Preseason ratings were posted over the weekend. Here’s a reminder as to what goes into them…
The components and weighting is based on a regression of the past nine seasons. The system is, by 2015 standards, pretty simple. It doesn’t try to project playing time for individual players. It doesn’t know about transfers, and all but 5-star recruits are virtually ignored. If you think your favorite team is ranked too low, the reason is probably that there are really good transfers or recruits arriving.
In the most general sense, the main ingredient in the system is inertia. If a team has been good in the recent past, it’s likely to be rated well in the preseason. As much as we like to think of college basketball as this crazy sport where anything can happen, there’s just not much class mobility in the game. I think we all understand that the Big Ten will always be better than the SWAC, but even within conferences there’s a clear power structure that might vary from year to year but is very predictable over the long term.
Over the next decade, it’s a near certainty that Arizona will win more Pac-12 games than Washington State, Kansas will win more Big 12 games than TCU, and Duke will win more ACC games than Boston College. So in the absence of looking at specific players, the projection first relies on recent team performance. Projected offense is largely determined by the quality of a team’s offense over the previous three seasons and its defense from last season. Projected defense uses similar variables.
Returning personnel is considered as well. Generally, the more players returning, the better. However, the quality of the player is also a factor. Losing a high-usage/high-efficiency player hurts a team’s offense a lot more than losing a role player. In fact, a low-usage inefficient role player that returns can actually hurt a team’s rating. So while chasing off a player is not the most ethical practice, it is apparently a good sign for the program when a struggling player seeks a new school.
A new addition this year is accounting for injured players from the previous season. If a returning player played a partial season, then his future impact is considered to be more than it was in previous versions of the system. Note that his doesn’t account for cases where a player completely missed the previous season. Michigan is the team that benefits most from this, with both Caris LeVert and Derrick Walton returning after having missed large portions of last season.
There is also a penalty for a coaching change, with a greater hit for teams ranked higher. Basically, teams that change coaches tend to underperform their counterparts that have not changed coaches, all other things being equal.
The predictions are calibrated on end-of-season ratings. However, in a departure from previous practices, the national averages I’ve subjectively imposed are designed to capture conditions at the beginning of the season. Currently, I’ve got an efficiency of 100 as the national average. The end-of-season number will undoubtedly be higher than that. Likewise, the average tempo of 68.5 should be an overestimate of the season-long average. Although, that is less certain.
While the projections for offense and defense are no longer state-of-the-art, I think the tempo projections are the best around. (Mainly because I’m not aware of any competition.) It’s far more accurate to project a team’s pace using their head coach’s history than the team’s history, and that is how this season’s tempo projections work. In the case of rookie head coaches, there’s just some regression to the mean applied to the team’s tempo from its previous season. But for coaches with a history there’s a lot of weight placed on that history.
For the sake of accountability, here’s how various pre-season systems did last season, measured by average error in predicting conference wins across all 350 teams.
Hanner/SI 2.15 Hess/TeamRankings 2.16 Pomeroy 2.26 Adamson/Matchup-Zone 2.26 Torvik/T-Rank 2.35
It would be better to use overall regular-season record but the issue of in-season tournaments makes this quite a bit more difficult. At any rate, as has occurred in previous seasons, Hanner sets the standard in projections, although TeamRankings is a very, very close second. If you want your preseason ratings considered in future editions of this post, please send your conference predictions along before the season begins.