Summary of Florida-Nebraska Match

This past weekend saw the opening of women's college play in the U.S., with the nationally televised (on CBS College Sports cable channel) Runza/AVCA Showcase from Omaha, Nebraska taking center stage. Each match featured a Big 12 school (either Nebraska or Iowa State) taking on an SEC school (Florida or Kentucky). As it turned out, tournament organizers saved the best for last, as yesterday's closing match between Florida and Nebraska came down to an exciting finish, with the Gators prevailing 15-12 in the fifth (boxscore).

For this match, I created the two figures below (one for each team), which convey two aspects of offensive attack: players' hitting percentages (on the vertical axis) and number of hitting attempts (horizontal axis). Players are arranged left-to-right in descending order of hitting percentage. You may click on the figures to enlarge them.

The ideal would be to have rectangles that were both tall and wide, indicating that a player maintained a high hitting percentage over a large number of attempts. The Gators' Kelly Murphy epitomized this combination. Undesirable shapes are tall-and-thin (a player who hits well, but gets few attemps) and short-and-wide (a player who hits for a low percentage, but gets a lot of attempts). I've depicted these suboptimal situations in the figures with paler shades of red and blue.

Whether volleyball coaches and analysts find these graphs useful remains to be seen. One immediate application of the graphs could be in goal-setting. Players with large numbers of attempts but low hitting percentages could be shown the graph, with the coach setting the goal of some specific, higher hitting percentage for the player to work towards.

Alexis Lebedew on Evaluating Setters

I recently received an e-mail from Alexis Lebedew of the Australian Institute of Sport, bringing to my attention some of his writings. Lebedew's focus is the evaluation of setting, a skill that has gone relatively unanalyzed over the years. The statistic of a setting "assist" exists, but because it represents the number of balls leading to kills, it overlaps considerably with hitting statistics.

In a piece entitled, "A Reconceptualisation of Traditional Volleyball Statistics to Provide a Coaching Tool for Setting" (link), Lebedew proposes a way to rate the quality of sets by taking into account not just the spike attempt following the set, but also the pass preceding the set. In short, setters are most rewarded for making "lemonade" from a "lemon" pass. As Lebedew states more technically, "...the combination of a [high-quality] spike and a [poor] pass has the top Rating... within the ‘Excellent’ outcome."

In fact, sets can be graded on a scale of 0-12, based on combinations of quality ratings for pass and spike. Lebedew notes that coaches who are used to grading passing and hitting performances on a metric different from his own (e.g., rating hit attempts on a 3- rather than 4-point scale) will still be able to construct a meaningful scale for setting, although the top value may differ from 12.

Lebedew also attempted to validate his setting metric in two ways. He first showed that computer software designed to link passes and hit attempts within the same sequences to derive set attempts only rarely missed a set attempt when compared to video footage. Second, he charted teams' percentages of sets (games) won for different averages of setting proficiency. For example, teams won roughly 95% of time when their set quality averaged 9 or higher, roughly 90% of the time when it averaged 8.5 or higher, etc., down through roughly 55% when averaging 6 or higher on setting. Lebedew encourages coaches and setters to strive for setting-proficiency averages of around 7.5-8.

All of the data were from international beach volleyball, which qualifies the generalizability of the findings in some important ways. With two-person teams, of course, there's no way to assess the setter's savviness in choosing which hitting-eligible teammate to set (as noted by Lebedew). Also, at levels of play beneath international caliber, more realistic setting-proficiency aspirations than the aforementioned 7.5-8 may need to be established.

JQAS Article on Quality of Skill Performance and Winning Points

A recent issue of the Journal of Quantitative Analysis in Sports(Volume 6, Issue 2) contained an article by Michelle Miskin, Gilbert Fellingham, and Lindsay Florence entitled "Skill Importance in Women’s Volleyball." Access to articles is by subscription, but the journal has guest-visitor privileges for single articles.

Miskin and colleagues analyzed data for a particular women's Division I team (not identified by name) during the 2006 season. When the team played at home, play on its side of the net was videotaped and later coded. Serves, passes, and digs were rated by judges on quantitative scales (e.g., 0-to-5), sets were evaluated in terms of their distance from the net, and spike attempts were coded by area of the court from where they were hit.

