IPL Auction & Match Simulator
A regression-based player valuation model and Monte Carlo match simulation engine built on 5 seasons of IPL data.
Overview
IPL franchises invest hundreds of crores during player auctions — yet valuation often relies on reputation and intuition rather than data. This project built a unified framework combining 4 composite performance metrics, temporal weighting, regression-based auction price prediction, and 600+ iteration Monte Carlo match simulation to answer the fundamental question: are franchises paying fair prices for players?
Problem Statement
- —Player auction prices influenced by brand value, not just performance
- —No objective framework to identify undervalued or overvalued players
- —Match predictions based on intuition rather than statistical modelling
- —Franchises lack data-driven tools for strategic squad planning
Data & Analysis
Actual vs predicted auction price (₹ Cr)
Above line = undervalued (bargain) · Overvalued points in Red · Undervalued in Teal
Market efficiency: over vs undervalued players (₹ Cr)
Positive = Overvalued (Red) · Negative = Undervalued (Teal)
Simulated score distribution — 600 iterations: CSK vs MI
Calculated simulation densities
Win Probability
Simulated heads-up outcomes
Key Findings
Auction model performance
The regression model explained 68.5% of auction price variation using only performance metrics.
Cummins overvaluation
Pat Cummins was predicted at ~₹16Cr but sold for ₹24Cr — an ₹8Cr brand premium.
CSK win probability
600 simulations gave CSK a 56.2% win chance against MI with average scores of 207 vs 212 runs.
Undervalued players found
Faf du Plessis, Sam Curran, and Prabhsimran Singh were statistically underpriced by ₹2Cr+ each.