Tuesday, February 1, 2011

Introducing the AFL Power Rankings

Over at ESPN.com each week, columnists Marc Stein and John Hollinger produce their NBA Power Rankings. The aim of these rankings is to ‘look through’ the standings to see where each team really stands in the scheme of things. Stein’s rankings are based on his opinion, while Hollinger’s rankings are based on a statistical formula.

I have wondered if Hollinger’s approach could be translated across to the Australian Football League. After some pondering, I think a formula that could be used is as follows:

Team Score = 2/3*(Average relative adjusted net margin over past 22 games) + 1/3*(Average relative adjusted net margin over past 5 games)

Let me explain:

1) Like Hollinger, I think the strength of each team should be based on their average net margin (points scored less points conceded) rather than their wins and losses. I have not checked if winning or losing margin is indeed a better predictor of a team’s future success in AFL than wins and losses, but if makes sense that it would, given teams that keep winning close games are riding their luck to an extent.

2) Average net margin needs to be adjusted for the strength of teams played, hence giving us the concept of average relative net margin. Average relative net margin is calculated as the average over all a team’s games of their net margin against each opponent less the average net margin by all teams against each of those particular opponents. So if Team A beats Team B by 3 points but if, on average, teams beat Team B by 13 points then relative net margin for Team A from playing Team B is -10 points. Over the long-run, as they play most of the teams, a team’s average relative net margin should more or less simply approximate a team’s average net margin.

3) Average relative net margin in turn needs to be adjusted for the advantage from playing at home and/or the disadvantage from playing interstate, giving us the concept of average relative adjusted net margin. To do this, all the net margins for a team would need to be adjusted by the relevant advantage/disadvantage (which would need to be worked out) from playing home/away. So if a Team A wins by 3 points when playing interstate but the disadvantage from playing interstate is 18 points then the adjusted net margin of Team A is 21 points.

4) I have put two-thirds weight on a team’s success over the last 22 games, and one-third weight on their success over the last 5 games. That seems to me a reasonable estimate of how people generally weigh up a team’s performance in recent matches.

If/when I can put some figures together, I’ll try and put this system into practice.

1 comment:

Dixy said...

I like Stein's subjective power rankings better. Hollinger's just seem like overkill, and he makes his predictions based on them lseemingly to justify the results his formula creates.