FIFA Player Value Prediction
Comparing four ML approaches for predicting player market value from skill attributes.
A comparison study: how well can a player’s market value be predicted from their recorded skill attributes, and does model complexity actually buy accuracy on a problem like this one?
Four approaches were evaluated against the same dataset — linear regression as a baseline, random forest, a multilayer perceptron, and a CNN. The framing question is whether the added capacity of the neural approaches pays off on structured tabular attributes, or whether a well-specified tree ensemble is the more honest answer for this shape of data.
This page is a summary of the write-up rather than a runnable notebook.
Player value is the thing being predicted from recorded skill attributes. Photo via Boston Herald.