Wednesday, October 24, 2012

Hitting Allocation Graph from Last Friday's Stanford-Washington Match

Below is a hitting allocation chart I made for Game 2 of last Friday night's Stanford-Washington match. The Cardinal took both this particular game/set and the match as a whole, 10-25, 28-26, 10-25, 26-24, 15-7. The chart shows which players took hitting attempts off of serve-receipt, from where on the court, at what angle, and with what result. An  introduction to the notation is available from this earlier posting in which I introduced the chart. One new piece of terminology today is that an offensive "reboot" is when a spike attempt is blocked back to the attacking team, which then starts over. For this chart, I tried harder to catch the names of the specific players taking each swing, but I was not always successful. You may click on the graphic to enlarge it.

I don't think there are really any big surprises here. Stanford went heavily to middle-blocker Carly Wopat, both in the middle and on the slide play to the right, and she produced several kills. Washington often called the number of outside hitter Krista Vansant, with mixed results.

The Huskies go down to Los Angeles this weekend to face USC (Friday) and UCLA (Sunday) in a pair of marquee match-ups.

Friday, October 19, 2012

Passing Well and Setting the Middle: The Texas Tech Internal Data

Today, for my second analysis using Texas Tech internal team data, I look at the relationship between quality of the team's passes on serve receipt, on the one hand, and the location and success of the resulting hit attempts, on the other. Again, my thanks to head coach Don Flora and assistant Jojit Coronel for their willingness to share the data and answer any questions I have. (Here's a link to my first analysis of the Texas Tech data, which focused on side-out rates in different rotations.)

The better the pass a team can make on serve receipt, the easier it will be for the setter to get to the ball and, hence, the better should be the set. A good set should then increase the hitter's likelihood of achieving a kill. A further objective for many teams is to set the ball for the middle hitter, to quicken the offense. Other common plays involve high-arching sets to the outside, which give the other team time to get their blockers in place.

In general, teams collect more specific or "micro-level" data than what are available in published box scores. One form of micro data is an evaluative rating of each pass, made by a coach or other observer who reviews video of a match. On serve-receipt, pass quality can range from 0 (being aced) to 3 (an excellent pass that gives the setter the full range of options for feeding a hitter in an advantageous position).

In the following graph (on which you can click to enlarge), the left-hand column has headings for pass quality 3, 2, and 1. For each level of pass quality, you can read horizontally to see Texas Tech's distribution of hit attempts taken from the outside (left), middle, and right. (The hit attempts were summed for the Red Raiders' first eight matches of the season.) For example, on passes of quality 3, Texas Tech went to the left 11.6% of the time, to the middle 68.8% of the time, and to the right 19.5% of the time.

As Coach Coronel clarified for me, the classification of left, middle, and right refers to the location where the attack took place, not to a player's position as listed on the roster. Thus, if an outside (left) hitter migrated into the middle to attack a ball, the play would be designated as a middle attack. Another example is a slide play, in which a middle hitter runs toward the right antenna to elude the opposing middle blocker, and hits from there.

The graph confirms that Texas Tech was incrementally more likely to set the middle with increasing quality of pass: roughly 10% of the time on a 1-quality pass, 43% of the time on a 2-quality pass, and 69% of the time on a 3-quality pass. For those with some statistical background, a 3 X 3 chi-square test on the raw frequencies indicated that the distributions of spike attempts into left, middle, and right were significantly different for the varying levels of pass quality (X2 = 83.6, p < .001).

Also included in the information I received were Texas Tech's hitting percentages from each of the Red Raiders' first eight matches, broken down by pass quality on serve-receipt. (The cryptic column headings, such as "hp_psql1," refer to hitting percentage pass quality 1 or whatever number.) These results are presented in the following table.

When Texas Tech's initial pass of the opponent's serve was of the lowest quality (1), the Red Raiders' average hitting percentage across the eight matches was essentially zero. In four of the matches, the team's hitting percentage was negative, indicating more hitting errors than kills. One thing that prevented the team's average hitting percentage on poor-quality passes from ending up markedly negative was a positive .75 hitting percentage on quality-1 passes vs. Binghamton; however, the .75 value was based on only four hitting attempts.

On medium-quality passes (2), the Red Raiders averaged a .232 hitting percentage, and on their best passes (3), their hitting percentage averaged .326. Analysis of Variance (ANOVA) showed that the linear trend of increasing average hitting percentage with better passing was statistically significant (F = 8.70, p < .05).

Some volleyball analysts offer the analogy between volleyball hitting percentages and baseball batting averages, arguing that .300 signifies high-quality performance in either sport. If one accepts this notion, then Texas Tech seemingly needs to receive serve with a passing proficiency close to 3, in order to have a good chance of hitting .300.

It should be noted that Texas Tech won all eight of the matches that formed the basis for the present analyses. Big 12 conference play has proven to be a more challenging proposition. I hope to be able to analyze statistics from conference play at some time in the future.

