BIG BLUE FANS FOR

BASKETBALL

2000-10 Season Analytical Writings

45
2009-2010 Analysis of Efficiency, Pace, and Schedule Strength

Review of 2009-2010 Basketball Season
Based on Pace-Strength of Schedule-Efficiency Factors

Overview; Scope of Data Base:

In 2009-10, the NCAA D1 included 347 teams. Assuming that these teams played an average of 30 regular season games each, and that post-season conference tournaments to eliminate all but the conference tournament champions included another 316 games, the Post Season NIT included an additional 31 games, and the NCAA Tournament included another 64 games, these teams played approximately 5,617 games. Those games averaged 67.3 possessions for each team, or approximately 378,000 total possessions.

Over the course of those games and possessions, the average efficiency for D1 basketball was 1.011 points per possession [ppp], on offense and defense. The better offensive teams scored more efficiently than this average, and the poorer offensive teams scored less efficiently than this average. In 2009-10, the best offensive team averaged 1.166 ppp and the poorest D1 offense averaged 0.807 ppp for their season. Kentucky averaged 1.121 ppp during 2009-10. The best defensive team held opponents to only 0.888 ppp, and the weakest NCAA D1 defense allowed opponents to score 1.167 ppp. Kentucky's defense held opponents to 0.913 ppp during 2009-10.

Kentucky clearly played offense and defense better than the NCAA average, but the performance at each end of the floor fell short of the gold standards of the best. Kentucky's offense was 12% better than average, and the best offense was 17% better than average. Fifteen D1 teams posted stronger offensive efficiency for the season. Kentucky's defense was 9% better than average, and the best defense was about 11% better than average. Ten D1 teams posted stronger defensive efficiency for the season.

Net Game Efficiency[NGE]:

Clearly, Kentucky's 2009-10 team was better on a comparative basis than top 10 or even top 16 based on the defensive and offensive efficiency rankings noted above. That is because few teams ever combine the quality of offensive and defensive efficiency that Kentucky posted during the 2009-10 season. The real measure of success cannot be located in offensive efficiency or defensive efficiency values alone. Rather, the real measure of success is the average spread between the two, and teams that maximize that spread are the teams that will win most frequently.

This spread is called the Net Game Efficiency [NGE] and is really nothing more than a measure of average margin of victory or loss, normalized to eliminate pace of play as a variable. In 2009-10, the highest NGE value for the season was 0.248 ppp [Kansas] and the lowest was a negative 0.270 ppp. The average for all NCAA D1 teams is 0.000 ppp. Kentucky's NGE was 0.208 ppp, which ranked 5 th in the nation. The top 4 teams were Kansas [0.248 ppp], Duke [0.242 ppp], BYU [0.229 ppp] and Murray State [0.212 ppp].

Strength of Schedule [SOS]:

It should be intuitively clear that the raw NGE values discussed and presented above could not tell the entire story. For example, while Murray State had a nice team, and turned in an impressive season, they were not truely worthy of inclusion in any serious discussion of contenders for national honors in 2009-10. To a lesser extent, the same can be said about BYU's national aspirations last year.

Why not? Each of their respective raw NGE values were impressive are rarely high. This dichotomy between the raw NGE values and the actual ability of a team to compete on the national stage indicates that one additional factor must be considered to arrive at legitimate efficiency based rankings. That final factor is the strength of schedule. Pomeroy reports schedule strength data for all D1 teams for 2009-10. For these top 5 raw NGE teams, their respective Pomeroy Schedule Strengths are as follows:

There is no disputing that schedule strength is an important and relevant issue. It is clear that the BYU and Murray State raw efficiency values do not possess the same inherent value as Kentucky's, Duke's or Kansas' because neither BYU nor Murray played schedules of the same strength.

The real issues with respect to schedule strength are first how can it be objectively measure and second how should the SOS be applied to the raw NGE values to arrive at an appropriate adjusted NGE that will accurately reflect the relative strengths of teams.

On the first issue, I have found two widely used measures of schedule strength. The RPI system provides a measure of schedule strength based on each team's opponents and opponents' opponents' winning percentage. The RPI SOS values seem to provide a method to assess whether Team A or Team B played a stronger schedule, but there does not seem to be any analytical foundation for that methodology. Pomeroy also provides a measure of SOS. Pomeroy uses the average offensive and defensive efficiencies of opponents to provide the basis for relative comparisons.

I have examined and used both systems over recent years with only limited success. At present, I am leaning toward the Pomeroy system because it appears to have a slightly stronger analytical basis than the RPI system. However, I continue to hold the view that much better methods for SOS measurement can and should be developed in the future.

How Does 2009-10 Compare Statistically To Four Previous Seasons?:

For those who are not familiar with the data, and the absolute values of the various numbers presented above, it is helpful to see how a particular season's results compare with other recent seasons. The following Table provides that comparison for the last 5 seasons. This span of time includes Tubby Smith's last 2 seasons, Billy Gillispie's 2 seasons, and John Calipari's first season at UK.

