IPL Win Probability Predictor (2008-2024)

A Hybrid Machine Learning model that predicts the outcome of IPL matches in real-time. It uses a dual-architecture approach to handle both pre-match uncertainty and live-match mathematical pressure.

Features

• Hybrid Architecture:
• 1st Innings: Uses Random Forest Classifier to predict outcomes based on historical Team Strength, Venue Bias, and Toss Decision.
• 2nd Innings: Uses Logistic Regression to calculate Win Probability based on the "Required Run Rate" pressure equation.
• Dynamic Feature Engineering: Calculates real-time metrics like CRR, RRR, and Wickets Left ball-by-ball.
• Context Aware: Accounts for Venue history (e.g., Wankhede chasing bias) and Head-to-Head records.

Model Accuracy

| Scenario | Model Used | Accuracy | | Pre-Match / 1st Innings | Random Forest | ~55-60% (beats random chance by 5x) | | Chasing / 2nd Innings | Logistic Regression | ~70-80% (converges as match progresses) |

Tech Stack

• Language: Python 3.10
• Libraries: Pandas, Scikit-Learn, NumPy, Matplotlib
• Concepts: Pipeline Integration, Column Transformation, Probability Calibration