ESPN

Footy Forensics: Ladder predictor

After simulating every match of the 2017 season 10,000 times, we've predicted the likely finishing position of every AFL team. How is your club shaping up?

Greater Western Sydney, by general consensus, are the favourites to take the 2017 AFL premiership, but don't be fooled into thinking they have the season stitched up just yet. In fact, Elo ratings suggest the Giants only have the fourth-highest probability of making the finals, behind the Swans, Crows, and Bulldogs.

Our estimates of teams' chances of making the finals are based on Elo ratings, which are a popular way of rating sports teams. In an Elo ratings system, teams are rated based on their match results, after taking into account home-ground advantage and the quality of their opposition.

We use these ratings to predict future matches, then simulate the season 10,000 times to calculate each club's chances of ending up at each position on the ladder. These projections take into account the quality of the team at the end of 2016 and the difficulty of their schedule in the coming season.

While the Giants are heavy favourites with bookies, experts and fans alike, our projections suggest there is hope for Leon Cameron's rivals. In fact, our projection of the 2017 season makes it clear that almost every team has a meaningful shot at playing September football this year.

The Swans sit in pole position, with a 53.9 percent chance of a top-four finish to the regular season, but five other teams also have a better than one-in-three shot at the top four. Virtually every club besides the lamentable Lions have a meaningful chance of making finals. Essendon only have a relatively slim one-in-15 (6.6 percent) chance, but that might be understated for reasons we'll get to in a moment.

Perhaps the key insight from simulating each team's finals chances is the level of uncertainty about the coming season. Even the most fancied teams have a 20-25 percent chance of missing the finals, because of the possibility of bad luck with injuries or just because they turn out to be worse than expected.


The limitations of Elo

Elo ratings are fairly simple - they only take into account teams' past performances, adjusted for home-ground advantage and the quality of their opposition. But they're powerful - match predictions from Elo ratings are pretty accurate. They tend to get more accurate as the season goes on, as the ratings system learns about the quality of the teams. As such, we will be updating our ratings on a monthly basis throughout the season.

At the start of the year, Elo ratings are less accurate as a measure of teams' true strength. Pre-season Elo ratings really just reflect how good teams were last year, with an across-the-board adjustment to bring the good teams down to earth and lift the bad teams back up towards average. The ratings don't take into account changes in personnel in the off season.

Essendon provides perhaps the best illustration of the limitations of Elo ratings. When the Bombers lost a large chunk of their list to the supplements saga in 2016, the ratings system took a few games to catch up to how bad the 2016 Bombers were. Similarly, the ratings system doesn't know the Bombers have regained a lot of their suspended players for the current season, so it's probably underrating their finals chances for 2017 at the moment. By contrast, the Kangaroos are almost definitely worse than their rating suggests at the moment, because the rating system doesn't know that North dropped a gaggle of ageing champions on the scrapheap.


What do punters think?

Elo ratings don't take into account trades and other list changes in the offseason. So what are teams' finals chances after taking those things into account? To see what clubs' chances might be when we take the playing list into account, we can use two different sets of predictions.

The first is the betting market odds, which embody the judgement of punters and bookies about each team's chance at playing September football. The second comparison is a set of season projections by Champion Data, based on the AFL Player Ratings. The betting market odds, we confidently assume, take into account all relevant factors, including the quality of the playing list, the difficulty of the fixture, and the chance of catastrophic injuries. The Champion Data projections are based on assuming that each team has its best 22 players available for each game, excluding Nic Naitanui, Mitch Wallis and Jarryd Roughead (the projections were released in November, before we knew Roughead would return).

The Elo ratings and the betting markets have similar estimates of almost every team's finals chances, bar North Melbourne and Essendon. As expected, Elo is too generous to the Kangaroos, giving them close to a 50 percent chance of making the top eight. The betting markets make it more like a 25 percent chance, while the AFL Official Player Ratings are very unimpressed with their squad, giving them only a 3.1 percent chance of playing in September. Similarly, the Elo model underrates the Bombers' chances, with the betting markets giving them a roughly one-in-three chance of making the eight.

The closeness of the Elo and betting market estimates for all teams other than the Bombers and Kangaroos suggests that off-season changes to the player list - unless they're as dramatic as they have been at Arden Street and Windy Hill - don't make that much of a difference to teams' finals chances, or if they do, punters don't have a systematically accurate way of taking them into account.


How many teams move up or down each year?

Perhaps the most straightforward way to forecast each team's performance is to look across AFL history (1990 to 2016) and calculate what proportion of teams fell or rose up the (regular season) ladder. For example, 46.2 percent of clubs that finished in the top four one season went on to finish top four the following year, with most of the rest finishing in the bottom part of the eight. Only a very small proportion (6.7 percent) of top-four teams ended up in the bottom section of the ladder the following season. Other than the 2016 Dockers, we have to go back to 2008 to find a team who went from top four to the bottom bracket, with both Port Adelaide and West Coast plummeting in that season.

Similarly, teams that finish in the bottom part of the ladder one year tend to remain there the following year, with 50.9 percent of teams that finished 13th or below in one season went on to do the same the next year. Only nine teams in the quarter century of AFL football have managed to climb from the bottom bracket of the ladder up to the top four in the following year - the most recent was the 2012 Adelaide Crows.

Dramatic changes from year to year are uncommon. Gold Coast and Carlton fans shouldn't get too excited about their flag prospects this year. But experience, modelling, and the betting markets all show that the outcome of the season is far from certain. Don't believe anyone who tells you they know how this season will play out.

The Elo-based projections will be updated every month or so at ESPN. Check back to see how your team's finals chances evolve after the first few rounds.

Matt CowgillMatt Cowgill is an economist by day and footy stats obsessive by night who thinks there's no such thing as too many graphs. He uses data analysis to tell stories for ESPN's Footy Forensics, as well as his own blog The Arc.

Start your AFL tipping competition now at footytips.com.au

More Stories