Volleyball Analytic System

Game Analysis System - A Road Map to Win

Accurately Analyzing Team’s Earning Points Capabilities

Dr. Iradge Ahrabi-Fard

Mark Jacobson

Dr. Mark D. Ecker

University of Northern Iowa

March 24, 2008

Statistics are taken to accurately evaluate individual and team performance during the game of volleyball. Even though the current statistics taking procedure is informative it isnot sufficient. It needs to be moredetailed so that based on the available information coaches could accurately analyze their individual, team, and game plan to plan corrective training content. “Volleyball Analytic System” is a detail data gathering procedure exploring the individual, rotation and team point earning capability against a given opponent, from varieties of court position. Data obtained is open to intricate statistical scrutiny to find tendencies and performance colorations during winning games losing games winning points and losing points. The recommended data gathering system may be cumbersome but the benefits outweigh the costs of the time consuming process.

Rally scoring dictates that every serve is a scoring opportunity with a team that gets to 25/30 points earlier than the opposition with two points difference to win the game. This proposition compelsthat teams accurately assess their point getting capabilities, and improve the likelihood of reaching the end sooner than the opposition. This important point can promote a philosophy that should spend more time stopping the opponent to earn point or concentrate on my team earning point. No matter how well can slow down or postpone the opposition to earn point my team should have the capability of earning point or I have to rely on opponent error to score.In the game of volleyball points are scored from four categories as follows:

Ace serve

Block kill

Attack kill

Opponent’s error

Three of the above items are in a team’s control and one is not. A team can initiate performances to earn point from serve, attack, and block but not from the opponent’s error. In order to accurately analyze a team’s point earning capabilities, a system of data gathering mechanism should be designed to answer all our concerns. The data gathering tool should extract valuable information, suitable for statistical scrutiny towards better understanding of the team performance in a given match.Learning about the capabilities and possibilities of earning point by individual player or team from the set categories that can be controlled is the essential challenge. Volleyball Analytic System (VAS) evaluates every skill performance in light of their scoring capability or creation of opportunity for the team to earn point. Out of the four categories of earning point teams have no predictive and accurate instrument to measure the” opponent’s error”. So, it is advisable to focus on the performances that a team has control over the outcome.For all practical purposes the formula of A (points obtained from attack + B (points obtained from block) + S (points obtained from serve +E (Points from opponent’s error = 25/30 sooner than the opponent grant a win. The formula is:

A+B+S +E = winning or losing

This concept of reaching 25/30 sooner than opposition dictates the emphasis on team’s capability of earning point from A, B, or S. Finding the proportion of points earned from each of the category from the above formula provides the opportunity to calculate the percentage of points earned by each category during winning or losing games. The comparison can provide insight of the potency during winning games and shortcomings during losing games.Since there is no predictable control over earning E, teams should concentrate on individuals’ or team’s skill performances that are initiated and can result in earning point. Therefore, serve, attack, and block are the only predictable point earning method under direct control of the team trying to get point. Attack and serve are skills that performer initiate and dictates the earning of points. Block is a reaction skill, reacting to the means by which the opponent intends to earn point. Teams can count on initiating the efficiency of serve or attack, but their blocking efficiency depends on the reaction to the attack performance of opponent. In other word one may intentionally put their team in a position to earn point from serve or attack, but avoid intentionally put the team in position to earn point from block.

Each categories of A, B or S are evaluated by the actual point earned and by the opportunity provided to the team to earn point called “probable point”. The actual earning points are the actual attack kills, block kill and serve aces or errors.The probable points are when a team is put in a position to get points from an attack. For instancea pass or dig may put the team in full attack capability (in system offense), partial offense capability (out of system offense) or no offense opportunity (free ball or down ball). So the value of each serve or spike is not only the actual value of +1 or -1 point. The effects of each of these skill performances either enable the opposition to have a full attack capability, partial attack capability or no attack capability. The value of each serve and spike should be identified not only with their kill or error but also with their affect on the next performance.

The extend outcome of eachrally is not one point but two points differentiation since from +1 to -1 is 2 points difference. So all the values assigned to the skill performance are from 1 to -1.

