Efemkay FPL Analytics wiki!

This is my playground to learn about XGBOOST and get my feet wet in machine learning. With the hype surrounding Fantasy Premier League, it provides a good motivation. Also, FPL captures a lot of data from each game and made it available just as frequent through public domain (read Getting FPL Data for more details).

Predictive Model

Here I will attempt to predict the best starting 11 and 4 reserves player to be selected at each Game Week. Overall idea is to utilise classification model to filter out bottom quantile (either 50th or 66th) and proceed with regression model to choose among the rest.

I've managed to train a XGBOOST model for the classification with an acceptable performance (~20% error rate). You can learn what I've done by perusing my R notebook in the link below.

• Top 1/3 Classification Model

Weekly Analysis

• Week 01
• Week 02
• Week 03