Predicting Match Outcomes in the Indian Super League: An Application of the Goal Supremacy Model
Abstract
This study expands upon the methodologies established in "Rating Systems for Fixed Odds Football Match Prediction" (2003), by applying the goal supremacy model to the Indian Super League (ISL). The research adapts this established model to predict football match outcomes, with a specific focus on the unique dynamics of the ISL. Analysing data from 377 matches across seven seasons, the methodology involves calculating team ratings based on the goal difference from their last five matches. Early-season games were excluded to ensure a consistent performance metric. The study employs regression analysis, revealing a significant relationship between team ratings and match outcomes, particularly in predicting home wins, away wins, and draws. The results indicate the model's robust predictive ability for home wins, as shown by an R-squared value of 0.730, and display significant correlations for losses and draws. This study validates the effectiveness of the goal supremacy model in the ISL context, offering valuable insights for sports analytics and demonstrating the adaptability of predictive models across different football league environments.
Introduction
The concept of a rating system, as described in "Rating Systems for Fixed Odds Football Match Prediction" (2003), offers a quantitative measure of a football team's superiority over its opposition. This superiority is assessed by analysing various aspects of past performance, including league points, positions, and notably, goals scored and conceded.
The 'goal supremacy model' is a predictive system that calculates a team's rating based on their goal difference - goals scored minus goals conceded - over a set number of recent matches, here the last 5 matches are chosen. This model operates on the assumption that teams demonstrating greater goal dominance are more likely to succeed in subsequent matches. It is a form of recent form goal-difference rating system, which simplifies the complex dynamics of football into a quantifiable measure of superiority, without accounting for the quality of the opposition faced in these matches.
The fundamental assumption of the goal supremacy model is that goal difference is a reliable indicator of a team's current form and potential to win. It hypothesises that a higher goal difference, indicating more goals scored and fewer conceded, correlates with an increased likelihood of winning the next game. This model simplifies the multifaceted nature of football performance into a singular, quantifiable metric, offering a streamlined yet effective approach to predicting match outcomes.
Applying predictive models like the goal supremacy model in football analytics is crucial for developing effective team strategies and informing betting markets. However, the model's effectiveness can vary significantly across different football leagues and contexts.
This study seeks to adapt the goal supremacy model for the Indian Super League (ISL), a league with distinct characteristics, to evaluate its accuracy in predicting match outcomes in a non-European football environment.
The paper will commence with an examination of the theoretical underpinnings of the goal supremacy model and its conventional applications. Following this, the adapted methodology for the ISL will be detailed, the results of applying the model to the league will be presented, and the findings will be discussed considering the results in the Indian Super League.
Methodology
The objective of this study was to apply the goal supremacy model to predict match outcomes in the Indian Super League (ISL). Data was gathered from fbref.com, encompassing a total of 541 matches across seven seasons, starting from the league's inception in 2014 until the 2022-23 season.
Due to the unique conditions imposed by the COVID-19 pandemic, the 2020-21 and 2021-22 seasons were excluded from the analysis. This decision was informed by the fact that matches during these seasons were played in a bio-bubble at a centralized venue in Goa, which removed the typical home and away team dynamics, a key consideration in the goal supremacy model.
The methodology involved a careful selection of matches. To align with the requirements of the goal supremacy model for recent form analysis, the initial matches of each season were omitted. This exclusion was critical as there would not be previous five matches at the start of the season to establish recent form, a foundational aspect of the model. As a result, 164 matches were excluded, narrowing the focus to 377 matches where sufficient data for recent form was available.
In applying the goal supremacy model, the study calculated each team's rating based on the goal difference from their last five matches. The rating for a specific match was determined by subtracting the away team's goal difference from the home team's goal difference.
Following the calculation of team ratings, the matches were grouped accordingly. For each rating category, the occurrences of home wins, losses, and draws were meticulously tallied, along with their respective percentages. This organization of data was crucial to analyse how the home team's rating influenced the outcomes of the matches. To ensure the data was robust and representative, ratings based on fewer than three matches were not included in the analysis.
Regression analysis was conducted using SPSS to explore the relationship between the home team's rating and the probabilities of various match outcomes.
Results
The application of the goal supremacy model to the Indian Super League (ISL) yielded significant findings. The analysis, based on 377 matches from various seasons, focused on understanding the relationship between the home team's rating and the probabilities of different match outcomes (home wins, losses, and draws).
