Football Analytics
Why the most common arguments against the stop-clock are nonsense
In which I refute the most common arguments given against the introduction of a stop-clock in professional football.
Frenkie de Jong on song
An analysis on rising star Frenkie de Jong, that was originally posted on the OptaPro blog.
Improving my ‘xG added’ model
A write-up of a research paper I did for university, mostly around an improvement to the xG added model I introduced earlier.
How accurate are xG models II: the ‘Big Chance’ Dilemma
A test of how good our xG models are and reflections on the usage of the 'Big Chance' variable.
What is the best location to pass from?
Some positions yield higher xG added values on average than others. How does this influence the xG added score for players?
Measuring dribbling skill
Improving the ‘xG added’ model by including dribbles (not only takeons) into the ball progression types. Who is the best dribbler in the Premier League?
What is a possession-based model (and why does it matter)?
Improving the ‘xG added’ model to a possession-based model, and explanation of what this changes and why it’s better.
These are the best passers at EURO 2016
Using the recently created metric ‘xG added’ to try and find the best passers at EURO 2016.
Measuring passing skill
Introducing a new metric called ‘xG added’, which tries to quantify the value a player adds to a team through his passing.
Biases in our xG models
Third and final blog in a series on evaluating xG models. In this part I focus on testing whether existing xG models have certain biases.
How good are our xG models?
Second in a series of three on evaluating xG models. In this part I introduce my way of evaluating a model and apply it to models currently out there.
How NOT to evaluate your xG model
First in a series of three on evaluating xG models. In this part I focus on why the most used method of evaluating an xG model is wrong.
Introducing my Expected Goals model
My first blog on football analytics, in which I introduce my own Expected Goals model, its methodology and shortly evaluate its performance.