Frequently asked questions

AFL questions

What is Australian football or the Australian Football League?

My friend, let me introduce you to by far the best sport on the entire planet.

Model questions

Why is this called the “Footy Forest”?

Many of statistical techniques used in this model are “tree-based”, so I thought I’d take the tree concept and run with it like Ed Langdon on a fast break. Also it predicts … footy.

What variables does the Footy Forest use?

It uses four variables. These are:

  1. each team’s form, based on an Elo formula.

  2. venue experience

  3. player ratings for each selected team, and

  4. the distance each team has had to travel.

Each of these are calculated for each team and then converted as a relative measure showing the advantage for the home team compared to its opponent on the day.

How can I check the performance of this model?

I update the performance metrics on the home page every week. I’ve also entered the model in the official AFL footy tipping competition, so you can check the performance by searching there.

Betting and tipping

Are you doing this to make money?

Hell, no. I am doing this for fun and for a challenge. Yes, I am a nerd. And yes, I am obsessed by Australian football.

Should I use this model to place bets?

Definitely not. Betting is for mugs and niff-naffs.

Should I use this model for footy tipping?

Absolutely. Go bananas.

Technical questions

What program did you use to write this model?

I use R for this project, which is what I use for a lot of my academic work. The website you’re looking at now is also produced with R with RStudio and Quarto.

Does this model use ChatGPT, or a similar large language model?

No. This model is one hundred per cent pure, old fashioned, home-grown human-written computer code, born free right here in the real world.

Isn’t that a line from The Matrix?

Yes. Yes it is.

What techniques are used for this model?

This model is an ensemble of models compiled with the R Tidymodels package. It uses a stacked combination of gradient boosted trees, generalised linear models and random forest models to make its predictions.

Where do you get your data from?

I use the FitzRoy package for R. It is wonderful. Cheers to the legends to have developed it and maintain it.

Personal questions

What football team do you barrack for?

I barrack for the original Australian football team – the Melbourne Demons.

Who is your favourite footballer?

As a kid my favourite player was Stephen (Stinga) Tingay. But right now it is probably young Caleb Windsor. I guess I have a thing for wingers.

What is your background?

I am a quantitative social scientist who is mainly interested in researching empirical problems in science communication. While this gives me some useful skills in statistical analysis, and I have helped to publish some academic work using machine learning techniques, there are many more qualified people than me who have produced similar models. Prior to this I worked in various communication roles.

Are you a real doctor?

I am not a medical doctor. I have a PhD in science from ANU.

How frequently are you actually asked these questions?

Never. I’ve never been asked any of these questions.