Defensive Development in the Carolina System

New readers –

This is launch week on Tobacco Road Blues, the new site dedicated to Duke and UNC sports. This afternoon's post comes from Adrian Atkinson, a contributing writer. He'll be covering UNC sports, and basketball in particular. If you're interested in writing for TRB, follow the contact link to the right. Enjoy.

-Shane

A Quantitative Look at Defensive Development in the Carolina System

Conventional wisdom declares that experience matters in college basketball, especially on the defensive end of the court. While effective defense is a function of myriad physical traits—lateral quickness, length/wingspan, and strength, to name three—it is also heavily dependent on a player’s mental attributes. The ability to communicate with teammates, to make crisp and timely help rotations, and to execute the nuances of a complicated defensive scheme are all things that can be mastered irrespective of a player’s level of athleticism. But they can rarely be mastered without a healthy dose of experience. Shared court time with teammates—both on the practice floor and in game situations—is essential for developing the type of defensive IQ and chemistry that all championship-caliber teams have.

Moreover, there is an interaction effect between a player’s experience and his physical traits. By reacting instinctively rather than thinking, a defender can gain a half-step of quickness that might be the difference between a successful help rotation and an unsuccessful one. On the other hand, a player who is not yet acclimated with a defensive scheme can often look paralyzed with indecision on the court. This lack of experience and defensive understanding can erroneously manifest itself as a lack of quickness or effort.

Not all defensive systems are created equal: depending on the degree of complexity in a given system, the steepness of a player’s learning curve will vary. But, conceding that point, coaches universally agree that there is a clear link between experience and defensive aptitude (as reflected in performance and results). The question is—and here’s where the quantitative part comes in—how valuable is a marginal year of defensive experience in Roy Williams’s system at North Carolina?

Methodology And Metrics

To answer that question, we’ll introduce a concept called defensive charting and use a metric known as Stop Percentage (Stop%). Defensive charting was introduced by Dean Oliver in Basketball on Paper, his seminal book about the use of quantitative techniques to measure basketball performance. Basically, one creates a defensive box score that is analogous to the standard offensive box score. By re-watching the game and charting the results, it is possible to allocate credit and responsibility for all of a team’s defensive possessions.

This exercise isn’t as simple as merely assigning credit or blame to the defender whose opponent shot the basketball (or the defender closest to the shot). In many cases, due to the team-oriented nature of defensive basketball, one defender’s breakdown early in a possession will lead to a basket against another defender later in that possession. Individual credit or blame is often split between two defenders too—especially in situations involving high ball screens and a defensive hedge-and-recover. In general, if a defender is beaten off the dribble and allows penetration into the middle of the court, any subsequent basket scored off of that penetration are charged to the initial defender (regardless of the number of passes and help rotations that follow it). The bottom line is that the defensive charting system involves a meticulous review of the game tape, including multiple rewinds of each Carolina defensive possession.

Once the defensive data is charted and recorded, Stop% can be calculated for each player. Stop% is a summary statistic that measures the percentage of individual defensive possessions for which a player records a stop (with 100% meaning he allowed no points, and 0% meaning he recorded no defensive stops). A full stop is earned for a forced turnover. Partial stops (that sum to a full stop) are earned for a forced miss shot and a defensive rebound (in cases where the same defender forces the miss and corrals the board, he earns a full stop).

The value of the forced miss component and defensive rebound component depend on the defensive profile of the team. Teams that are relatively strong on the defensive glass will earn more credit for a forced miss (since these are deemed more valuable to such a team). Conversely, a team that is relatively strong at forcing missed shots will earn more credit for a (relatively more valuable) defensive board. Scoring possessions include both made free throws and made field goals allowed by a defender, and his Stop% is simply: Defensive Stops / (Defensive Stops + Scoring Possessions). A full description of defensive charting and how to calculate Stop% can be found in Basketball on Paper.

Analysis And Findings

It should be noted that this analysis was constrained by the amount of available data; I’ve only been charting North Carolina games since the 2005-06 season. For that reason, sample sizes are fairly small in this study. Even so, findings were consistent with a priori expectations. Like for offensive box score production (for which there is ample statistical evidence across large samples), it is hypothesized that the greatest defensive improvements will be realized between a player’s freshman and sophomore years.

In reality, a good deal of defensive improvement likely happens during the course of a player’s freshman campaign during which he is fully immersed in the new defensive scheme. As the scheme becomes more and more familiar (as a result of practice time, individual instruction, film work, and game experience), team defensive concepts become more instinctive to young players. Due to the level of granularity in the data, we can only analyze between-year improvements—not within-year improvements. With this caveat in mind, it is expected that the rate of defensive improvement will be most rapid during the freshman-to-sophomore year interval.