Essentially, the authors appear to be looking at correlations (or associations) between characteristics and quality of skill performance, and likelihood of winning the point. As they state on page 2:

The importance score incorporates not only the impact of a specific skill..., but also the uncertainty associated with the performance... Thus, a skill whose association with scoring a point is less certain will be penalized when using this metric when compared to a skill where performance at a given level is more closely associated with a positive outcome.

The article throws a barrage of statistical terms at the reader (e.g., Bayesian analysis, Markov Chains, Dirichlet prior, Gibbs sampling, gamma distributions), some of which I was familiar with, but many of them not. Fortunately, the authors translated the complex statistical results into plain English recommendations for the team that was investigated:

1. Keep sets and passes away from the net.

2. Force the attack to the middle and right side if at all possible.

3. Devote a considerable proportion of practice time to transition offense.

4. Get to blocking positions more quickly following a serve.

Presumably, if a team wanted to apply the analytic tools described in the article in their full glory, it would need to hire a pretty high-powered statistical consultant (in addition to acquiring the videotaping and coding resources). Perhaps similar analyses could be done via more basic correlational and regression techniques, but I suspect that the resulting conclusions may be somewhat imprecise, compared to those from the fully sophisticated analyses.

Lawson Powers Stanford to NCAA Men's Title

Stanford's Brad Lawson had an incredible offensive night as the Cardinal blew out Penn State for the NCAA title, 30-25, 30-20, 30-18. Lawson, a 6-foot-7 sophomore outside hitter who was one of four players from the state of Hawaii to take the court for Stanford tonight, compiled the following line: 24 kills with only 1 hitting error, in 28 attempts, for a remarkable .821 percentage (box score). For those who don't follow volleyball closely, a hitting percentage in the .300's would be considered very good and in the .400's, outstanding. For the season (including the championship match), Lawson hit .387 (522 kills and 143 errors on 980 attempts).

This NCAA men's volleyball records page (current only through 2006) presents two championship records, for a single match and for both games of a tournament combined:

HITTING PERCENTAGE, MATCH (MIN. 15 ATTEMPTS)
.867--Jeff Nygaard, UCLA (3) vs. Ohio St. (0), 5-7-93.

HITTING PERCENTAGE, TOURNAMENT (MIN. 20 ATTEMPTS)
.788--Rick Tune, Pepperdine, 1998 (.833 vs. Princeton, 10-0/12; .762 vs. UCLA, 17-1/21).

Nygaard's record, based on a 13-0-15 line, was achieved in a semifinal match, arguably making it slightly less impressive than a comparable hitting percentage in a championship match. Also, Nygaard was a middle blocker, as was Tune.

A couple of other notes:

*Interestingly, this year Stanford also saw an .800 hitting performance on the opposite side of the net. On February 19, Pepperdine's Cory Riecks recorded a 17-1-20 night against the Cardinal.

*As a follow-up to yesterday's posting (immediately below), Penn State managed only 4.5 total team blocks against Stanford in the title match.

Preview of Penn State-Stanford NCAA Men's Final

In anticipation of tomorrow night's (7:00 Eastern) NCAA men's championship match between Penn State and Stanford, the Nittany Lion athletic department has put out a press release that includes some interesting statistical facts.

According to the release, Penn State is 21-6 when two or more players record double-digit kills, 7-3 when two or more players record double-digit digs, and 13-2 when achieving 10 or more blocks (among other things). Such statistics can potentially provide useful insights in assessing a team's chances of winning a particular match. However, caution should be exercised for a few reasons. Before I go any further in my comments, though, I want to state that I am thrilled any time I see statistically oriented writing in the coverage of volleyball and that I intend my remarks in a constructive spirit.

First, the presented statistics do not make use of all the known information. Using the last statistic given above, the Nittany Lions are 13-2 when getting 10 or more blocks. What is their record when getting fewer than 10 blocks? As shown below, we can fill out the picture by knowing that the team's overall record is 24-7.