Tuesday, October 16, 2012

New Kind of Hitting Allocation Charts

This past weekend featured as much televised women's collegiate volleyball as I can remember seeing over a similar period. A big reason is the new Pac 12 Network and its extensive coverage of volleyball, but ESPN 2 and ESPN-U also played a part. With so many matches available, I decided to test a new type of chart to track teams' hitting allocations off of serve receipt, including the locations and angles of the spike attempts. I did something similar for a UCLA-Texas match in the 2010 NCAA women's tournament, based on video coverage from the Longhorns' website using an "end-zone" camera. Here's a link to that previous analysis.

Traditional television coverage uses a sideline camera (for the most part), however, so I developed a new graphing approach using this perspective. One benefit of mapping the locations and angles of spike attempts is that doing so provides richer information than a typical box score. For example, a box score would list how many attempts, kills, and errors were recorded by a middle-blocker. However, there would be no way to distinguish spikes taken by middle blockers from the actual middle of the front row from those taken on slide plays, on which the middle blocker runs over to the far-right side to evade his or her blocker and attacks from there.

I graphed four different games/sets from this past weekend, from three different matches. Let's take a look first at Game 4 from Stanford's five-game win over UCLA last Friday night (you may click on the graphics to enlarge them).

I think most of the terms should be self-explanatory. Just to clarify a few that may be unclear, "In Play" refers to an attack that remained in play due either to being blocked back to the attacking team, which could then start over, or being dug by the defensive side. All plays refer to hard-hit spike attempts, unless the attack is described as a "Tip." All attacks were categorized according to the location from which the ball was attacked, not necessarily the position with which a given player is identified (e.g., the aforementioned slide attack, which is undertaken by a middle hitter, is listed as emanating from the right-hand side).

As can be seen, the teams showed contrasting approaches. UCLA focused on the "pins" (the right and left antennae at the ends of the net), whereas Stanford stormed the middle. For the Bruins, Tabi Love went to town with cross-court spikes from the left end, drawing the exclamation noted in the chart. Shoring up the right side was Kelly Reeves. For Stanford, Carly Wopat and Inky Ajanaku recorded kills in the middle; it also looked to me like Rachel Williams hit some balls in the middle, but she is officially listed as an outside hitter (I could be wrong).

Before looking at the remaining charts, I want to mention four limitations in these depictions:

1. They only show the attacks launched upon serve-receipt, not any attacks in transition.

2. For the most part, they present only team totals for a given avenue of attack, rather than statistics for individual players. I do have a few exceptions in which individual hitters are identified. However, the speed of the game makes it difficult to identify particular players in many instances.

3. As long as the television broadcasts stuck to sideline views, my charts should be highly accurate. However, when a different angle (e.g., end-zone) was presented, I faced a spatial-rotation challenge to convert the shown angle to the sideline-angle format of the charts.

4. There were a few plays (noted in the fine print) that I just flat-out missed, for example, because I was still writing down the information for one play while the next serve was being struck.

Next, let's look at Game 2 of USC's three-game sweep of Cal across the Bay. The Golden Bears enjoyed success on slide plays featuring Correy Johnson, but little else. USC went several times to outside hitter Samantha Bricio, but the Mexican frosh had difficulty on the evening, hitting only .167 (she also struggled two nights earlier against Stanford, hitting -.043).

Lastly, I looked at two games (1 and 2) from Sunday's Minnesota-Nebraska match. One game may or may not provide a valid microcosm of an entire match, so I decided to look at two games in this instance. The Golden Gophers took the first game, but the Cornhuskers roared back to take the next three.

In Game 1, the Gophers amassed 17 kills, which were relatively evenly distributed among five players: OH Katherine Harms 5, MB Tori Dixon 4, OH Ashley Wittman 3, OH Daly Santana 3, and MB Dana Knudsen 2 (play-by-play sheet). As shown below, eight of those Minnesota kills occurred immediately off of serve-receipt. Five of the eight stemmed from slide plays or quick sets down the middle.

In winning Game 2, Nebraska exhibited a variety of ways to produce kills on serve-receipt: a couple of cross-court spikes from the outside hitters, a couple of slides, a tip here and there. The Huskers were also aided by four Gopher service errors.

In conclusion, I think these charts offer a lot of information at a glance, even with their imperfections (e.g., difficulty in identifying particular players from television broadcasts). If I get a chance, I will try to map the action from watching a match in person. If archived videos are available, analysts can go back to enhance the detail provided in the chart.

Thursday, October 4, 2012

Controversy Over NCAA Rule Change on No. of Subs

On this week's episode of the Internet-radio show The Net Live, a spirited discussion broke out on the merits of this year's new NCAA women's rule expanding the number of substitutions from 12 to 15 per game/set (see link to the archived broadcasts in the right-hand column). Substitution policy clearly has analytic implications, thus making it a worthy topic for VolleyMetrics; I provide no statistics in this write-up.

This July 31 article focusing on the Nebraska program provides a lot of background perspective. A key impetus for the rule change appears to be the opportunity to get more players into matches. In terms of volleyball training, the substitution issue raises questions of player specialization vs. well-roundedness.  As the article notes:

With 15 substitutions, coaches will likely not have to worry about reaching their limit and can take out their top hitters when it is time to rotate to the back row, replacing them with passing and serving specialists...

