The Illusion of ‘Form’: An Empirical Analysis of the Premier League
The fascinating phenomenon of the “Hot Hand Fallacy” continues to intrigue scholars and sports enthusiasts alike. Famously coined in a 1985 study by Gilovich, Vallone, and Tversky, this psychological bias convinces us to believe in the ‘form’ or ‘streak’ of an individual or a team in a game or a series of games. Is there such a thing as a winning or losing streak? Or is it merely a psychological trick we’re playing on ourselves? We explore this conundrum through a rigorous analysis of football — one of the most passionately followed sports in the world.
In this study, we scrutinize the performance of the English Premier League teams across the complete 2022/2023 season, comprising of 38 matches. The primary focus was to observe any apparent relationship between sequential positive results (i.e., wins or draws), sequential losses, and the subsequent outcomes. We also investigated the correlation between a team’s result on a given day and its performance in the subsequent match.
The following table represents a statistical analysis of football matches for the different teams:
Team: This column lists the football teams being analysed.
Total Positive Matches: The total number of matches where the team has had a positive outcome.
Total Positive %: The proportion of matches the team has won out of all matches played.
Positive after 2 positives: The percentage of times the team won their next game after winning two games in a row. The number in brackets represents the number of times the team had two consecutive wins.
Positive after 3 positives: The percentage of times the team won their next game after winning three games in a row. The number in brackets indicates the number of times the team had three consecutive wins.
Positive after 2 negatives: The percentage of times the team won their next game after losing two games in a row. The number in brackets shows the number of times the team had two consecutive losses.
Positive after 3 negatives: The percentage of times the team won their next game after losing three games in a row. The number in brackets reveals the number of times the team had three consecutive losses.
Correlation: This column represents the correlation between a team’s past outcomes (wins or losses) and the outcome of the next game. A positive correlation would suggest a streaky performance, where wins follow wins and losses follow losses. In contrast, a negative correlation suggests a pattern where wins often follow losses and vice versa. A value near zero implies that previous match outcomes bear little to no impact on the next match outcome.
Indeed, as we delve deeper into the nuances of our findings, it becomes apparent that the relationship between a team’s current performance and its subsequent results is far from straightforward. The lack of consistency in the correlation between a team’s performance on a particular matchday and the next further illuminates the notion that each match should be treated as an independent event.
Interestingly, the teams with higher overall positive results, such as Manchester City and Arsenal, demonstrated a remarkably high percentage of maintaining two and even three consecutive wins. Manchester City and Arsenal, boasting 87% and 84% total positive results, respectively, showed 82% and 85% positive outcomes after two consecutive wins. This percentage remained impressively high even after three consecutive positive results, with Manchester City registering 83% and Arsenal 86% positive results. However, when examining the correlation between match results, we find the figures surprising. A negative correlation of -0.14 for Manchester City and a negligible positive correlation of 0.01 for Arsenal indicate that the outcome of a particular match has little bearing on the result of the subsequent one. This observation is intriguing, as one might expect teams with high overall victories to display a stronger positive correlation.
The enigma intensifies when we turn our attention to teams with lower total positive results. Consider Wolverhampton Wanderers (Wolves) and West Ham United. Wolves, with a total of 19 positive results (50% of the total matches), and West Ham, with 18 (47% of the total matches), demonstrated a significant percentage of consecutive losses. Yet, the correlation between the outcomes of their matches was negative (-0.41 and -0.08 respectively). This observation is noteworthy as it counters the common notion that a team ‘out of form’ or on a losing streak is likely to lose the next match. These teams have shown that previous losses do not necessarily predict a negative outcome in the following match.
Moreover, the correlation data underscores that there is no uniform pattern when examining individual matches across the teams. Some teams showed a positive correlation, some a negative one, while others had no correlation at all, irrespective of their total positive outcomes. For instance, Aston Villa, with a total positive percentage of 66%, demonstrated a positive correlation of 0.22, indicating some consistency between the outcomes of consecutive matches. In contrast, Brighton, with a higher total positive percentage of 68%, showed a negative correlation of -0.32, suggesting that a positive result in one match did not necessarily imply a positive result in the subsequent match.
Our analysis supports the ‘hot hand fallacy’, implying that the notion of ‘form’ in football might not hold significant predictive power. Some teams show a higher correlation between match outcomes, but it’s not a universal trend. Hence, a new match presents fresh probabilities, regardless of a team’s previous streaks.
The captivating narratives of ‘winning’ and ‘losing’ streaks enrich the drama of football. However, each match is an independent event influenced by its unique dynamics, reinforcing the unpredictability of the sport. As the famous saying goes, “Football is a game of two halves”, indicating anything can happen by the final whistle.
The ‘Hot Hand Fallacy,’ as mentioned in Gilovich, Vallone, and Tversky’s 1985 study, suggests our tendency to misinterpret random sequences and see patterns where none exist, often leading us to believe in the concept of ‘form.’ This fallacy is supported by our analysis of the Premier League data.
The following infographic will provide a visual representation of the data, highlighting each team’s matchday outcome in light blue for positive results and orange for losses. This will offer an engaging summary of the intricate data and its implications.
Our research challenges the practical value of ‘form’ in football as a reliable predictive tool. Each football match is primarily an independent event, and the outcome of one game appears to minimally influence the next one.
This understanding extends beyond football, impacting our interpretation of sequences of success or failure in various fields like stock market trends and academic performance. The tendency to find patterns can lead us to overestimate the existence and significance of ‘form’ or ‘streaks’, emphasizing the need for informed decision-making.
Every football match begins on a blank canvas. Past performances, while vital for evaluating a team’s abilities, might not predict future results reliably. Managers, coaches, and players often base their strategies and preparedness on previous games’ outcomes. However, recognizing this fallacy could encourage them to treat each game independently.
Football isn’t merely a game of chance. It involves strategic decisions, adjustments, and unexpected factors such as injuries and suspensions. These factors underline the dynamic nature of football, where lineups continually change due to both strategical choices and unpredictable circumstances.
SportGPT-1, a new generation predictive model, emerged in response to these intricacies. It bypasses the traditional notions of ‘form’, focusing instead on individual players’ performances likely to feature in the upcoming game. It evaluates each player’s capability based on their past performances, undeterred by the team’s previous ‘streaks’.
This approach provides a more nuanced, informed prediction system. It challenges conventional notions of ‘form’ and ‘streaks’, using individual player data for forecasting future outcomes, thereby avoiding the hot hand fallacy.
In conclusion, the accepted notion of ‘form’ in football, often deemed crucial in predicting match outcomes, might be more of an illusion. Our study, along with others, suggests the inherent randomness of the sport, proposing that each match is largely an independent event. It’s crucial to remember that each match is a fresh event, independent of the preceding ones, as we relish the thrill of football.