Data Science2024
Formula 1 Data Analysis
A structured tour of data-visualization craft, using seven decades of Formula 1 as the racetrack.
- Chart-form judgment
- Design principles as analysis tools
- 14-table relational wrangling
- Curriculum
- 17 concept groups, 4 tiers
- Data span
- 1950 — present
- Tables
- 14 relational CSVs
- Companion
- Written viz curriculum
01
Overview
Formula 1 generates beautifully relational data — drivers, constructors, circuits, lap times, pit stops, qualifying, standings — which makes it the perfect syllabus for visualization craft. This project pairs a working analysis notebook with a written curriculum (Explanation.md) that organises the field from first principles to advanced technique: basic charts through time series, distributions, correlation, multivariate and geospatial views, network graphs, and animation — closing with the design principles (color theory, layout, typography, storytelling) that separate a chart from an argument.
02
Problem
Most visualization learning is a grab bag of chart tutorials with no map of the territory. I wanted the map: a single dataset rich enough to motivate every chart family, worked through in one notebook, with a companion document that explains when each form earns its place — so the skill compounds instead of accumulating as disconnected recipes.
03
Goals
- Build a complete taxonomy of visualization forms, from bar charts to animated and 3D views
- Ground every technique in real questions about real F1 data — championships, lap pace, pit strategy
- Document design principles (color, layout, typography, narrative) as first-class material, not an afterthought
04
Solution
The notebook works through the Kaggle Formula 1 World Championship dataset — a relational set spanning 1950 to the present across circuits, races, drivers, constructors, lap times, and pit stops — applying each visualization family to a question it genuinely answers: line charts for championship arcs, box plots for lap-time spread, heatmaps for driver-circuit performance.
Explanation.md is the curriculum spine: four tiers (basic, intermediate, advanced, design principles) covering seventeen concept groups, each mapped to where it appears in the analysis. It reads as a course outline because it is one — written to be reusable on any dataset.
05
Research
The taxonomy synthesises standard visualization theory — chart-type selection, perceptual color considerations including color-blindness, and data-storytelling structure — into a personal reference validated against the F1 questions.
Tooling
Tech stack
- Python
- Jupyter
- pandas
- matplotlib
- seaborn
- plotly
Friction
Challenges
Relational data, flat charts
The F1 set is a dozen linked tables; every chart first needs the right join. Half the notebook work is shaping data — a deliberately visible half, because that is the honest ratio in real analysis.
Advanced does not mean better
Radar charts and 3D plots demo well and mislead easily. The curriculum treats them with explicit caveats about when a plain grouped bar chart is the stronger choice.
Takeaways
Lessons learned
- A taxonomy turns tool knowledge into judgment — knowing all the chart types matters less than knowing the question each one answers
- Design principles belong in the analysis phase, not the polish phase; color and layout choices change what readers conclude
- One rich dataset beats ten toy ones for building durable skill
Evidence
Dataset
Formula 1 World Championship (1950—present)
Kaggle
Relational tables covering circuits, races, drivers, constructors, lap times, pit stops, qualifying, and season standings across the championship era.
View datasetRoadmap
Future improvements
- Rebuild the strongest views as a race-weekend strategy dashboard on live timing data
- Add race-strategy analysis with live-timing data via FastF1
- Publish the curriculum as a standalone reference essay
Rebuilt for this site from the original repository — source history