VOLLEYBALL STATISTICS: A NEW APPROACH
by Dr. L. R.(Rod) Schall
Men's USA 1984 Olympic Volleyball Team Statistician
Graceland College Men's Volleyball Coach (1967-1993)
INTRODUCTION: A three year study was conducted for the purpose of gathering statistical data at several international volleyball events culminating at the 1984 Olympics. This study was funded in part by the United States Volleyball Educational Foundation, 1982-1984. The primary purpose of keeping statistics of any sport is to provide the media with data about players and teams in order to evaluate their performance in general terms. More interest has been given to statistical systems that will allow a coach to critically evaluate players for the purpose of making line up adjustments and specializing players according to particular skills and abilities, and to use as a diagnostic tool to help players improve.
The main thrust of this study was to evaluate a new approach in keeping volleyball statistics. The system being studied was developed by the author in the early 1970's and converted for the microcomputer in 1977 where it was first used live in a men's match between the USA and Mexico in Moscow, USSR.
The system was studied in relationship to seven major ingredients of a good statistical system. These include: (1) it must be as objective as possible by providing similar results between different statisticians; (2) it must provide a rapid and accurate method of handling large amounts of data quickly; (3) it must show differentiation well between players in each skill; (4) it must be able to evaluate players as a whole across all skills for all around comparableness; (5) it must be comprehensive with total data, not a small sample; (6) it must correlate positively with success; and (7) the rating scales used should have values that are related to scoring. Another objective was to find out which skills correlate most with success.
The first year of the study included a tour with the USA Men's National Team as they played five matches with the Korean National Team in 1982. It also included a match with the Poland National Team. In the second year, data were gathered at the 1983 NORCECA Championships, and the Men's Pre-Olympic Tournament. The 1984 Olympics provided the event for the third year. Funds for this study were made available in part from the United States Volleyball Educational Foundation.
THE STATISTICAL SYSTEM: The volleyball statistical system gathers data in the six basic skills of volleyball; serve, pass (receiving the serve), dig (balls received from the opponent excluding the serve), set, attack, and block. Each player as he plays the ball in one of these skills is rated on a scale from zero to five. For the purposes of this writing, a scoring play refers to termination plays when either a point is scored or a side out is awarded. These ratings are made using the following scale of values: *(See note at end of paper.)
5 - service ace, ball not controlled, immediate score.
3 - aggressive serve that results in no attack.
2 - serve that no quick attack can be made.
1 - serve that is passed well for multiple attack.
0 - serving error.
PASS AND DIG
5 - perfect pass for multiple attack.
3 - pass that can be set for non-quick attack.
1 - pass that results in no attack advantage.
0 - passing error.
5 - perfect set when spike is killed.
4 - perfect set but spike is not killed.
3 - set that is not ideal but can be spiked.
1 - set that cannot be spiked.
0 - setting error or foul.
SPIKE AND BLOCK
5 - kill, stuff, or return for point or side out.
4 - terminal intimidation scoring play, or block assist with partner making stuff.
3 - play that results in attaining attack advantage.
1 - play giving opponent attack advantage.
0 - error or foul.
In using the above rating system, data are entered into microcomputers at courtside, or cassette recorders for later entry, to calculate each player's percentage for each skill, as well as an average and composite for all skills combined. Totals of all players are also computed for the team performance level. The results are displayed on a TV screen immediately, and printouts can be made at any time during a game. Below is a typical printout of the statistics for a 3-game match.
