This is something I’ve been working on for a while now, and thinking about for even longer. FanFooty Guru is a set of numbers published for each round of the AFL home & away system giving a prediction of each selected player’s Dream Team and Super Coach fantasy point scoring for the week. They are now showing up in the Fanplanner pages and will be a regular feature of our Exteam and live scoring pages in future.
Projections are a well-established part of American fantasy sports. Publishers such as Sportsline, ESPN and Yahoo have experts for all their fantasy sports who prepare huge tables full of statistics indicating past performances and crunch the numbers to produce predictions for each player each week. I’ve just touched the surface of what is possible to do with AFL statistics, but here’s what goes into the FanFooty Guru projection, in order of importance and how they are applied:
• Rolling average in all recent games
• Rolling average versus opponent
• Rolling average at the ground of this week’s game
• Rolling average at the time that this week’s game is played (day, night or twilight)
Then there are modifying factors, which change the results of the above on a percentage basis:
• Health
• Guru nudge, a subjective push of the numbers by me when I think a number doesn’t look right, or the above factors produce a weird result
• Normalisation, a squishing up or down of all of a team’s scores to take into account the average amount of total points the opponent has given up in recent matches (i.e. this solidifies things like the Sydney effect)
The formula, while it looks complicated, is still relatively simple at the moment. I will investigate more granular factors like team vs type to take into account factors like Richmond and Carlton being harsh on midfielders, or how tall forwards usually escape the Sydney effect. Also there is currently no allowance for taggers or weather.
Note: The Guru points for round 1 are based on teams which I’ve collated from the various “best 22” threads on BigFooty club boards and elsewhere. Thus you’ll see players like Ricky Petterd and Patrick Dangerfield benched, even though they are every chance to play in R1. Of course I’ll update these when the teamsheets come out. At the moment there are no Guru points predicted for rookies and draftees either – another aspect I’ll have to work on before the start of the season.
So, what do you think the weightings of the various factors should be? Are there any other numbers you’d like to see tossed into the cauldron to produce the projection?
