This project involves the development of an Excel-based prediction model that forecasts football match outcomes based on key team performance statistics such as possession, attacking (ATT) and defensive (DEF) ratings, and expected goals (xG). Using regression analysis and simulations, the model predicts match results between two teams and simulates thousands of matches to provide probability distributions of outcomes.
Project Overview
The model uses team performance data (matches played, wins, draws, losses, goals for, goals against, goal difference, points, possession percentage) to calculate both ATT and DEF ratings. By incorporating regression analysis, the model predicts each team’s expected goal (xG) based on possession, attack, and defense correlations. The final output includes a prediction of the match outcome and a series of simulations to illustrate potential results.
Key Features
Team Performance Analysis:
The dataset includes key statistics for 20 teams, including Liverpool, Manchester City, Manchester United, and others.
Attack and Defense Ratings are calculated using a regression formula with possession as a key variable:
Attack (ATT) Rating Correlation:
Intercept: -0.39, Possession Coefficient: 0.027.
Defense (DEF) Rating Correlation:
Intercept: 1.78, Possession Coefficient: -0.015.
Expected Goals (xG) Calculation:
The model predicts each team’s expected goals (xG) based on their ATT and DEF ratings. For example:
Manchester City: xG = 2.09
Manchester United: xG = 1.27
Match Outcome Prediction:
Using the calculated xG values, the model predicts the probabilities of various match outcomes, such as:
Manchester City Win: 56.47%
Draw: 20.20%
Manchester United Win: 23.33%
Simulations:
The model runs 10,000 simulations of the match to predict possible outcomes. For example, over 12 trial runs:
Manchester City wins 6 times.
Manchester United wins 5 times.
2 matches result in a draw.
This allows users to gauge the variability of results and understand the likelihood of different match outcomes.
Key Statistical Outputs
ATT and DEF Correlations:
The ATT rating is positively correlated with possession, indicating teams with higher possession tend to have better attacking performances.
The DEF rating is negatively correlated with possession, showing that teams with higher possession also tend to concede fewer goals.
Regression Summary:
ATT Model: R² = 0.69, showing that possession explains 69% of the variation in attacking performance.
DEF Model: R² = 0.38, showing that possession explains 38% of the variation in defensive performance.
Challenges and Solutions
Data Interpretation: Handling the complexities of multiple variables (e.g., possession and defense) and understanding their influence on match outcomes was crucial for accurate predictions. The regression models ensured the data was processed correctly, making predictions more reliable.
Simulation Accuracy: Running thousands of simulations helped account for variability in match outcomes, providing a more comprehensive understanding of the probabilities of each result.
Final Deliverables
Match Prediction Sheet: Provides the expected win, draw, and loss probabilities for each team, based on possession, ATT, and DEF ratings.
Simulation Sheet: Includes the results of 10,000 simulated matches, showing win-loss patterns and the most likely outcomes.
Statistical Summary: Detailed regression outputs and residuals for ATT and DEF predictions, ensuring the model’s reliability.
Conclusion
This Excel-based prediction model offers a data-driven approach to football match forecasting. By utilizing team statistics and regression analysis, it predicts match outcomes and simulates results for thousands of games, giving users a clearer picture of probable outcomes. The tool is ideal for football analysts, bettors, or enthusiasts looking to make informed predictions based on solid statistical foundations.
Excel File Link: Download Here