USA VS BRAZIL - MEN - 8/11/84
1984 OLYMPICS GOLD MEDAL MATCH
ACCUMULATED TOTALS FOR 3 GAMES. SCORES 15-6 15-6 15-7
USA DEFEATS BRAZIL 3-0 & WINS GOLD MEDAL. TEAM PERFORMANCE LEVEL = 560
NO. SER PAS DIG SET SPK BLK ATT AVG CMP P/L
TEAM 233:115 882:75 441:78 914:107 665:116 335:117 608 578 561 1
1 211:17 : 533:12 919:95 446:3 228:14 141 471 723 163
2 : : : : : :
3 : : : : : :
4 : : : : : :
5 : : : : : :
6 209:21 : 381:11 800:1 686:23 338:26 82 482 414 -146
7 266:18 : 600:5 700:2 777:18 385:27 70 545 480 -80
8 : : : : : :
9 200:1 0:1 500:2 : : : 4 233 300 -260
10 200:4 : : : : : 4 200 200 -365
11 : : : : : :
12 228:21 914:40 352:17 900:4 680:10 400:19 111 579 589 29
13 261:13 200:1 400:8 999:3 647:38 307:13 76 469 505 -55
14 : : : : : :
15 240:20 890:33 460:23 800:2 608:24 277:18 120 545 550 -10
The headings show abbreviations for player's number, serve, pass, dig, set, spike, block, total attempts, average, composite, and performance level. The average is figured by summing the percentages for each skill and dividing by the number of skills the player participated in. The composite is figured from all the attempts for each skill. The team performance level was set at 560, which was the performance level at that time. The P/L column shows how much the team and players' composite are above or below the team performance level.
The digits to the left of the colon (:), are the performance percentage, and the digits to the right are the number of attempts. (Note: Whenever 999 is shown, it is really 1000, but changed to a three-digit number to keep the output in even columns.) This percentage is computed by adding the total number of points a player earned in a particular skill, divided by the number of attempts, and divided again by 5, which is the total possible points that can be attained, (Points/Attempts/5). For example, if a player served an ace for a five, one serve for three, two serves for two, five serves for one, and one for a zero, the total points would be 17 for 10 serves. Seventeen divided by 10 equals 1.7, and then divided by 5, equals .340.
The average gives equal weight to all skills regardless of the number of attempts in the skill. A skill with few attempts has equal weight as one with many. The average indicates a player's all-around ability; whereas, the composite shows the player's contribution to the team. This is why a setter has a higher composite than an average, as he sets a great number of balls at a high percentage level. A printout showing the positive results of the match give the press needed information for each player. Included are the number of service aces, perfect passes and digs, setting assists, spiking kills, blocking stuffs, and blocking assists.
PERFECT PERFECT SETTING SPIKING BLOCKING
NO. SERVES PASSES DIGS ASSISTS KILLS STUFFS ASSISTS
TEAM 0 58 26 66 68 16 14
1 0 0 5 61 1 2 1
2 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0
5 0 0 0 0 0 0 0
6 0 0 3 0 15 4 2
7 0 0 3 0 12 4 5
8 0 0 0 0 0 0 0
9 0 0 1 0 0 0 0
10 0 0 0 0 0 0 0
11 0 0 0 0 0 0 0
12 0 33 3 2 6 5 1
13 0 0 2 3 22 1 1
14 0 0 0 0 0 0 0
15 0 25 9 0 12 0 4
An abbreviation of the first printout is displayed on 40 column screens. The IBM PC display is 80 column and is similar to the printout except for the headings. Additional monitors can be fed so broadcasters can have immediate access while the game is going on, as done at the Men's Pre-Olympic Tournament. The 40 column screen is displayed as follows:
USA GAME 3 TPL=560
NO SER PAS DIG SET SPK BLK ATT CMP P/L
TM 220 910 410 920 700 330 228 558 -2
> 1 206 403 920 201 407 56 713 153 2
> 6 200 203 556 319 28 409 -151
> 7 266 502 801 657 380 29 478 -82
> 9 201 01 2 100 -460
10 202 2 200 -360
>12 227 930 366 994 227 44 579 19
13 254 03 991 720 403 29 495 -65
>15 227 880 650 717 204 38 540 -20
SCORE 15-7 PASS/PT. = 2.00
The heading is the same as for the printout except the average is not included. With 40 column displays, the information in the printout is abbreviated. Instead of a three digit percentage, only two are displayed as the left two digits, and the third digit represents the attempts. The colon is eliminated. If the number of attempts is 10 or more, the third digit is zero. (Note: Whenever 99 is shown, it is really 100, but changed to a two-digit number to keep the output in even columns.) The pointer (>) indicates the players in the game at that time.
