How AFL Fantasy Pricing Works (Part 2)
Understanding the behaviour of the AFL Fantasy pricing mechanism
--
Postscript : Beta version of app to estimate next round player breakevens here. Please leave me a clap if you find the analysis to be useful.
In this article I present a guide to understanding the definitions and order of the calculations used for the AFL Fantasy player pricing mechanism. Hopefully the article will also help unpack some of the conundrums of why the numbers aren’t what they always seem to be on the footywire pages.
My understanding of this mechanism is based on a previous analysis of player prices for the 2015–2019 seasons.
Order of Calculations
To fully appreciate the definitions we need to first consider the order of the calculations for a given round.
- Before the match is played, the aggregate prices of all players who will be playing in that round is calculated.
- After the match is played, the weighted average score for each player is calculated.
- Using (2) — the weighted average score for each player — the aggregated weighted average scores of all players for that round is calculated.
- Using (1) and (3) — the aggregate player prices and the aggregated weighted average scores — the magic number is calculated.
- Using (2) and (4)- the weighted average score and the magic number — individual player prices are calculated.
- Using (4) and (5) — the breakevens are estimated.
The section below discusses the definitions of each of the highlighted terms and mulls over implications for individual players.
Weighted Average Score
The weighted average score takes into account the most recent games played during that season. The weightings are such that the the most recent game played carries the greatest weight.
The weighted average score for each player can be estimated as follows -
The weighting depends on up to five of the most recent games played for the season. Note that, if only two games have been played then only the first two weights are used in the calculation.
Magic Number
The magic number is intended to be a rebalancing factor which ensures that for a given round, the aggregate prices before and after the match has been played for all players playing that round remains a constant.
This does not imply that the individual player prices before and after the match is a constant.
The magic number for a given round can be calculated as follows -
As a general observation, the magic number decreases over the course of the season because a proportion of players who haven’t played the full previous season(s) are priced at a discount to their previous season average.
- The preseason magic number is artificially high because it uses only the scores for players who have played the previous season in the denominator of the calculation.
- Once the initial total value of all players who played in the prior season is determined, the preseason magic number can be calculated.
- Following the calculation of the preseason magic number, the preseason prices for individual players can be determined, using either their previous season average scores or a discounted figure for rookies and other players who have not played a threshold number of games in the previous season.
From the above, the aggregate value of player prices can be determined, as well as the aggregate value of those who played in R01 which becomes inputs to the R01 magic number calculation.
After R01 is played, because the total scores of players will by definition be more than the implied preseason score, the new round magic number will decrease relative to the preseason magic number.
This mechanism, where not all players are priced at their expected scores (particularly rookies), causes the magic number to fall during the season to a normal value which implies that a self-correction process occurs to player prices by the end of the season.
Magic numbers have fallen by 4–7% over the course of the 2016–2018 seasons, roughly 0.25% on average per round, although this is not evenly distributed for each round — it depends on the makeup of the cohort of players who participate in that round.
For example a round with a greater relative proportion of rookies vs the norm (perhaps the byes?) would likely have a lower aggregate weighted average score, hence the magic number would be higher, all else being equal.
In terms of starting price for the 2019 season, a player is priced in relation to their 2018 average. If a player played 10 plus games, they will be priced at their 2018 average. A 3% discount is applied for every game less than 10, however 2017 season average is used as the base if the 2017 average is higher than the 2018 average.
Player Prices
The player prices at the end of the round are determined using a combination of their prices before the round begins and their weighted average scores for the 5 most recent games played.
Depending on the number of games played so far, the weight attributed to the new price can be between 75–91% of the previous round price, and conversely the remainder is attributed to their scores.
As the season progresses, a reducing percentage of their most recent price can be attributed to the preseason price, that is, their prices become more dependent on the weighted average scores — hence their price will converge towards the product of the magic number and the weighted average score.
Breakevens
The breakeven is by definition the score that a player needs to obtain such that his price remains unchanged after playing that game. The breakevens on the footywire website are only a best guess of what that score might be.
This is because the individual players prices at the end of a round are determined by the magic number which is in turn dependent on the aggregate weighted average scores of players playing that round, which is an unknown until all matches have been played for that round.
Breakevens can be reverse engineered in conjunction with your own magic number estimate using the pricing equation above. This is a useful technique to estimate the path of future breakevens given assumptions about future player scores.
As noted earlier, the final player price will converge to the product of the magic number and the weighted average scores — hence the breakeven at the end of the season will be approximately equal to the weighted average score for a given player.
In the context of trading players, once a players weighted average score is around his breakeven during the season, his change in price will be small but decreasing as it is dependent on the (downward) path of magic number. This is another way of explaining why the prices of the gun players decreases.
Conclusion
In this article I’ve hopefully managed to give you a more intuitive understanding about the AFL Fantasy pricing mechanism — the relationship between the prices, scores, magic numbers and breakevens and a few observations about how this impacts player prices.
Side note : This is my first season as an AFL Fantasy player and as at the end of Round 6, my team is ranked 70,000 out of around 150,000 players and made up of names I’ve not even heard of before. Hopefully gaining an understanding of this piece of the puzzle will give me ammo to rise up the table for next season.
Data for the analysis was scraped from individual player profile pages and the breakeven page on footywire.
Special thanks to my fantasy footy buddies — Jack, Justin and Selby — for their guidance and encouragement which made writing this article possible.
