(In the very last post of 2013 on the Statistical Analyses journal Andrew introduced us to the concept of Set of Games Ratings - team ratings created by fitting Ordinary Least Squares regressions to portions or full seasons of actual results.
Andrew returns today with some additional results.)
The Set of Games Ratings post from late December introduced a twist on Tony’s VSRS concept. For any given set of games, the SOGR approach produces a rating for each team indicating its relative scoring ability within those games. Each SOGR model is optimised for the set of games on which it was fitted (in the least squares sense).
That initial post looked at the performance of Carlton across the ratings period. This post presents the charts for all the teams for the same period: 1999 to 2013.
To reiterate, the annotations on the chart are as follows:
- Black line: tracks the ratings from the simple VSRS model
- Red line: marks the SOGR for each team in each season
- Green line: marks the SOGR for each team in the first half of each season
- Blue line: marks the SOGR for each team in the second half of each season
Analysis of this rating data is ongoing and I'll be writing about the outcomes of that analysis in another post. In the meantime, here are the pretty charts for your perusal.
The charts reveal, unsurprisingly, the broad agreement of the week-by-week VSRS rating and the various SOGR ratings. In a post-modernist vein, it does occur to me that the detachment of data analysis belies the intense commitment that teams and players invest for even modest movements in their ratings. (Is post-modernism allowed on a statistics blog?) [Yes - ed.]