Čuk I., Forbes W. HOW APPARATUS DIFFICULTY SCORES AFFEST ALL AROUND... Vol. 2 Issue 3: 57-63 HOW APPARATUS DIFFICULTY SCORES AFFECT ALL AROUND RESULTS IN MEN'S ARTISTIC GYMNASTICS v i 2 Ivan Cuk and Warwick Forbes 1 University of Ljubljana, Faculty of Sport, Slovenia 2Australian Institute of Sport, Canberra, Australia Original research article Abstract On a sample of 49all-around male gymnasts at the 2009 European Championships the implications of the difficulty scores were tested in relation to the success in all-around competition. After the regression, cluster and ANOVA analysis, three groups of quality all-around gymnasts were determined, while only one group had a chance to win an all-around medal; difficulty scores between all six apparatus were not equal; the highest prediction of the all-around score was the parallel bars difficulty score. Keywords: artistic gymnastics, man, strategy, all around. INTRODUCTION After the Olympic Games (OG) in 2004 the International Gymnastics Federation (FIG) made changes to the Code of Points. One of these changes was the implementation of a new philosophy of an open scoring system, prepared by Fink and Fetzer (1993), which had previously been introduced at the FIG symposium in Lugano in 1993. Prior to 2006 all disciplines in men's artistic gymnastics (FIG, 2000) were limited to a maximum final score of 10 points. In the past, different maximum scores were allowed, before World War II (WWII) the maximum score was sometimes between 11 and 16 points. After WWII the maximum score was limited to 10 points (Stukelj, 1989). Despite changes to what 10 points actually represented, it was decided that this represented exercise content and exercise presentation. The proportion of exercise content and exercise presentation had also changed; in the end it became equal to approximately 50:50 (Bucar 1998). Exercise content was mostly characterized by difficulty and special requirements. In the Code of Points 2006 (FIG, 2006) the whole philosophy of evaluating gymnastics exercises changed. No longer was one maximum score (10 points) for evaluating exercises used. New rules (FIG, 2009) defined D and E score, where D score evaluates exercise content (difficulty, special requirements, and bonus points) and the E score evaluates exercise presentation. 'D scores' start at zero points and increase according to the difficulty the gymnast demonstrates, how the exercise is constructed (the exercise must include elements from all five element groups, and no more than 4 from one group), and how difficult elements are connected (bonus points). The system works well for apparatus specialists; the more you show the greater the score, however in all around (AA) gymnasts a problem can exist. The problem is with the apparent equality between apparatus i.e. the vault has special rules compared to floor exercise, pommel horse, Science of Gymnastics Journal 5 8 Science of Gymnastics Journal Čuk I., Forbes W. HOW APPARATUS DIFFICULTY SCORES AFFEST ALL AROUND... Vol. 2 Issue 3: 57-63 rings, parallel bars, and horizontal bar. Gymnasts in AA competition only perform one vault, and compared to the other apparatus the vault is similar to only one element from the other exercises. Therefore, on the vault the D score is known in advance (FIG, 2009). According to the results of men's AA qualifications at OG 2008 Cuk and Atikovic (2009) found that the vault is considered to be the most valuable apparatus, and the pommel horse was undervalued among AA gymnasts. Using the Code of Points, it is very hard to obtain a high D score on the pommel horse, whereas it is easier to obtain a high D score on the vault. Pairwise t-tests showed that D scores between the vault and other apparatus, and between the pommel horse and other apparatus were significantly different. Table 1. Average D score (multiplied by 1000) and standard error of sample (N=44) at OG 2008 for MAG all-around gymnasts Cuk and Atikovic (2009) Mean Std. Error FXA Score 6015.91 50.572 PHA Score 5677.27 69.189 RIA Score 5943.18 95.257 VTA Score 6445.45 65.306 PBA Score 6090.91 84.834 HBA Score 5897.73 80.530 However, the new Code of Points presented in 2009 (FIG, 2009) has a number of changes that impact the D score. In the past (Hadjijev, 1989), it was expected that the least amount of training time was spent on the vault, and the most amount of time was spent on the pommel horse. Training times on other apparatus were similar ( the gymnasts preferences, abilities, and individual characteristics are also important in determining training time spent on each apparatus). Using the 2009 Code of Points, one of the most experienced Slovenian international judges Enis Hodzic calculated maximum difficulty scores for each apparatus. Results were Floor exercise=7.9; Pommel Horse=7.6; Rings=7.6: Vault=7.4; Parallel Bars=8.1; High Bar=8.5. It is clear that the maximum difficulty scores are different for each apparatus. The 2009 European Championships (EC) in Milan was the first major competition in the world to use the 2009 Code of points. It is therefore interesting to see how the AA gymnasts coped with the new rules, as their performances might be a guideline for the Olympic cycle up to the OG 2012 in London. The number of AA gymnasts has diminished over the last two decades (at OG in 1992 all the gymnasts were competing in AA in order to get into finals, while at OG 2008 and at WC 2007 only half of them competed in the AA competition). It is interesting to see how all-around gymnasts are coping with the new Code of Points and what kind of strategies they are using to improve their results. METHODS Our sample was composed of 49 AA gymnasts who competed at the EC in Milan 2009 qualification event. From official results we made 6 variables of D scores: Floor Exercise (FX) , Pommel Horse (PH) , Rings (RI), Vault (VT), Parallel Bars (PB) and Horizontal Bar (HB) . To evaluate the AA we used the AA final score (AAFS). To assist the statistical presentation, D and E scores were multiplied by 1000; so a score of 6 points had a value of 6000. SPSS 15.0 was used to calculate Kolmogorov-Smirnov to test the normality of the variables distributions, Pearson correlations, pair-wise t-tests between D scores of all apparatus, and a linear regression analysis between AAFS and D scores (method enter). We also prepared the classification of gymnasts with the method of Euclidian square distances using D scores. Clusters were then compared with one way ANOVA and Tamahne 2 post hoc test. All statistics used an alpha level of p < 0.05. Science of Gymnastics Journal 5 8 Science of Gymnastics Journal Čuk I., Forbes W. HOW APPARATUS DIFFICULTY SCORES AFFEST ALL AROUND... Vol. 2 Issue 3: 57-63 RESULTS AND DISCUSSION Table 2. Descriptive statistics Std. N Minimum Maximum Mean Std. Error Deviation K-S test FXDscore 49 3900 6400 5381.63 77.956 545.693 Normal PHDscore 49 3000 6800 4997.96 115.709 809.961 Normal RIDscore 49 2500 6500 5273.47 103.624 725.366 Normal VTDscore 49 4600 7000 6012.24 76.578 536.048 Not Normal PBDscore 49 3400 6500 5202.04 99.808 698.656 Normal HBDscore 49 3100 6800 5269.39 114.070 798.489 Normal AAFS 49 64325 89150 81395.41 693.419 4853.936 Normal Table 3. Pairwise t-test (N=48) Pair t Sig. (2- tailed) FXDscore - PHDscore 3.532 .001 FXDscore - RIDscore 1.170 .248 FXDscore - VTDscore -10.537 .000 FXDscore - PBDscore 2.151 .037 FXDscore - HBDscore 1.149 .256 PHDscore - RIDscore -2.713 .009 PHDscore - VTDscore -8.881 .000 PHDscore - PBDscore -2.148 .037 PHDscore - HBDscore -2.390 .021 RIDscore - VTDscore -7.475 .000 RIDscore - PBDscore .825 .413 RIDscore - HBDscore .042 .967 VTDscore - PBDscore 9.105 .000 VTDscore - HBDscore 7.087 .000 PBDscore - HBDscore -.781 .439 Table 4. Pearson correlation matrix HBDscore PBDscore VTDscore RIDscore PHDscore FXDscore AAFS .720* .830* .606* .743* .697* .710* HBDscore 1.000 .682* .452* .605* .511* .537* PBDscore 