawson	VCARB	14	13.673	3.69537	0.3	20.4
15	Yuki Tsunoda	Red Bull Racing	15	14.239	3.76484	0.3	17.9
16	Nico Hülkenberg	Kick Sauber	16	15.083	3.58678	0	12.4
17	Gabriel Bortoleto	Kick Sauber	17	15.68	3.50655	0.3	9.9
18	Esteban Ocon	Haas F1 Team	18	15.933	3.6221	0.4	9.1
19	Jack Doohan	Alpine	19	16.404	3.5326	0.4	7.5
20	Lance Stroll	Aston Martin	20	16.779	3.45458	0	7.9

---

Model Summary

We use a hybrid approach for accuracy + engagement:

• Model Type: Random Forest Regressor (from sklearn)
• Training Data: FastF1 results from 2022–2024
• Features:
  • Grid position
  • Track-specific performance
  • Driver form (rolling 5-race average)
  • Team strength (recent trends)
  • Driver experience & special case handling
• Simulations: 1000 Monte Carlo runs per driver
• Extras: Heuristics for rookies, team changes, rain risk, Suzuka experts

---

File-by-File Breakdown

File	Purpose
suzuka_f1.py	End-to-end race simulation and visual generation for Suzuka GP
suzuka_v1.py	Race simulation with minimal visuals
suzuka_2025_predictions.csv	Predicted results (driver, team, average finish, probabilities)
*.png	High-res images for use on social platforms like Instagram, X (Twitter), TikTok
README.md	You’re reading it

---

For Non-Technical Readers

You don’t need to understand code to enjoy this:

• We’re using past F1 data to predict who finishes where in future races.
• We simulate 1,000 races based on real qualifying data, driver history, team performance, and random events like crashes or rain.
• The charts help show how confident we are in each prediction (e.g. Verstappen might win, but is he 90% likely or 55%?).
• Think of it as fantasy F1 meets data science.

---

Built With

• fastf1 for live/historical race data
• pandas, numpy for data processing
• scikit-learn for modeling
• matplotlib, seaborn, pillow for graphics
• Python 3.9+

---

Contributions & Replication

We welcome forks and feedback! Try replicating predictions for other 2025 races by:

• Updating drivers.py with real qualifying results
• Switching the circuit name and location flags
• Running the script and sharing results

---

Author

• Frank Ndungu – Software Engineer | CX at Otto Rentals

---

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Let’s Talk

Love cars? Like data? Run a motorsport or F1 page?

• Let’s collaborate on custom graphics.
• DM @otto.rentals on IG or X.
• We can help you bring data storytelling to racing.

---

Powered by Otto

The SaaS platform digitizing car rentals across Africa.
We build products for hosts, renters — and now, Formula 1 fans.