Notre Dame defeats Kansas during in the 2013 NCAA Regional. (Photo by Teri Priebe)
Notre Dame defeats Kansas during in the 2013 NCAA Regional. (Photo by Teri Priebe)

Numbers' Game: What do winning percentage and national rankings really mean?

October 28, 2013 - 12:14pm

Dave Dagostino, a college basketball analyst for Full Court, is a former Division I head coach and has also served as a Division I recruiting coordinator in the Big East, Atlantic Sun, and TAAC conferences. A former four-year college basketball player and point guard, he also serves as a game analyst for Krossover and is a member of the Association for Professional Basketball Researchers. Dave has been active in scouting since 2008, serving as a national evaluator for Blue Star, Executive Director of Scouting for the Round Ball Journal, and a consultant on international basketball through both DDJR LLC and SINO. You can follow him on twitter @dagostinodm.

Maybe former NFL coach and current commentator Herman Edwards said it best: “We play to win the game -- HELLO -- we play to win the game.” 

Over the course of a grueling college basketball season, coaches set their teams up for a series of roughly 35 one-game seasons. It’s a psychology, regardless of the conference or level,  that allows a team to focus on one game at a time and accumulate as many wins as possible. Isn’t that what we are all after? 

Regardless of the endeavor, coaches want to win every possession, every practice, every community event -- it is a mindset that allows student-athletes to understand that they cannot turn it on and turn it off. You have to win every day.

As we continue our journey in analytics, I offer you this notion: Decisions that we make as coaches are about earning wins. In the analytic world, wins are a great indicator of success. (I didn’t shock the world with that statement, did I?)

So the next question in our analysis is this: Which statistic correlates most closely and reliably with wins? 

As we have established in an earlier article, point differential has been proven to have such a stable relationship with wins. Below I offer a table reflecting team rankings based on winning percentage for the 2012-13 season, and comparing those rankings to both rank by season-average point differential, an objective measure, and the teams' respective national rankings in the March 18, 2013 NCAA Coaches’ Poll, which is, of course, a subjective analysis.

Take a look and see which is closer to predicting and defining success -- the objective measurement of point differential over the course of the season or the poll rankings. (And remember, we are not proffering analytics as a substitute for the eye test; it is merely a way to make sure we are not victimized but what we see.)

Rank       Team                    Winning %            Rank/Pt Diff         Poll Rank

1               Notre Dame        94.6                          3/21.9                    2

2               Baylor                 94.4                          2/27                       1

3               Stanford             91.7                          6/18.6                    4

4               Duke                   91.7                          4/20.6                    5

5               Quinnipiac         90.6                          26/13.2                  NR

6               Green Bay          90.6                          5/18.9                   20

7               Dayton                90.3                         12/16.8                  15

8               Connecticut       89.7                          1/32.7                    3

9               Delaware            88.9                          16/14.8                  16

10            California            88.9                          24/13.4                   6

11            Chattanooga       87.9                          23/13.7                  NR

12            Toledo                 87.9                          20/14.1                  NR

13            Albany                87.1                          9/17.6                     NR

14            Kentucky            83.3                          10/17.5                   7

15            Hampton            82.4                          13/16.4                    NR

16            Gonzaga             81.8                          19/14.3                   NR

17            Penn State         81.3                          17/14.6                    8

18            Charlotte            81.3                          32/10.4                   NR

19            UNC                    80.6                          48/8.7                     18

20            Georgia              80.0                          27/12.4                    13

21            Boston U            80.0                          34/10.1                    NR

22            Liberty                79.4                          14/16.2                    NR

23            FGCU                  79.4                          11/17.2                    NR

24            SD State              79.4                          15/15.6                    NR

25            Marist                  78.8                          29/11.6                    NR

End of story, right? Not exactly.

Let’s consider the following points:

  • UConn finished with the ninth best winning percentage but was first in point differential -- and the Huskies won the title, didn’t they?  So is point differential a better indicator of a team’s ability to perform than winning percentage?
  • Though point differential correlates far more closely within winning percentage than do poll rankings, plainly neither measure is infallible. For example, North Carolina, a team boasting an .806 winning percentage (ranked 19th in the nation in that category), weighed in at just 48th nationally in point differential. Princeton diverged in the opposite direction with a .759 winning percentage (No. 32 nationally) but an 18.2 point differential (No. 7 in that category). How and why might a team’s winning percentage and point differential diverge? Perhaps the team won a lot of close games, or, conversely, suffered a number of losses but also racked up a lot of double-digit wins. In such cases, one might want to ask: What was the average point differential of their opponents? Did the team play to the level of its competition, letting lesser teams hang around?
  • Should we take a closer look at the record of games decided by five points or less (two possessions)?
  • Should we place a greater emphasis on fourth-quarter point differential as opposed to the spread for the whole game?
  • Where are traditional powers Maryland, Tennessee, UCLA and South Carolina?
  • How does point differential compare with RPI as a predictor of winning?

These are just a few of the questions that will drive our season series on analytics. 

Take a look at some of the questions and run them against your own evaluations. How does your perception of the quality teams match up with these various ways of measuring success?

It does seem that point differential could be a strong leading indicator of success as measured by wins and losses, so our next step will be to examine the specific statistics that have the greatest impact on point differential -- and also to test some of the statistics that may be inflated in regard to their importance to point differential.

Keep the questions coming and I’ll continue to reply.