With all of the cells filled in, we can thus see that Penn State has a pretty good record, too, when getting fewer than 10 blocks.

Second, there is considerable variation in the quality of Penn State's opposition during the season. Playing other eastern (or midwestern) schools presumably is not as difficult as going against the traditional Mountain Pacific Sports Federation powers the Nittany Lions faced during the regular season (USC, Hawai'i, UC Irvine, BYU, Cal State Long Beach, UC Santa Barbara, and Cal State Northridge). As accessed from Penn State's game-by-game log, here are the Nittany Lions' total blocks (in red) and match outcomes against MPSF opponents:

USC 4 L(0-3)
Hawai'i 16 W(3-2)
UCI 9.5 W(3-2)
BYU 6 L(1-3)
CSULB 5 W(3-0)
UCSB 6 L(3-0)
CSUN 8.5 L(3-0)

As shown, only once in these seven matches did Penn State achieve 10 or more blocks, and it happened in a five-game match, which provides more opportunity to accumulate blocks (and other statistical markers). Blocks per game might be more appropriate to cite.

Of course, though, in a stunning turnaround from Penn State's 0-3 loss at Cal State Northridge on April 10, the Nittany Lions turned things around on Northridge in last night's NCAA semifinals, winning 3-0 on the strength of 11 blocks. Eleven blocks in three games yields a robust 3.67 average. Penn State's opponent in the championship game, Stanford, piled up 12 blocks in a three-game sweep over Ohio State.

Lastly, as is drilled into the heads of all students taking social-science research methodology courses, correlation (i.e., that two things co-occur) does not by itself prove that one thing has actually caused the other. In basketball, for example, one might find that when a given team makes under 10% of its three-point attempts, it loses the game a high percentage of the time. One might intuitively interpret such a statistic to mean that poor shooting caused the team to lose. However, the team could have been trailing in some of its games for reasons having little to do with three-point shooting and then taken a lot of desperation threes (that were missed) in an attempt to get back into those games. In other words, it may have been the losing that caused the missed shots from behind the arc. Similar examples for football are discussed here.

I could envision a volleyball scenario where a team would have a poor won-loss record in matches in which it committed a large number of service errors. Maybe the service errors cost the team a lot of points early and paved the way to eventual defeat. But, it could be that the team fell behind for reasons unrelated to serving and then decided to serve aggressively in an attempt to catch up, only to have the high-risk/high-yield serves mostly fail. Something to think about.

Chuck Rey's Analysis of Penn State-Texas NCAA Women's Final

Prompted by a re-airing of last December's NCAA women's final between Penn State and Texas, fellow volleyball blogger Chuck Rey has produced a statistical analysis of the match. Interested readers can compare and contrast Coach Rey's analysis to the one that I did.

Karch Kiraly's Hypothesis on "Better" Kind of Hitting Error

There are two main types of hitting error. According to the NCAA volleyball statistical manual, one type of error involves hitting the ball somewhere other than in-bounds on the opponent's side of the court (i.e., "Hits the ball out of bounds" or "Hits the ball into the net resulting in a four-hit violation"). The other major type of hitting error is when the attacker is stuff-blocked (where the ball is "blocked down by the opposition to the same side as the attacker, and cannot be kept in play as a direct result of the block"). There are additional types of attack error such as the hitter contacting the net, back-row attack violations, and “thrown”/double-hit balls; the present analysis is not all that concerned with this last set of errors, however.

Of the two main types of error -- failing to hit the ball in bounds, and getting blocked -- ESPN commentator and former UCLA and Olympic great Karch Kiraly feels that one of these types of error is more encouraging for the team committing the faux pas and the other, more discouraging. During the Hawaii-Penn State NCAA women's semifinal broadcast about a month ago, Kiraly referred to "the better kind [of hitting error], getting blocked" (the remark occurs at roughly the 4:00 mark on this video).