Also, at the end of a match, all the raw data can be printed out for an analysis of performance. The team total performance is first, followed by each player. The team and one of the players are printed below to show the format of this printout.
PLAYER NO. TEAM USA
DEFEATS BRAZIL 3-0 & WINS GOLD MEDAL.
SKILL 0'S 1'S 2'S 3'S 4'S 5'S TOT CMP TEAM
SER 4 93 13 5 0 0 <--ACES 115 233 233
PAS 2 2 0 13 0 58 <--PERFECT 75 882 882
DIG 36 3 0 13 0 26 <--PERFECT 78 441 441
SET 0 1 0 5 36 66 <--ASSISTS 107 914 914
SPK 16 25 0 7 0 68 <--KILLS 116 665 665
BLK 58 14 0 15 14 16 <--STUFFS 117 333 333
TOTAL COMPOSITE FOR ALL ATTEMPTS 608 561 561
AVERAGE OF ALL SKILLS 578 578
PLAYER NO. 1 USA DEFEATS BRAZIL 3-0 & WINS GOLD MEDAL.
SKILL 0'S 1'S 2'S 3'S 4'S 5'S TOT CMP TEAM
SER 0 16 1 0 0 0 <--ACES 17 211 233
PAS 0 0 0 0 0 0
DIG 4 1 0 2 0 5 <--PERFECT 12 533 441
SET 0 0 0 4 30 61 <--ASSISTS 95 919 914
SPK 0 2 0 0 0 1 <--KILLS 3 466 665
BLK 9 2 0 0 1 2 <--STUFFS 14 228 333
TOTAL COMPOSITE OF ALL ATTEMPTS 141 723 561
AVERAGE OF ALL SKILLS 471 578
By looking down the column of 5's, the number of serving aces, perfect passes and digs, setting assists, spiking kills, and blocking stuffs can be easily determined as shown in the previous printout. The blocking assists are shown in the column of 4's for blocking. The number of attempts at each rating level is given for each skill. Not only is this information available for the match just concluded, but an accumulation for the entire tournament or season is also available.
RESEARCH FINDINGS: In developing a system that was as objective as possible with a various levels in rating players (a scale of 0-5), the study shows a high degree of success. The correlation between different statisticians gathering data from the same games was .9818, which was extremely high. Defining each level of rating objectively reduced the amount of subjectivity in making these judgments. Giving a different rating based on which team has attack advantage on in-play balls, makes it possible to increase the number of scale values in rating the player's performance.
Likewise, trying to keep up with the rapid pace of international volleyball, and input the data live into the computer with a minimum of error, was possible with a correlation of .9990, a very high degree of reliability.
The third objective dealing with being able to differentiate well between players. This was the major reason for developing a new system. There are two weaknesses of the present method of figuring the spiking efficiency, (kills minus errors divided by attempts). First, spikers with different spiking abilities can get equal ratings. Second, negative percentages can be obtained which are awkward. Table 1 shows the effect the K-E method and 0-5 system developed by the author, have with ten attackers.
TABLE 1. SPIKING PERFORMANCE
COMPARISON BETWEEN K-E AND 0-5 SYSTEMS.
Pl. 0 1 3 5 K-E 0-5
A 2 0 4 4 200 640
B 2 1 3 4 200 600
C 2 2 2 4 200 560
D 2 3 1 4 200 520
E 2 4 0 4 200 480
F 4 0 4 2 -200 440
G 4 1 3 2 -200 400
H 4 2 2 2 -200 360
I 4 3 1 2 -200 320
J 4 4 0 2 -200 280
The table shows clearly, when using the K-E system, the top five attackers are all equal and the lower five are all equal, but the 0-5 system has them all well differentiated and with no negative percentages. The NCAA system computes percentages using the K-E method for attacking only. The 0-5 system computes percentages for each of the six skills. In the three rating systems that were discussed in the full report, (Olympic, NCAA, and the 0-5 systems), the 0-5 system used in this study demonstrated much more differentiation in rating players according to their ability than the other two methods.