Serve

In order to win a game, a team must serve more than the opponent. So developing a good serving team is a necessity not an opulence. Tough serve and position serving should be evaluated in light of their actual point earning or the level ofits empowering or debilitating the opponent to earn point. The following value system is recommended to evaluate the actual and probable affect of serve to earn pointin either side of the net.

Data gathering values of serve is recommended as follows:

Ace serve receives +1 actual points

A serve that forces the opponent to send the ball over the net with no offense receive .25probable points, since the effect of the serve provide an opportunity for a serving team to attack and get point.

A serve that causes out of system attack receives .0probable points. The effect of serve took the offense out of their full power performance.

A weak serve that provides the opposition the opportunity for full organized attack receive -.25probable points for giving the best opportunity to the opponent to earn point.

Serve error receives -1actual point.

Block

Blocking is a team effort to reject the ball from crossing the net and an attempt to earn point. A team intends to organize full blocking power in front of an attack. Organizing one,two or three blockers provide one blocking attempt that can earn, lose or enable either team to organize an attack for earning one point. In this system of data analysis individual blocker does not get evaluated. Teams are supposed to have maximum number of blockers on each attack attempt. If a team is forced to one blocker it is the flaw of their performance.So all block attempts, from one to three blockers can only receive single evaluation. There are two types of blocking affects. Block that prevent the ball from crossing the net or a flawed block that after the touch let the ball cross the net referred to as the “soft block”.

Data gathering considers the following variability of the block:

Blocker(s) left, middle, right

Rotations

Block value that prevents ball crossing the net

Block value of the block that after the touch ball crosses the net

Data gathering of block that prevent the ball cross the net:

Block performance in which attacked ball is not touched will not be evaluated since there could be many immeasurable reasons for or against it.

Blocking effort that prevents the attacked ball cross the net and terminates the rally in favor of the block effort receives 1 actual point.

Block that prevents the ball fromcrossingthe net and causes no organized offense from opponent (free ball over) receives .25 probable point

Block that prevents the ball fromcrossingthe net and result in anykind of in system or out of system attack (a limited offense)receives .0probable point

Blocking effort that prevents the attacked ball cross the net but results in an error by putting the ball outside the court or by blocker(s) touch the net receives -1 point.

Data gathering of soft block that ball crosses the netafter block touch:

A blocking attempt that touches the ball and let it cross the net is failure. In this case floor defense is utilized to either save the rally or empower the team for organize offense.

A block that touches the ball let it cross the net and result in full in system attack capability receives .25 probable points.

Block that touches the ball and let it cross the net resulting in an out of system attack receives 0probable points.

Block that touches the ball and let it cross the net resulting in no organized offense, overpass, or free ball receives -.25probable points.

Block that touches the ball and let it cross the net resulting in an unplayable ball(block error) receives .-1 actual point

Attack

Attack is a skill by which most of the points in a given game are obtained.Each serve has only one opportunity to score or affect the outcome of a rally but block and attack may have several occasions in a single rally to obtain point. As discussed previously, block is an important part of earning point or debilitating the opposition offense effectiveness. But, block is not a skill that teams would like to initiate putting themselves in that situation for earning point. Attack is the most effective and common method of earning point. A team can not rely on winning a game by serve and block as the major contributors of reaching 25/30 points. Therefore empowerment of individual attackers within a system of multiple offense is the most effective way to win a game.

There are four components of attack evaluation:

Individual attacker performance

In system/out of system

Zone of attack

Team rotation capability

Sophistication of the offensive to create full, one, no blocker or flawed and delayed block attempt against the attacker

Data gathering of attack is recommended as follows:

Attack kill receives +1 actual points

Attacks that cross the net and force the opponent to send the ball back with no organized offense receive .25probable points.

An attack that crosses the net and causes an out of system attack from opponent receives .0 probable points.

A weak attack that crosses the net and provides the opposition an opportunity for full organized attack receives-.25 probable points.

Attack error receives -1 actual point

Blocked attacks that are retrieved but have to be free balled over or overpasses receive -.25 probable points.

Blocked attacks that are retrieved for an organized attack receives 0 probable points, since this attempt has failed and the next attempt has to be evaluated for its merit.

All the gathered data are statistically analyzed to discover the relationship between the different variables. The result gives the coach a much better detail understanding of the game outcome. Based on the findings of data analysis, coaches can develop training emphasis by which a team can improve on their point earning efforts.