In other words, given that you're inevitably going to commit errors, it's better to do so by getting blocked than by hitting out of bounds (or into the net), according to this line of argument. As the writer on College Volleyball Coach.com puts it, "I rarely get upset with my hitters when they are blocked, because I view it as the responsibility of the rest of the team to cover the hitter…" Stated differently, when an attempted spike gets blocked back toward the hitting team, the latter possibly can dig the ball back up and run another play. On the other hand, a spike attempt hit long or wide loses the attacking team the point immediately, with no chance to recover. Further, spraying a lot of balls out of bounds may indicate general shoddiness on the part of a team.

Getting blocked a lot wouldn't necessarily reflect positively on a team's offensive attack, either, though. It could mean a team's hitters and setters aren't all that adept at reading the block or taking steps to counteract it (e.g., "tooling" the block). It could also mean that a team's player's are not especially quick at getting to a ball that's been blocked back (i.e., covering the hitter).

Ultimately, it's an empirical question whether teams that commit larger shares of their hitting errors by getting blocked do better over the long run than do teams that commit a smaller share of their errors that way. We'll call this proposition the Kiraly Hypothesis. In testing the hypothesis, it's important to hold constant the number of errors made by a team, in order to focus exclusively on the type of errors.

To address the Kiraly Hypothesis (using the 2009 women's season), I first examined the final statistics for several conferences to find two (or more) teams in the same league that committed the same number (or virtually the same number) of hitting errors. One example, shown in the table below, is that Cal and Arizona St. each committed 368 hitting errors in Pac 10 competition (in all cases, only conference play is used). By going through a team's box scores, I could record the opponent's total blocks for a given focal team (e.g., when Cal played Stanford, I would record the Cardinal's total number of blocks, which indicate how often the Golden Bears got blocked; when Cal played Oregon, I would record the Ducks' total blocks; etc.). I did this for each of a focal team's conference matches and summed up the total number of times a focal team got blocked. If one wanted to, one could subtract this number from a team's total number of hitting errors in conference, to arrive at how often the focal team hit the ball out of bounds or into the net, touched the net, violated the back-row-attacker rule, etc. The latter step is not necessary for present purposes, however.

What we're left with is six paired comparisons (plus one trio). Within each pair (or trio), the teams committed essentially the same total number of hitting errors in conference play. One of the teams in a given pair would generally have a greater number of "got blocked" errors than the other. I then checked whether the team with the greater number of "got blocked" errors finished higher in the conference standings than did the team with fewer "got blocked" errors. That's what the Kiraly Hypothesis would imply, to me at least. Let's look at the following chart (which you can click on to enlarge):


As described above, Cal and ASU were paired together because they each amassed the same number of total hitting errors. Cal got blocked (i.e., erred in the "better" way) 160.5 times, whereas ASU got blocked only 151.5 times. By this logic, the Golden Bears should have finished higher in the Pac 10 standings than did the Sun Devils and, indeed, this is what happened. Score one for the Kiraly Hypothesis. (As an aside, I cannot understand how a team -- as opposed to an individual player -- can finish a match or a season with a total ending in half-blocks.)

In the fourth comparison, I grouped Utah St. and Nevada together, due to their similarity in number of times getting blocked and in their WAC win-loss records, and compared them jointly to Louisiana Tech. In all, four comparisons supported the Kiraly Hypothesis -- the team that more frequently got blocked during conference play also finished higher in the standings than its comparison school (note that the colors in columns two and three are the same in these cases).

As also shown in the chart, three comparisons contradicted the Kiraly Hypothesis. New Mexico St. is an interesting team. The Aggies finished 13-3, second in the WAC standings only to Hawai'i. Yet, despite compiling one of the highest hitting-error totals (363) in the WAC, NMSU didn't get blocked very much (see the box scores of the Aggies' matches hosting San Jose St., blocked 1 time; hosting Boise St., blocked 3 times; and hosting Nevada, blocked 3 times; NMSU additionally had three matches in which it was blocked 4 times).

Seven comparisons, of course, do not make for a very large sample. It is possible that if, say, 50 or 100 comparisons were done, the Kiraly Hypothesis would receive greater support. However, it is time consuming to compile the data for these comparisons. The most that can be said at this point is that the Kiraly Hypothesis is not so overpowering a phenomenon that its existence can be documented with only a small number of tests.