The quadratic regression analysis revealed distinct patterns:
Home Win Percentage:
The relationship between the home team's rating and the home win percentage was best described by a quadratic equation. The equation, y = 0.098x² + 1.935x + 41.436, demonstrated a significant correlation with an R-squared value of 0.730 and a p < 0.001. This high R-squared value indicated that approximately 73% of the variance in home win percentage could be explained by the team's rating.
Away win Percentage:
Similarly, the relationship between the home team's rating and away win percentage was captured by a quadratic equation, y = 0.027x² - 1.483x + 29.301. This model also showed statistical significance, with an R-squared value of 0.537 and p < 0.001, suggesting that about 53.7% of the variability in away win percentage was accounted for by the home team's rating.
Draw Percentage:
The analysis for draw percentage also followed a quadratic relationship, defined by the equation y = -0.125x² - 0.452x + 29.263. This relationship had an R-squared value of 0.417 and p = 0.003, indicating a moderate level of predictability for draw outcomes based on the team's rating.
Discussion
The application of the goal supremacy model to the Indian Super League (ISL) offers a detailed perspective on how team ratings influence match outcomes, with each outcome exhibiting a unique relationship characterized by its curve shape, statistical significance, and R-squared values.
Home Win Percentage and Rating Curve: The quadratic equation for home wins (y = 0.098x² + 1.935x + 41.436) is statistically significant (p < 0.001) with a high R-squared value of 0.730. The upward-opening parabola indicates that as the rating increases, the probability of a home win also increases, but at a diminishing rate. At a rating of 0, the model predicts a base home win probability of approximately 41.4%, reflecting the intrinsic home advantage. The positive coefficients suggest a steep increase in win probability with rating improvements.
Away Win Percentage and Rating Curve: The away win percentage is modeled by y = 0.027x² - 1.483x + 29.301, with a significant negative linear coefficient and an R-squared value of 0.537 (p < 0.001). This upward-opening parabola shows that as the rating improves, the likelihood of an away win decreases, but this decrease slows down at higher ratings. At a rating of 0, the model predicts an away win probability of around 29.3%, indicating an inverse relationship between home team rating and away wins.
Draw Percentage and Rating Curve: The draw outcome analysis (y = -0.125x² - 0.452x + 29.263), with an R-squared value of 0.417 and a significance level of 0.003, forms a downward-opening parabola. This shape reveals that the highest probability of a draw occurs when the rating is near 0, suggesting evenly matched teams. As the rating moves away from 0, the likelihood of a draw decreases on both sides, indicating that a larger disparity in team strengths reduces the chances of a draw.
The examination of these curves alongside their statistical significance and R-squared values provides a comprehensive understanding of the relationships. The curves illustrate not only the direct impact of ratings on match outcomes but also the subtleties in these relationships. Particularly, the analysis of draw outcomes at different ratings highlights the complexity of predicting these outcomes and the importance of closely matched team strengths.
Conclusion
This study ventured into assessing the applicability of the goal supremacy model within the Indian Super League (ISL), aiming to evaluate its effectiveness in the ISL. The investigation primarily focused on understanding the relationship between team ratings, derived from recent goal differences, and the probabilities of various match outcomes.
The research uncovered a notable correlation between team ratings and the distribution of match outcomes. It was observed that as the home team's rating increases, reflecting a stronger recent form, the likelihood of securing a home win also rises substantially. Conversely, the chances of away wins diminish with higher ratings for the home team. A critical insight emerged regarding draws; these were most probable when teams were evenly matched, as indicated by ratings near zero, and became less likely as the disparity in team strengths increased.
However, there exists a potential limitation of the goal supremacy model: its primary focus on recent team form without accounting for the differentiated dynamics of home and away performances. Teams often exhibit varying levels of performance depending on the venue, some teams play very well at home but significantly drop of away from home or vice versa, a factor that the current model does not capture. This limitation points to an area ripe for improvement in the model.
The next enhancement of the model, therefore, could involve incorporating elements that reflect a team's specific home and away performance trends. Such an advancement could lead to a more refined and accurate predictive analysis, offering a comprehensive understanding of a team's capabilities across different playing environments.
In conclusion, the application of the goal supremacy model to the ISL not only demonstrates its adaptability to different football league environments but also brings to light its current limitations and the potential for further refinement.
References
Football-Data. (2003). Rating Systems for Fixed Odds Football Match Prediction. Available at: https://www.football-data.co.uk/ratings.pdf (Accessed: 2nd January 2024).
Fbref - Indian Super League Seasons (2014 - 2023). Available at: https://fbref.com/en/comps/82/history/Indian-Super-League-Seasons (Accessed: 2nd January 2024).
