Wednesday, August 14, 2013

Those AFL Power Rankings Again: What If An Opponent’s Worth Was Weighted Towards More Recent Games?

In my AFL Power Rankings, which you could not have failed to miss if you have read these pages before, a team’s ranking is a weighted sum of its ‘relative adjusted net margins’ over its past 22 games. The weights are higher for more recent games, and a team’s relative adjusted net margin for a game is its net margin (points for less points against), adjusted for home ground advantage, and adjusted again for the ‘worth’ of its opponent.

However, one thing that has always slightly bothered me about this formula since its inception is that while a team’s performances is weighted towards more recent games, the ‘worth’ of its opponent is not. The ‘worth’ of its opponent is simply the average of that opponent’s net margin adjusted for home ground advantage over the opponent’s past 22 games. Therefore, if an opponent is on the rise, a team will not get as much credit for its performance against them than if the opponent’s worth was weighted towards more recent games (think Bulldogs at the current point), and will get ‘too much credit’ if the opponent is falling (think St. Kilda).

Tonight, I decided to finally see what would happen if I altered this. Under my ‘alternative system’, each team’s adjusted net margin (i.e. their worth as an opponent) is weighted according to the same formula as its relative adjusted net margin usually is. (Note that I’m not planning to switch systems mid-season.)  You can see the difference this would make in the table below. I have also included each team’s adjusted net margin under each system to show how its worth as an opponent varies.

First compare columns C and D (which I should probably have made columns A and B). Teams that are improving in form, such as North Melbourne, Collingwood, and the Western Bulldogs, are considered harder opponents under the alternative system (column C) than under the current system. Conversely, teams that are deteriorating in form, such as St. Kilda, Carlton, and Essendon, are considered easier opponents under the alternative system.

Now compare columns A and B, which compares ranking points under the alternative system with the current system. You can see there is not a lot of difference in each team’s ranking points. Adelaide moves up a spot, because its most recent opponent North Melbourne is considered tougher (it has a higher adjusted net margin) under the alternative system. (Every team’s ranking points also move up a bit - I won’t go into the details, but essentially it’s because the alternative weighting system lowers the effects of last year’s final series.) Gold Coast and the Bulldogs also switch spots, with the Bulldogs’ win against Carlton last week not considered as highly under the alternative system. But overall I’d say this shouldn’t be keeping up at night … ahem, not that it was.

Later on, I might try showing the effects of different weighting systems on performances. For example, what if even more weight was given to recent games? There might be a few other experiments I run before next season.

No comments: