The 2025 AFL season is just around the corner and fans are pondering the big questions: who will play finals? Who will finish in the top four? Who's getting the wooden spoon ?
Authors
- Tara Lind
PhD Candidate, La Trobe University
- David Carey
Senior Lecturer in Sport Analytics and Data Science, La Trobe University
The start of a new season brings with it many unknowns, hopes, and in some cases, trepidation.
Hawthorn finished 2024 playing some of the most exciting footy in the competition - can they keep that momentum going?
Collingwood enters 2025 with the oldest and most experienced list - will that be the key to another deep finals run? Or are they over the hill?
Can Carlton finally break its premiership drought? Can West Coast, North Melbourne, or Richmond get back on track? What can Fremantle do with its young list and high expectations?
With so many unknowns, we turned to data.
Simulations and predictions
In La Trobe University's Master of Sport Analytics , students need to build their own footy tipping algorithms and use them to simulate future matches.
We've seen lots of different approaches to this problem. Each comes with its own set of assumptions and blind spots.
One straightforward way to try to forecast what will happen in the upcoming season is to just look at history: how often does a team that finishes first on the ladder stay on top the next?
That's happened seven times since 1990, so about 20% of the time.
We can model probabilities like this for every ladder position to get a gauge on how rankings typically shift from season to season, and apply this to the end-of-season 2024 ladder to predict the 2025 standings.
This approach does not take into account last year's finals results, the different age profiles of teams, the 2025 fixture, or other team changes such as trades, retirements, or injuries.
Taking age into account
How about if we consider player ages as well? This should give us a better sense of a team's expected change between seasons.
Research has suggested AFL players reach their peak performance levels at around 24-25.
A quick look at team median ages since 1990 agrees: teams with a median player age over 25 typically have a worse winning percentage the following year, and teams younger than 24 usually improve (with plenty of exceptions).
Combining last year's ladder with age profiles gives a different view of the upcoming season.
There is more shuffling, with older teams like Collingwood and Melbourne expected to fall, while the younger Fremantle, Gold Coast and Adelaide lists are given higher probabilities of finishing near the top.
We're still left with some important blind spots though: information from last year's finals (Brisbane performed far better than a typical fifth-place finisher), and the difficulty of the upcoming fixture, have not been considered.
The Elo rating system
To take the full 2025 fixture into account, we need to simulate the entire season game by game.
That can be done if we use the Elo rating system to get a "strength" rating for each team.
Elo ratings track team strength over time: ratings go up with a win and down with a loss. The amount it changes depends on the opponent - beating a strong team boosts the rating more than beating a weak one, and the ratings update after every game played.
We'll use the Elo ratings that each team ended up on at the end of last year (including finals) as a baseline for 2025.
With these ratings, we can calculate the probability of one team beating another in any given matchup. The method also considers home ground advantage by giving the home team a small rating boost.
Once we have probabilities for each match outcome, we can simulate the entire season. Here's how it works:
- Each game needs a winner. To decide, we use a computer function that picks a winner based on probability, kind of like flipping a weighted coin. If a team has a 70% probability of winning, it's more likely to be chosen, but there's still a 30% chance they lose
- This is done for every game in the season
- We then repeat this 10,000 times - simulating 10,000 different versions of the season
- In each version, we create an end-of-season ladder, based on the simulated games results
- After all the simulations, we can see how often each team finishes in each ladder position. This gives us a prediction for their chances of finishing first, second, third and so on.
The Elo approach favours Brisbane much more and is less kind to West Coast (35% chance of finishing last).
It does not predict the decline of Collingwood and Melbourne because, although it takes into account the finals and fixture, it doesn't have an age component.
The 'wisdom of the crowd'
If each approach comes with its own set of limitations, then we might expect to get a better forecast by combining lots of predictions from different sources because of the " wisdom of the crowd ".
The idea is that you get more accurate predictions if you combine multiple independent sources.
Luckily for us, each season, several AFL stats experts build models to estimate the probability of each match outcome and generously post them online .
What goes into each model is not always known, but they consider a mixture of different factors such as attacking and defending strengths, in-game statistics, home ground advantage, player lists and trades, last season's performance and more.
For our analysis, we'll combine the Elo model with the average of all these expert tips to get a "wisdom of the crowd" prediction for each game's probability. The ladder can then be simulated using the same method as above.
Four groups emerge from the wisdom of the crowd:
- Brisbane, Hawthorn, Geelong and the Western Bulldogs are predicted to lead the pack, surpassing last year's top three
- Sydney, Port Adelaide, GWS, Carlton, Fremantle, Collingwood and Adelaide have a wide spread of predicted finishes, skewed more towards finishing in the top eight - but there won't be enough room for all of them
- Essendon, Melbourne, St Kilda and Gold Coast might challenge for a spot in the finals, but the models are less confident in their chances
- West Coast, North Melbourne and Richmond are hard to separate from each other, a cut below the rest.
Uncertainty and excitement
Each table tells a potentially different story but the most universal theme is uncertainty.
Team sports are hard to predict, especially before we've had a chance to observe any games, and even the most confident predictions are under 40% (meaning they are more likely not to happen).
Uncertainty leads to excitement, and this data only makes us more excited to see what will play out this season.
The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.
