: What do these ratings mean?
A: They represent each team’s projected
per-100 possessionperformance — schedule-adjusted and relative to league average — for the coming week, taking into account the quality of players on each roster, as well as injuries and expected minute allocations.
Q: How is player quality measured?
A: Using ESPN’s
Real Plus-Minus (RPM), which attempts to isolate each player’s contribution to the team’s scoring margin while on the court by adjusting for the quality of his teammates and opponents faced. While the version of RPM
listed at ESPN.com is a single-season metric, these power ratings use the more predictive multiyear version of RPM.
Q: Where do the rosters come from?
A: ESPN’s
depth charts and
injury wire.
Q: Who generates the projected minute allocations?
A:
Jeremias Engelmann, the creator of Real Plus-Minus, provides the minute projections for each team.
Q: How are these different from other computer power ratings available, such as the
Hollinger Power Rankings?
A: Most power ratings are, to some extent or another, backward-looking; they can only generate ratings using inputs from games the team has played. Given a large enough — and relevant enough — sample of played games, this is usually not a problem. But in the case of early-season rankings, or when a team experiences roster changes midseason (via trades or injuries), it takes time for traditional power ratings to catch up to the team’s new quality.
These RPM power ratings, however, are based on the talent of the players on hand for each team. The advantage of this approach is that when a player is added to or subtracted from a team, a talent-based rating can adjust immediately, without waiting for new games to be played. In other words, injuries, trades and signings are instantly accounted for in these rankings.
The other side of that coin is that, barring personnel changes, these ratings aren’t going to change drastically from week to week. RPM player talent estimates have a strong grounding in
Bayesian statistics; and for veteran players, their
prior rating carries a good deal of weight. So, while a team’s “statement win” in a given week might have a tangible effect on human or even recency-weighted computer power rankings, it’s unlikely to move the needle much with these ratings