Defensive box score data is available for 12 Tar Heels who were members of the rotation as both freshmen and sophomores: Bobby Frasor, Marcus Ginyard, Danny Green, Tyler Hansbrough, Wayne Ellington, Ty Lawson, Alex Stepheson, Deon Thompson, Will Graves, Ed Davis, Larry Drew II, and Tyler Zeller. Table 1 reports the defensive box score summary statistics for this group of players in each of their first two collegiate seasons. The final row of Table 1 shows the percentage change for each defensive category.

As seen, the dozen Heels included in this analysis were better across the board defensively as sophomores (with the exception of on the defensive glass where they were slightly worse). The most significant improvement occurred in the area of drawing offensive fouls—an increase of 67% from freshman to sophomore year (from 0.30 per 40 minutes to 0.50 per 40 minutes). Since drawing charges is generally the result of a help rotation, it is not surprising that this skill would be most closely tied to gains in experience. The half-step-late rotations of freshman year that resulted in blocking fouls are replaced by crisper sophomore-year rotations that lead to offensive fouls.

Table 1: Freshman-to-Sophomore Defensive Development—UNC 2006-2010*

Three-point defense is a second category in which significant sophomore-year strides were made. In general, experience allows a defender to take better angles when navigating a screen (and to learn some of the tricks that good shooters use to get open off of a screen). Also, the types of mental mistakes that can lead to clean perimeter looks (e.g., ball-watching when the ball is entered to the post, not locating shooters in transition, etc.) are reduced as a player’s tenure in the system increases. As it specifically relates to preventing 3s in the Carolina system, experience pays dividends in defending the high ball screen (communicating and rotating on the hedge-and-recover).

It is also very important in making proper help decisions from the wing against dribble penetration. In Carolina’s system, the wings pinch in early to stop penetration (instead of allowing the dribbler to get into the paint and force help from the post). This can lead to drive-and-kick 3-point opportunities if the wing makes a poor read (over-helping/committing too soon) or a sloppy rotation. With additional experience, Carolina defenders tend to make better help-side decisions (of course, physical attributes like length and quickness are very important in recovering to/closing out on perimeter shooters too—but that variable is largely held constant in our analysis; while quickness can improve from season to season due to strength and conditioning work, it can also decrease due to injuries/variations in physical well-being; in most cases, though, physical attributes stay fairly constant from year to year). Better decision-making, fewer mental mistakes/lapses in concentration, and improved communication all combine to cause Carolina sophomores to defend 3-pointers much more effectively than they did as rookies.

As sophomores, Tar Heel defenders also commit fewer fouls, force more turnovers (primarily driven by the increase in drawn charges), and get more deflections. When accounting for all of these defensive improvements, Carolina sophomores raise their average Stop% from 56.5 as a freshman to 59.1. Moreover, the average defensive On-Court/Off-Court for a UNC sophomore is +1.3 (interpreted to mean that the team allows 1.3 fewer points per 100 defensive possessions with that player on the court as it does with him on the bench). This is an improvement of 2.4 points per 100 possessions from the average freshman’s defensive On-Court/Off-Court of -1.1. It is not surprising, perhaps, that better individual defense and more stops are associated with an improvement in defensive +/- at the team level.

So we’ve seen that significant defensive improvement occurs between a Tar Heel’s freshman and sophomore seasons. How about between his sophomore and junior campaigns? Table 2 summarizes the data for the 8 UNC players who have freshman-through-junior-year data since 2006 (Frasor, Ginyard, Green, Hansbrough, Ellington, Lawson, Thompson, and Graves). As might be expected, defensive improvement continues into the junior season, but its rate slows. For the 8 players studied in the Table 2 analysis, freshman-to-sophomore Stop% increased by 6.7% (from 55.4 to 59.1) while sophomore-to-junior Stop% increased by just 2.9% (from 59.1 to 60.8). Likewise, the rate of improvement in defensive On-Court/Off-Court also slowed (from 4.8 points per 100 possessions between freshman and sophomore seasons to 1.2 points per possession between sophomore and junior seasons).

Table 2: Freshman-to-Junior Defensive Development—UNC 2006-2010*

There was no evidence, however, that defensive improvements continued into a player’s senior season. At that point, it is possible that mastery of the defensive system has been attained. Since there are no more (or very few) gains to be made from additional experience, the average player’s Stop%/defensive performance as a senior will not deviate much from his junior-year performance.