For each major statistical category, the table provides the maximum, average, minimum, and UK's result. In addition, the 5-year average for each category appears across the bottom of the table. The values for efficiency, SOS, and pace in 2009-10 were similar to those that have occurred in each of the four previous years. My experience tracking these statistics for nearly 2 decades indicates that the 2009-10 values are typical of those that have occurred in college basketball for nearly all years since I began tracking this data in the early 1990s.

Adjusted NGE:

In 2009-10, the average Pomeroy Strength of Schedule was 0.515. Pomeroy has develop an algorithm for the adjustment that he prefers. I have also been working for several years independently along parallel, but not identical lines to develop a relationship as well. The strongest schedule in 2009-10 has a Pomeroy value of 0.828 and the weakest was 0.163. All values deviate from the mean between a maximum deviation of 0.313 and a minimum deviation of 0.3517. The adjustment for schedule strength is based upon the magnitude of the individual team's value from the mean. The following table shows the deviation for each of the 347 D1 teams on the X Axis and the Pomeroy Adjustment for NGE on the Y Axis. The trend line shows that the adjustment is approximately 39% of the actual deviation.

When these strength of schedule adjustments are applied to all teams, The Top 10 and their respective Adjusted NGE are:

The distribution of all 347 D1 teams by their year ending Adjusted NGE is shown in the following graph.

Pace of Play:

Pace has been eliminated from the NGE analysis, but pace remains a topic of interest nonetheless. The average NCAA game in 2009-10 was played at a pace of 67.3 possessions for each team. That is just under 135 total possessions in 40 minutes, or about 17.8 seconds per possession. A possession for this analysis is defined as each trip into offensive court by a team, and the possession does not end until the other team takes possession of their own. Therefore, when a team shoots, misses, and retains possession due to an offensive rebound, it is simply a continuation of the original possession, not a new possession. The fastest team in 2009-10 was VMI who played at a pace of 84.8 possessions per game [[14.2 sec/possession] and the slowest team in the nation last year was Samford at 57.5 possessions per game [20.9 sec/possession]. The following graph illustrates the distribution of pace from fastest to slowest teams.

The vast majority of NCAA D1 teams play at a pace between 62 and 73 possessions per game, and only a handful of teams play at a pace outside this range, either on the high or low sides. Those that did play outside these general limits generally did not rank in the top 75, especially on the high end. There are top 25 teams that play at both ends of this common range, and the average pace for the top 25 teams in 2009-10 was 67.9 possessions per game. Kentucky's pace was 70.3 possessions per game, which ranked tied for 53 rd in the nation with Vanderbilt and Siena.

The 10 fastest teams in 2009-10 were:

Only Texas and Providence appear from BCS conferences in this list.

The 10 slowest teams in 2009-10 were:

Only Wisconsin appears in this list from a BCS conference.

The pace of the top 10 NGE teams were:

Only BYU appears in this list from a non-BCS conference.

The following graph shows the pace of all 347 D1 teams, based on their final NGE ranking.

The next issue is how the strength of the top 60 teams [Simulation of NCAA Tournament Field] varies from year to year. In 2009-10, 2 teams ended the season with adjusted NGE values above 0.300 points per possessions, and one of those finished above 0.350 ppp. There were 9 additional teams with adjusted NGE values between 0.250 ppp and 0.300 ppp, and 20 more between 0.200 and 0.250 ppp. Teams with adjusted NGE values above 0.200 ppp represented about 1/2 of the NCAA tournament field. The following Table shows the distribution of the top 60 ranked teams for seasons 2003 through 2010 by the range of teams' final adjusted NGE values. The right most column provides the average adjusted NGE value for the top 60 teams for each season.

As this table indicates, the strength of the tournament field does vary from season to season. The strongest fields occurred in 2007 and 2008, and the weakest fields during the 2003-2010 period occurred in 2003 and 2010. Typically, there is 1 team above 0.350 ppp, and between 3 and 4 additional teams with adjusted NGE values between 0.300 and 0.350 ppp. This indicates that teams desiring a #1 seed in the tournament need to post adjusted NGE values above the 0.300 ppp threshhold. At the other end of the spectrum, about 25 to 26 teams post adjusted NGE values between 0.200 and 0.300 ppp in a season. These teams generally correspoind to teams that advance into the second round and battle it out for trips to the Sweet 16 round.

The following table shows the adjusted NGE values for the top 60 teams, sorted on ranking categories. The categories presented are the top ranked, second ranked, and 3rd and 4th ranked teams in the first 3 columns. These teams represent the teams that earn the #1 seeds. The 4th column represents teams ranked 5 to 8 [#2 Seeds], and the 5th column is for teams ranked 9 to 16 [3 and 4 Seeds]. As this Table demonstrates, #1 Seeds generally require adjusted NGE values above 0.300 ppp, and #2 Seeds have averaged 0.289 ppp over the last 8 seasons. 3 and 4 Seeds have averaged 0.259 ppp, and seeds 5 to 8 have averaged 0.221 ppp. The 9-16 seeds have averaged about 0.173 ppp.

Submitted by Richard Cheeks

Submitted by Richard Cheeks

44
2009-2010 NCAA Tournament Prediction Analysis

Copyright 2010
SugarHill Communications of Kentucky
All Rights Reserved