The 0-5 system provided a method of not only rating individual players for each skill, but also provided a percentage that relates to his total ability as a player in all skills combined. The other two systems did not provide this kind of statistic. Two types of percentages are available in the 0-5 system, the average and the composite. The average is figured by summing the percentages of each skill and dividing by the number of attempts and dividing again by five, which is the maximum rating that can be given. The composite is influenced by the number of attempts in a skill, whereas the average treats each skill with equal weight. The evidence gathered in this study show that the average is more highly correlated (.8659) with scoring level than the composite (.7308). Both, however, were found to be very high correlations.
The 0-5 system demonstrated that it was more comprehensive than any of the other systems. The Olympic system only gathered 52.7 percent of the data gathered by the 0-5 system at the 1984 Olympics. When gathering data on the Men's USA team, the Olympic statisticians recorded 2,254 touches of the ball, whereas the 0-5 system recorded 4,275. The Olympic system gathered nothing for digging and setting. The NCAA stats gather approximately the same amount of data as the Olympic system, which is not every play of the ball. A more accurate assessment of players was demonstrated in the 0-5 system when total data was collected.
The sixth objective was to determine if the system's team statistics were correlated positively with success. The Olympic and NCAA systems have no comprehensive statistic to correlate with success. The 0-5 system does provide two such statistics, the team average and composite. Correlations with scoring level were .8659 for the average and .7308 for the composite, which were found to be quite high.
Questions were raised regarding the values given in the rating scales. When these scales were first established, they were set up rather arbitrarily and intuitively. Part of the objective of this study was to determine if these ratings are what they should be as related to scoring. The results were very clear that having intermediate levels between the top and bottom ratings are valid. Those skills that were executed at a higher level, showed significant difference between those rated at a lower level in terms of leading to a scoring play on the next possession of the ball in the rally. For example, a spike rated a three, lead to a scoring play 51.3 percent of the time on the next possession, whereas a spike rated a one, lead to a scoring play 31.1 percent of the time.
Percentages of scoring were computed for each intermediate rating of each skill, and from these data, adjusted ratings were established based on the actual percent scoring plays were achieved. Games from the 1984 Olympics were replayed by computer to observe the results using the adjusted ratings. It was observed that each of the six skills had lower correlations with scoring level, and the average and composite slightly higher. Because of the confusing results which do not correlate as high as the original ratings, and the fact that the correlations were really not much different than the results obtained by the original rating system, it was recommended that the ratings be left as they are. This would provide for a less complex system when otherwise, ratings for each skill would be different.
The last objective of this study was to find out which skills are most correlated with success. Many have indicated that the attack is the most important, as exemplified by handling that particular skill in a more advanced way than other skills in most rating systems using the K-E method. This study supports this notion and ranks the other skills accordingly. The ranking and correlations of the skills with scoring level were (1) spiking, .7568; (2) blocking, .6513; (3) setting, .5931; (4) serving, .5037; (5) passing, .4979; and (6) digging, .4963.
The number of volleyball plays gathered for this study number around 50,000. This includes both men's and women's play. The results give strong support to the system that was tested as being a valid statistical system. It met the challenge of the seven ingredients of a good statistical system with great strength in every category. It is hoped that through this study that the sport of volleyball will be able to move ahead in this modern day of technology in using a system that is more comprehensive. Volleyball statistics needs a new approach.
This computerized volleyball statistical system gathers more data than any other system and makes it immediately available to the press and others without a long delay. The data is presented in familiar style using percentages, which is similar to the reporting of statistics for other sports. The Men's and Women's USA National Teams used this system at the 1984 Olympics for self analysis and scouting other teams to help win Gold and Silver Medals.
*Note: A few years after this study, the rating scale was changed to 0-4, since that was the number of values in the 0-5 scale anyway. The ratings of 2 or 4 were not used for certain skills and made it confusing, therefore the change was made with no significant difference in the results.
Click here to return to SPORTISTICS home page.