GR Cup Performance Intelligence Platform

Slash lap times with AI trained on 4,343 corners. We pinpoint exact time loss and deliver real-time, physics-based coaching to perfect your line instantly.

Comment

Inspiration

Motorsports drivers know they're slow—but not where or why. Traditional analysis requires coaches to manually review hours of telemetry. We asked: what if AI could do this instantly, identifying not just WHERE drivers lose time, but WHY and HOW to fix it?

What it does

Our platform analyzes championship-scale telemetry data (4,343 corners from 64 drivers across 7 tracks) and provides:

  • AI-Powered Coaching: Top 3 improvement opportunities per driver, prioritized by lap time impact
  • Physics-Based Insights: Explains the root cause (over-braking, late throttle, etc.) and provides actionable advice
  • Predictive Modeling: Random Forest model (R² = 0.899) predicts lap time improvements from technique changes
  • Cross-Track Analysis: Identifies driver strengths/weaknesses across different circuits
  • Multi-Driver Comparison: Teams can compare multiple drivers to optimize lineup and training

How we built it

  • Data Processing: Processed 100+ million telemetry rows using Pandas
  • Feature Engineering: Extracted 8 physics-based features per corner (braking zones, throttle application, steering smoothness, lateral G-forces)
  • Machine Learning: Trained Random Forest and Gradient Boosting models to identify what drives lap time
  • Driver Clustering: K-means clustering identified 4 distinct driving styles
  • Dashboard: Built with Streamlit for real-time interactive analysis
  • Visualizations: Plotly for professional, publication-quality charts

Challenges we ran into

  • Data Inconsistency: Different tracks had different telemetry formats (column names, time units, file structures)
  • Corner Detection: Automating corner identification across varied track layouts required adaptive thresholding
  • Scale: Processing 4,343 corners efficiently while maintaining interactive performance
  • Dynamic Comparisons: Pre-computing all driver pairs wasn't feasible—had to build on-the-fly comparison engine

Accomplishments that we're proud of

  • 89.9% model accuracy predicting lap times from corner features
  • Discovered NEW insight: Steering smoothness (34% importance) matters MORE than braking power (21%)—contradicts traditional coaching wisdom
  • Championship-scale: Successfully processed ALL 7 tracks, not just a sample
  • Production-ready UI: Glassmorphism design, smooth animations, fully interactive

What we learned

  • Corner exit throttle management is 3x more important than we expected
  • Driver style clustering revealed clear patterns: some drivers are naturally smooth, others aggressive—coaching should adapt
  • The gap between elite and mid-pack drivers is smaller than expected (1-2 seconds), making coaching ROI enormous

What's next for GR Cup Performance Intelligence Platform

  • Live telemetry integration: Real-time coaching during practice sessions
  • Video overlay: Sync telemetry insights with onboard camera footage
  • Automated reporting: Generate PDF coaching reports post-race
  • Predictive maintenance: Use telemetry patterns to predict mechanical issues
  • Expand to other series: F1, NASCAR, IndyCar, IMSA

Built With

Share this project:

Updates