Explaining the Stats Insider AFL Player Ratings
Stats Insider’s AFL Player Ratings model is an attempt to rate and rank the competition’s players based on their performance and output over the last two years of home and away and Finals football.
"Even after the Moneyball revolution, baseball didn’t get rid of scouts."
- Darryl Woodford, Hypometer Technologies co-founder
The idea was borne out of our ‘wisdom of the crowds’ approach towards modelling the 2019 Brownlow Medal. During that process we noticed that canvassing media votes, coaches votes, and match reports gave our models a much stronger idea of players who played important roles in each individual game. However, these aren’t the only inputs; we also use a custom, fantasy-like scoring system to split tiebreakers and sort the players who have received a mention elsewhere. Let’s take a look:
Component 1: Game Rank
After the data from each game has been collected, a score out of 100 is given to each player. To score a full 100, the player would need to be unanimously considered the best player on the field. From here, each player is given an in-game rank from 1st to 44th. A rolling average of the players’ game rank is built into the final model, disregarding a player’s lowest 20% of performances to account for unknown factors such in-game injuries, as well as to give players in less consistent positions some room to move.
Component 2: Award Tour
This part of the model I’ve named after the ‘A Tribe Called Quest’ song of the same name.
It solely looks at a player’s total score from every game over the last 40 matches of AFL. The key difference here is that it ranks each player against their current position, and turns it in to a league-wide percentile (from 0 to 99.9). This is built to stop midfielders filling all the top rankings like they have run away with the Brownlow Medal over history. There is no difference between being a Harris Andrews or Nat Fyfe on this model if they’re both atop their positions in total score over that period.
Component 3: Star Power
In keeping with the theme of scoring out of 100, the last part of the model is also a positional percentile. Taking a longer look back, we find the average of a player’s best 20 games over the last 40 AFL games, and rank them against the other players in that position.
“Every [player] rating model is flawed...”
– Darren O’Shaughnessy, St Kilda Football Club senior analyst (via Unpluggered Podcast).
Ultimately, we combine these three components into our model with multiple inputs to identify where each player currently sits in comparison to each other.
Our AFL Player Ratings will update weekly following the conclusion of all matches. Over the next few weeks and months you'll see lots more content and updates to the ratings ratings page/s to give you a deeper understanding of how the AFL player ranks stack up.
Did you enjoy this article? Join our free mailing list to get the best content delivered straight to your inbox, or join the conversation by leaving a comment below or on the Stats Insider Twitter or Facebook page.