This project involved the creation of an Excel-based prediction model designed to forecast the outcomes of football matches. The model uses team performance statistics and advanced regression analysis to predict the probabilities of a home win, away win, or draw. Additionally, Poisson distribution was utilized to model goal probabilities and generate expected match results. This tool provides valuable insights for match forecasting, betting strategies, and sports analysis.
Project Overview
The Excel-based model leverages team stats such as matches played (MP), wins (W), draws (D), losses (L), goals for (GF), goals against (GA), and points (Pts), along with ATT and DEF ratings. Multiple regression analysis was conducted to examine the correlation between possession and attacking/defensive performance. The model then applies Poisson distribution to predict match outcomes and provides probabilities for home win, away win, and draw.
Key Features
Team Performance Stats Analysis:
Data from team performance was used to calculate attack (ATT) and defense (DEF) ratings.
Regression Analysis: Statistical models were built to measure the correlation between possession and both ATT and DEF ratings.
ATT Correlation:
Multiple R: 0.83, R²: 0.69 (indicating a strong positive correlation between possession and ATT rating).
DEF Correlation:
Multiple R: 0.61, R²: 0.38 (showing a moderate inverse relationship between possession and DEF rating).
The output includes regression statistics, ANOVA tables, and residuals for each observation, ensuring that all necessary statistical checks are included in the model.
Poisson Distribution for Goal Predictions:
A Poisson distribution model was applied to predict the probability of each team scoring a specific number of goals in a match.
The matrix of goal probabilities allowed for precise forecasting of match outcomes between two teams.
Example:
Probability of Manchester City scoring 2 goals: 26.94%.
Probability of Manchester United scoring 1 goal: 36.33%.
Based on this, the model estimated the likelihood of a home win, away win, or draw.
Prediction of Match Outcomes:
Home Win, Away Win, Draw Probabilities: The model calculates the percentage chances of each match outcome. For example, it might predict Manchester City’s home win probability at 21.30%, an away win for Manchester United at 5.69%, and the probability of a draw at another percentage.
Betting Odds and Probability Comparison: Using these probabilities, the model compares calculated odds with actual betting odds to identify potential value in bets.
Comprehensive Visual Analysis:
The model is equipped with visual representations such as charts and tables that display the goal probabilities for each team, making it easy to interpret the results of each match simulation.
Challenges and Solutions
Statistical Accuracy: Ensuring the accuracy of predictions required meticulous data cleaning and multiple iterations of regression analysis. The model's success hinged on properly understanding the relationships between team performance metrics and match outcomes.
Goal Distribution Modeling: Poisson distribution was chosen due to its suitability for modeling discrete outcomes like goal counts in football. This added complexity but significantly improved the prediction accuracy for match results.
Final Deliverables
Prediction Model Spreadsheet: The final Excel model includes regression analysis output, residuals, Poisson distribution predictions, and detailed statistical tables for easy reference.
Match Outcome Predictions: For each fixture, the model provides percentage probabilities for a home win, away win, and draw, along with visualizations of goal distributions.
Statistical Summary: The model outputs include multiple R, R², adjusted R² values, and p-values to validate the significance of the predictors.
Conclusion
This Excel-based football prediction model is a powerful tool for forecasting match results, leveraging team statistics and advanced statistical methods such as regression analysis and Poisson distribution. The model is suitable for football analysts, bettors, and enthusiasts looking to predict match outcomes with a data-driven approach.
Excel File Link: Download Here
