November 1, 2011 - by David Hess
The preseason AP college basketball poll came out a few days ago, and if you’re not a hoops junkie — if you don’t know which players were graduating seniors, what teams bring back all their starters, or who fell victim to the injury bug — you might not have much of an idea why teams are ranked the way they are. Specifically, why are Memphis, Vanderbilt, and Baylor so high? Why did Kansas fall almost a dozen spots from last season?
The answer is pretty simple: the first three teams listed above bring back most of their key players, while other top squads from last year — Kansas, Arizona, Kentucky — lost a lot of talent. But what to we mean by “a lot” and “most”?
To quantify the amount lost by each team, we calculated the individual Offensive Efficiency, Usage Rate, % of Minutes Played, and Defensive Rating (using definitions from Basketball On Paper) for every player from last year, and used those values to create a single “percent of returning production” value.
Here’s a quick overview of what the above stats measure:
Using those four stats, we came up with simple offensive and defensive “Value Over Replacement Player” measures, roughly analogous to Cracked Sidewalk’s Value Add statistic, but without the opponent adjustments. [Nerds: the formulas can be found at the bottom of this post.] We then just summed all the player values to get a team value. From there it was easy to find what percent of the team value was contributed by players that will be returning for the 2011-12 season.
Finally, to get a single value for each team, we combined the returning offense and returning defense percentages using a weighted average. We figure it’s more important for good offensive teams to return their key offensive players than their key defensive players, and vice versa.
So, finally, here is the result, along with the final TR Predictive Power Ratings from last season, as well as the preseason AP rank for each team:
Team Returning Offense Returning Defense Overall Final 2011 TR Rank AP Poll
North Carolina 86% 80% 83% 13 1
Kentucky 54% 53% 53% 6 2
Ohio St 58% 61% 59% 1 3
Connecticut 50% 62% 55% 8 4
Syracuse 80% 77% 79% 14 5
Duke 46% 64% 55% 3 6
Vanderbilt 97% 95% 96% 34 7
Florida 58% 56% 57% 15 8
Louisville 67% 70% 69% 16 9
Pittsburgh 57% 53% 55% 4 10
Memphis 86% 89% 87% 81 11
Baylor 78% 79% 79% 80 12
Kansas 26% 32% 28% 2 13
Xavier 70% 59% 65% 39 14
Wisconsin 52% 50% 52% 10 15
Arizona 54% 57% 55% 19 16
UCLA 57% 59% 58% 49 17
Michigan 71% 84% 77% 26 18
Alabama 74% 57% 63% 50 19
Texas A&M 56% 66% 61% 43 20
Cincinnati 74% 71% 73% 17 21
Marquette 59% 63% 61% 24 22
Gonzaga 75% 69% 72% 42 23
California 86% 80% 84% 68 24
Missouri 79% 73% 77% 40 25
A few teams stand out in the above table, some for good reasons, and others for bad:
Of course, we didn’t calculate all these player stats (Offensive & Defensive Rating, etc) just for this post. We’re hoping to eventually add them to our site, along with a variety of other tempo-free player stats.
In the meantime, we’re using the team returning production values to create a set of preseason ratings, which will be used to drive our season projections. Those should up soon. Follow us on Twitter at @TeamRankingsCBB or subscribe via email (on the right side of this page) to be notified when they’re posted.
And now, that note for the nerds that we promised earlier…
If you’re familiar with the concept of a replacement player, you’ll find these fairly simple. Basically, we took the difference between each players ratings and those of a bad player, then multiplied them by the amount of possessions and/or minutes used:
Offensive Value Over Replacement Player = (Offensive Efficiency – 86) * Usage Rate * % Minutes
Defensive Value Over Replacement Player = (116 – Defensive Efficiency) * % Minutes / 5
You might think the defensive equation looks a little weird. Remember, the equation for Defensive Rating already accounts for defensive usage rate. It is essentially a measure of the team defensive rating, with an adjustment for the good/bad play of the player. So we take that team value and divide by the 5 players on the court. It’s a bit of a fudge, but it ends up creating Defensive VORP values that have about half of the range of the Offensive ones. This seems reasonable, given that box score data can shed so much more light on offense than defense.
As always, feel free to leave comments, questions, or suggestions below…
Printed from TeamRankings.com - © 2005-2024 Team Rankings, LLC. All Rights Reserved.