The Fantasy Premier League, according to the Four Four Two website, is an online game in which players collect points based on how real life footballers perform each week. Simply put: you choose a virtual team of Premier League footballers, and if their real-life counterparts do well, you get points.
I've been playing for a few years now and my selections were based on gut feeling and my domain knowledge of football. In a league with friends, I've done fairly well - never won, but fairly well. Since the FPL has their data publicily available, I decided to try a data based approach. The data is available here. Did I try using last year's data to predict a team for the first match day? No, I didnt. I tried using my guts again and ended up with this team (fig. 1)
The players I selected for this season did quite well last season, and even with my mishaps Shaw, Salah, Bamford, Saint Maximin, Gundogan and Lloris are doing well. However, if I had paid attention to the data, I would have known not to put anyone from Arsenal after Game week 1 (no attack on Arsenal fans, this is backed by data).
Looking at a box plot of total points by players on a team (fig. 2)
Manchester City, Liverpool and Spurs seem to be doing quite well. Spurs being the surprising one but not unexpected because they tend to do well at the start of the season (look at Arsenal). I wouldn't have added Saka even though he did well last season and has been great on the England team.
The FPL has some metrics it created, ICT, which stands for Influence a player has, the Creativity of the player and the Threat the player player poses. By players on a team, this is how it looks (fig. 3)
Although West ham isn't on average one of the teams with the best players based on the ICT criteria, they seem to have an outlier! Looking at the players' totals points per game over the past 3 game weeks (fig. 4),
And Anthonio from West Ham looks to be out performing everyone. Salah seems to have contributed the most points to my team and being my captain, that number is doubled. Lloris is unfortunately on my bench meaning I've missed out on all those points.
A deeper dives to see goals(fig. 5) and assists(fig. 6) per player shows
Anthonio is high on goals and assists. To be fair, they've played Newcastle, Leicester City and Crystal Palace who aren't the strongest of teams so it gives them a good start.
With this data, I'll update my team and see how that goes.
I've not done any predictive analytics yet, just looking at the data and finding the blind spots my guts missed out on. There are more features available and by combining them, I can do some predictive analytics. Data over guts!
Till next game day, May the Force and football gods be with me.