Fantasy Premier League 19/20, a review - Part 1, Basics

5 min readJul 28, 2020

So it’s a wrap … Premier League 19/20 has come to a close with Liverpool taking the honors this time around. Since there can only be so many players on the pitch, competitions for a lot of us supporters comes in the form of football simulation games like PES or FIFA or football management simulators like the Football Manager series or predicting the best performing XI of the week on Fantasy Premier League. In this article I’ll be reviewing the FPL 19/20 season, with some basic analysis techniques and visualizations like who scored the most points, who proved to be the best bargain buys and a lot more!

Data

The first and the most important step is ….. where to find the data. Anyone like me who is trying to look for sources of data for football matches like event or tracking data for matches knows the pain of looking for data for recent matches. Being a very competitive space all this data is not open to all and most of event data that is free to use online is of seasons gone by like Wyscout data for 17/18 season. Tracking data is even harder to find. (Note :- In case there are any other sources for more recent match data, do let me know!)

But but but…. FPL data is free for all and can be accessed using the following API. Since extracting data from here might be a little tedious for some you can get all this data from this Github repository. A huge shout out to the maintainers of this repository for scraping and compiling all that data. Also they provide FPL data for previous seasons too!

In this article we will make use of players_raw.csv for 19/20 season provided in the above mentioned Github repository.

So now that we have the data let’s…

Dive head first into the data

The file contains data for 667 players. Obviously all 667 players would not have played an equal part in the league. So I have removed all players who have played less than 15 full games(1450 minutes). Now, this does hurt players who come on as super subs or impressive January signings like Bruno Fernandes(who has been eliminated due to the constraint :P but has his own section at the end), but it reduced our dataset to 244 players. Now here’s our first plot

Cost v PPG plot. Annotations not provided in the bottom left cluster for the sake of clarity

Who proved to be the best signings?

So far so good. We’ve been able to generate a simple points per game plot from the data. Now from a macro view, how can we determine who have been the best signings in FPL this season. Do we plot a line via linear regression to determine who have performed well and over their cost? Well due to the skew in the data towards the bottom left of the graph, the line may not be representative of the best buys towards the top right. I found dividing the data into buckets and selecting a top percentile from that bucket to be a better approach to figure out the best signings.

We divide the players into 4 buckets:-

1. Premium Players (above 9M)

Pretty much the top, most expensive players of the league. Let’s look how they fared:-

Clearly De Bruyne is top of the list by some margin, followed by the two Liverpool wingers. At the bottom we can see Lacazette and Firmino, two strikers who clearly did not enjoy the best of seasons.

2. Top Players (7M–9M)

Now we look at the next few expensive players in the league

FPL which is known for being notoriously biased towards goal scorers finally has a defender up there, Trent Alexander Arnold (is he actually a defender though?), who enjoyed his most prolific campaign in the league in terms of assists. United strikers Martial and Rashford are up there too.

3. Mid Range Players (5.5M-7M)

Next up we have the mid range players, a lot of which are defenders or central midfielders, hmmm….

Finally some more defenders, with van Dijk(Liverpool), Doherty(Wolves) and Pereira(Leicester) forming the top 3. Teemu Pukki just manages to sneak in given his early season goal scoring exploits.