Data Analysis

The dataset analyzed in this paper consists of 12 matches from a Division I University. The University of Northern Iowa (UNI) women’s volleyball team played these 12 matches from August 24th to October 6th in 2007. UNI had a strong record of success, including the Missouri Valley Conference (tournament) championship and an NCAA tournament appearance during the 2007 season, although only winning 4 out of the first 12 matches during this early part of the 2007 season. (UNI had a record of 9 wins and 9 losses through the early part of the season, but complete data was only available for 12 of those 18 matches). Seven of the matches went to 3 games and 5 of the matches lasted for 5 games.UNI won 4 matches 3-0 and lost 3 matches 0-3. UNI lost 5 matches 2-3. This gave a total of 46 volleyball games in the data set, where overall, UNI won 22 games and lost 24 games.

The variables used in this study include the ratio of the kills to the points score for each game for UNI(AkPts); the game of the match (1, 2, 3, 4 or 5); whether it was a home or an away game for UNI; a binary variable for game5 (since scoring only goes to 15 points instead of 30 points for game 5); won or lost that game for UNI; kills by UNI (K); Attack Errors by UNI(AE); attack attempts by UNI (TA); attack percentagefor UNI (PctAttack); UNI service aces(UNISA); UNI service errors(UNISE); opponent’s service aces(oppSA); opponent’s service errors(oppSE); UNI blocking errors(BE); UNI ball handling errors(BHE); LeadChanges; Points scored by UNI (UNIPts); Points scored by opponent (oppPts); probable attack points (ProbAttackPts); and the points difference of UNI points score for game – opponent’s points score for game(PointsDiff).

One comparisonof interest isthe attack percentage (to UNI points) percentage for the 22 games that UNI woncompared to the attack percentage to UNI points for the 24 games that UNI lost. The ratio (attack percentage to UNI points) is required to account for fifth games that had only a winning score of 15 points; otherwise, we would only use attack points for games ones through four. The formal hypothesis test to compare the mean percentage of attack points per UNI point is a two sample t-test. The mean percentage of UNI attack points (per UNI point) in winning games is 0.5701 compared to 0.6611 in games where UNI lost. Surprisingly, UNI is not attacking as well in games that they won (compared to games that they had lost). We assume equal population variances after noting that one sample variance is not more than twice the other. We find no formal statistical difference (at the alpha=0.05 level) between the mean percent of attack points per UNI point for games in which UNI won compared to games in which UNI lost.

Another statistical comparison involved comparing the percentage of opponent’s errors caused by UNI, again accounting for fifth games where the total winning score is 15 winning points, in games UNI won to the opponent’s errors per total points in games UNI lost. Again a two sample t-test is the formal statistical procedure; however, the unequal population variance t-test is needed due to one sample variance being more than twice the other. We formally find (p-value) that there is no statistical difference between opponents errors in games won versus lost by UNI. If we eliminate the fifth games, then the p-value in comparing the means in games won versus lost drops to 0.044; primarily due to the large variability in opponent’s errors in the three game 5 losses.

We also compare the Probable Attack points in the games where UNI was at home to games where UNI was away. In home games, UNI averages 13.9 Probable attack points compared to 9.4 for away games. We again use a two sample t-test with equal population variances to formally conclude that there is a very strong statistical difference in Probable Attack Points in home versus away games (p-value = 0.0001).

A logistic regression (Hosmer and Lemeshow, 2000) models which factors most influence UNI’s chances to win a volleyball game. In particular, the probability of winning is the dependent variable while covariates including the Attack Probable points, number of kills, service aces, service errors, opponents service aces, opponents service errors, block errors, BHE, AE, TA, and whether the game was at home or not. The only factors that correlate with UNI’s chance of winning (at the 0.05 significance level) were whether the game was at home (p-value 0.045) and Probable Attack points (p-value = 0.09). Interpreting the results, UNI’s chances of wining are dramatically increased when at home; they are 27 times more likely to win at home, given the same exact values of the covariates, compared to being on the road. The more Probable Attack points, the higher the chance of winning. In particular, UNI is 1.67 times more likely to win for every extra Probable Attack point earned.

Reference

Hosmer, D. and S. Lemeshow. (2000). Applied Logistic Regression, Second Edition, Wiley.

1