FPL project overview and report. Download PDF for best quality.
Fantasy sports are online games where participants create virtual teams of real-life sports players. These players earn points based on the actual performance of their real-world counterparts. Fantasy sports players compete against each other in private and public leagues, and the player with the highest overall score at the end of an annual season wins.
Like any decisions made by humans, the process of selecting players for your Fantasy Premier League (FPL) team is susceptible to various inherent biases.
One of the most prevalent biases is team loyalty, wherein players often exhibit a preference for players from their favourite Premier League club. This loyalty, whether consciously or subconsciously, leads to a certain level of favouritism towards players from their team of choice, as well as a reluctance to include players from rival teams. As an Arsenal fan, for instance, it is often the case for my team to consist of multiple Arsenal players, even if they may not be the most optimal choices, while Tottenham players (Arsenal's biggest rival) are rarely considered. This trend seems to emerge every year, albeit unintentionally.
Recency bias also plays a role in influencing team selections. If one has recently witnessed a team or player perform exceptionally well or poorly, there is a tendency to overemphasize these recent performances when making decisions for the upcoming weeks selections.
Another powerful form of bias, which can be both helpful and misleading, is the players selected percentage metric. This metric, displayed alongside each player during the FPL team selection process, indicates the percentage of FPL players who have included that player in their teams. While high selected percentages often indicate players who have been performing well or offer good value, it is not always a reliable indicator. Nonetheless, many FPL players tend to select certain players simply because others have chosen them, without considering other relevant factors.
The objective of this project is to mitigate these biases in all their forms. By developing algorithms and selection processes solely based on historical data, my goal is to eliminate human bias and stick to purely data-based decision-making in FPL team selection.
Web Scraper: FBREF Football Player Data Scraper
Step 1: Wrangling data
Step 2: Predicting team goals
Step 3: Predicting clean sheets
Step 4: Predicting player points
Step 5: Selecting teams
Licensed under MIT.
FPL data: https://github.com/vaastav/Fantasy-Premier-League
The data used in this project is property of https://fbref.com and http://clubelo.com/ I don't own any of the data