“There are very few kids that are 6-foot-5 and athletic,” [Nebraska coach John] Cook said. “There are only so many to go around, and the top programs are going to get those. They won't have to worry about training them back row. There will be no equalizer when they have to go to the back row.

Of concern to some is how a greater trend toward specialization at the college level will impact the U.S. Olympic program. According to the article, "International rules allow only six subs per set and only one per set at each position, which means that players with all-around skills are preferred over elite specialists."

Some observers, such as UCLA athletic administrator and Net Live contributor Mike Sondheimer, believe having the number of substitutions at 15 facilitates running a 6-2 (two-setter) offense. Such an offense is potentially advantageous in that, by always having a setter in the back row (guaranteed by placing the two of them three positions apart in the rotation), the team can always have three hitters in the front row. The back-row setter thus runs up into the front row to set, once the ball is served (the only restriction is that a player originating in the back row cannot attack in front of the 10-foot line). In a 5-1 (one-setter) offense, half the time the setter is in the front row, leaving only two hitters.

The potential downside of a 6-2 is that, in addition to one setter always being in the back row, one also is always in the front row, where she or he can hit and block. Setters tend to be shorter than other players, however, potentially limiting their effectiveness as hitters and blockers. Bringing the discussion back to the substitution rule, my interpretation of Sondheimer's statement is the 15 opportunities make it easier to replace an undersized setter when she or he rotates into the front row, with a tall hitter/blocker specialist. 

Tuesday, October 2, 2012

Success of Texas Blocking Line-Ups at Texas Tech

I attended last Saturday afternoon's University of Texas at Texas Tech match. The Longhorns won 3-0, but the Red Raiders were very competitive in two of the games/sets, as reflected in the 25-16, 25-23, 25-22 score. My focus was on the Longhorns' blocking -- how they lined up in their rotations and with which combinations were they able to stuff the Red Raiders for points.

Texas's front lines in their six rotations (A through F), depicted schematically with the players' faces to the net, are shown in the graphic below. You may click on the graphic to enlarge it. The figure pertains only to Game 2, in which the Longhorns achieved 6 of their total 14 team blocks in the match (Texas Tech had only 3 team blocks total in the match). The Texas uniform numbers in the figure correspond to actual players, as follows:
  • 1 Kat Bell (MB, 6-1, soph)
  • 5 Molly McCage (MB, 6-3, frosh)
  • 10 Haley Eckerman (OH, 6-3, soph)
  • 12 Hannah Allison (S, 5-11, junior)
  • 14 Sha'Dare McNeal (RS, 6-1, senior)
  • 23 Bailey Webster (OH, 6-3, junior)

As can be seen, the Longhorns earned two points each on blocking in rotations A, C, and F. The blue arrows illustrate how pairs of players came together to block (i.e., did an outside player move toward the center to join the middle blocker, or did the middle blocker move outside?). As always, I tried to reconcile the notes I took at the match with what was reported in the play-by-play sheet. Due to one discrepancy, I could not pinpoint the location of one block. Recall that blocks are credited in the statistics only when immediately resulting in a point.

Also, as noted in the AVCA PowerPoint on keeping statistics (see links in right-hand column): " does NOT matter which player touches the ball – if 2-3 players go up for a block and one player touches it, each receives a block assist." I was nevertheless interested in which player's arm actually blocked the ball. In the figure, therefore, I listed the player first who actually touched the ball, then said "(with)" to refer to the additional player credited with the block (see asterisk in Rotation A as an example).

Obviously, the number of blocks in this exercise is small. However, with data from several matches, coaches could study their opponents' blocking success by rotation and location (left, center, and right) and instruct their teams to hit away from their opponents' advantageous blocking areas.


In other weekend action, Washington moved to 13-0 on the season with a 25-23, 26-24, 25-21 victory over USC on Friday night. If a match can be a tight three-game sweep, this was one of them. The box score doesn't leave a lot of clues as to why the match was so close, at least as far as I can tell. The Huskies outperformed the Trojans in hitting (.239-.221), blocking (11-4), and serving for aces (8-1, each team had 12 service errors). 'SC outdug U-Dub 43-29, but that would be reflected in the team hitting percentages (i.e., spiking a ball that is dug increases the hitting team's number of attempts, while depriving it of a kill). The Huskies' Krista Vansant continued her hot hitting, registering a .370 percentage on 12 kills (with only 2 errors) in 27 swings.

Last Saturday's Penn State at Minnesota contest featured a battle of coaching titans. Russ Rose has coached the Nittany Lions to five NCAA titles, whereas new Golden Gophers' coach Hugh McCutcheon has guided medal-winning U.S. teams in the last two Olympiad (the men to gold in 2008 and the women to silver in 2012). Not only that; both teams came in ranked in the top 10 nationally (PSU No. 1 and Minnesota No. 10). Though the match might have looked good on paper, in the end it was a rout, Penn State prevailing 25-23, 25-8, 25-20. In Game 2, the Gophers sided-out at a 29% clip (7-24), which is incredibly low for as good a team as Minnesota. In the same game, the Lions sided-out with 88% proficiency (8-9). For the match overall, Penn State dominated the hitting (.404-.098). The box score can be accessed here.