Of the 8 Tar Heels with junior-to-senior data since 2006, one was gotten significantly better (Wes Miller), two have gotten slightly better (Ginyard, Thompson), 2 have gotten slightly worse (Hansbrough, Reyshawn Terry), and 3 have gotten significantly worse (Frasor, Green, and Quentin Thomas). It is worth noting that many of the seniors whose numbers declined were battling or recovering from injuries during their senior campaign, including Hansbrough, Frasor, and Thomas. Green was also in a new role as a starter. Preliminary evidence has shown that coming off the bench is correlated with a slightly higher Stop% (since it involves a higher percentage of minutes against the opponents’ bench and, in fewer minutes as a reserve, a player is able to exert a higher level of defensive intensity/effort).

Miller, the only senior to significantly improve, went from a starting role as a junior to a reserve role as a senior—experiencing the opposite effect from Green. While more data and a larger sample would help to definitively quantify the impact of the junior-to-senior year of experience, the preliminary evidence is compelling enough to conclude that the rate of defensive improvement essentially flattens out after the junior season. A good general rule of thumb is: 4-7% improvement in Stop% as a sophomore, 2-4% improvement in Stop% as a junior, and no change in Stop% as a senior.

Table 3: Largest Individual Season-to-Season Improvements in Stop Percentage: UNC 2006-2010

What it All Means

As seen in Table 3 and discussed in the previous section, a Carolina player is most likely to have a significant jump in his defensive performance between his freshman and sophomore years. In addition to the three double-digit percentage increases listed in Table 3, Frasor and Green both improved their Stop% by 9% as sophomores. Hansbrough’s increased by 5%. This theoretically boded well for North Carolina’s sophomore class of John Henson, Dexter Strickland, and Leslie McDonald in 2010-11. All three of these defenders possess outstanding physical traits defensively, and all three were average or above-average defensive freshmen (Henson’s Stop% was 56.2, Strickland’s 58.4%, and McDonald’s 62.8% as compared to 56.5% for the average Roy Williams-coached UNC freshman).

Despite having solid rookie years defensively, all three of the rising sophomores were prone to mental mistakes and lapses of concentration on that end of the court. These are precisely the types of errors that an additional year in the system tends to reduce. By combining their prodigious physical gifts with a better mental approach and understanding of the scheme, Henson, Strickland, and McDonald all projected to be plus excellent defenders as sophomores (and beyond). Zeller, who made a significant defensive leap as a sophomore, seems poised to continue his improvement as a junior. Making the transition from reserve to starter figured to hurt his Stop%/defensive results some (a la Danny Green as a senior), but, due to his limited game action as a Heel, he should still have seen a slight uptick in effectiveness as a rising junior.

The results were mixed. Henson experienced great improvement, his stop%  rising to 66.4 for a +10.2 gain. Strickland's rise was marginal, up to 58.5 from 58.4, while McDonald actually dropped significantly; from 62.8 to 52.6. As predicted, Zeller's improvement continued. After rising +11.9 from freshman to sophomore year, he upped his stop% to 65.7 as a junior, another gain of +3.5. Table 4 below shows the full defensive statistics for the 2010-11 team.

Table 4: Defensive Stats, 2010-11

A final point which speaks very clearly to the value of defensive experience in Roy Williams’s system can be delivered by considering the correlation between Stop% and (adjusted) defensive efficiency. Not surprisingly (since all possessions end in either a stop or a score), there is an incredibly tight relationship between a team’s Stop% and its defensive efficiency (r-squared of 0.96). Using the results of a linear regression, we can estimate what UNC’s adjusted defensive efficiency would be assuming an all-freshman, all-sophomore, or all-junior/senior line-up (with the assumption that there’s no defensive improvement between junior and senior years).

Plugging the Stop%’s in Table 2 into this regression, an all-freshman line-up (of Frasor, Ginyard, Green, Hansbrough, Ellington, Lawson, Thompson, and Graves) would be predicted to have a defensive efficiency of 93.4—or 56th in the country in 2010. Based on the sophomore-year Stop%’s of that group of 8 Tar Heels, that number would drop to 84.3 (2nd in the country in 2010). As juniors, those 8 players defended at a level equivalent to a defensive efficiency of 80.7—a mark that would have led the country in 2010. For the given group of players studied in this analysis, a single year of additional experience is worth over 9 points per 100 possessions on the scoreboard. A second year of additional experience in the program (from sophomore to junior) is worth about 3.5 points per 100 possessions.

It’s certainly true that a great defender needs to possess some core physical characteristics and skills. He must also be willing to work hard, do the dirty work, and exert effort and energy on each defensive possession. But in addition to these traits, we shouldn’t overlook the role that experience plays in effective defense. As the numbers in this article confirm, the mental aspect of defensive basketball can be worth several points on the scoreboard each game, and can often be the difference between a win and a loss.

Adrian Atkinson is a contributing writer and the editor of MSP's Tar Heel Tip-off. You can follow him on twitter at @FreeportKid