1.000 .517* .639* .620* .583* VTDscore 1.000 .431* .350* .700* RIDscore 1.000 .576* .511* PHDscore 1.000 .425* FXDscore 1.000 *all correlations are significant p<0.01 Science of Gymnastics Journal 5 8 Science of Gymnastics Journal Čuk I., Forbes W. HOW APPARATUS DIFFICULTY SCORES AFFEST ALL AROUND... Vol. 2 Issue 3: 57-63 Table 5. Regression analysis (method Enter), predicted AAFS variable R R Square df1 df2 Sig. .920(a) .847 6 42 .000 Unstandardized Standardized Coefficients Coefficients T Sig. Std. Std. B Error Beta B Error (Constant) 38244.8 32 3455.33 2 11.068 .000 HBDscore .753 .536 .124 1.405 .167 PBDscore 2.324 .687 .335 3.384 .002 VTDscore .677 .780 .075 .868 .390 RIDscore 1.357 .578 .203 2.349 .024 PHDscore 1.163 .484 .194 2.401 .021 FXDscore 1.868 .826 .210 2.262 .029 Cluster analysis with the method of gymnasts were grouped. Those with three Euclidian distances gave the best results clusters were used in further analyses via a with 3 clusters, where 21, 6 and 22 one way ANOVA. Table 6. ANOVA results with Tamahne 2 post hoc test Sum of Squares df Mean Square F Sig. FXDscore Between Groups 7179162.028 2 3589581.014 23.210 .000 Within Groups 7114307.359 46 154658.856 Total 14293469.388 48 PHDscore Between Groups 15670726.654 2 7835363.327 22.784 .000 Within Groups 15819069.264 46 343892.810 Total 31489795.918 48 RIDscore Between Groups 12291159.555 2 6145579.777 21.806 .000 Within Groups 12964350.649 46 281833.710 Total 25255510.204 48 VTDscore Between Groups 6085986.395 2 3042993.197 18.163 .000 Within Groups 7706666.667 46 167536.232 Total 13792653.061 48 PBDscore Between Groups 17029687.693 2 8514843.847 61.199 .000 Within Groups 6400108.225 46 139132.788 Total 23429795.918 48 HBDscore Between Groups 18955315.399 2 9477657.699 37.426 .000 Within Groups 11648766.234 46 253234.049 Total 30604081.633 48 95% Confidence Interval for Mean Std. Lower Upper Group N Mean Deviation Std. Error Bound Bound FXD score 1 21 5057.14 344.342 75.142 4900.40 5213.89 2 6 4966.67 674.290 275.278 4259.04 5674.29 3 22 5804.55 342.925 73.112 5652.50 5956.59 Science of Gymnastics Journal 5 8 Science of Gymnastics Journal Čuk I., Forbes W. HOW APPARATUS DIFFICULTY SCORES AFFEST ALL AROUND... Vol. 2 Issue 3: 57-63 Total 49 5381.63 545.693 77.956 5224.89 5538.37 PHD score 1 21 4904.76 618.447 134.956 4623.25 5186.28 2 6 3650.00 476.445 194.508 3150.00 4150.00 3 22 5454.55 578.773 123.395 5197.93 5711.16 Total 49 4997.96 809.961 115.709 4765.31 5230.61 RID score 1 21 5119.05 520.211 113.519 4882.25 5355.84 2 6 4166.67 882.421 360.247 3240.62 5092.71 3 22 5722.73 417.398 88.989 5537.66 5907.79 Total 49 5273.47 725.366 103.624 5065.12 5481.82 VTD score 1 21 5666.67 425.833 92.924 5472.83 5860.50 2 6 5800.00 438.178 178.885 5340.16 6259.84 3 22 6400.00 385.450 82.178 6229.10 6570.90 Total 49 6012.24 536.048 76.578 5858.27 6166.22 PBD score 1 21 4952.38 400.773 87.456 4769.95 5134.81 2 6 4000.00 428.952 175.119 3549.84 4450.16 3 22 5768.18 328.614 70.061 5622.48 5913.88 Total 49 5202.04 698.656 99.808 5001.36 5402.72 HBD score 1 21 5061.90 529.600 115.568 4820.83 5302.98 2 6 3916.67 636.920 260.021 3248.26 4585.07 3 22 5836.36 437.031 93.175 5642.60 6030.13 Total 49 5269.39 798.489 114.070 5040.03 5498.74 The descriptive statistics and Kolmogorov-Smirnov test (Table 2) showed that only the vault data D scores were not normally distributed. The score distribution was leptokurtic and skewed to the left, meaning that higher values are more common. Despite trying to normalize variables with logarithmic functions (ln and logio), the abnormality persisted, so we decided to continue analyses with raw data. Comparing the average of all-around D scores on the apparatus from OG2008 and D scores from EC2009 it can be noted that there is a huge lowering of D scores when the2009 Code of Points were used. On average, D scores were lower by 0.6 points; the greatest lowering was on parallel bars, and least on the vault. The 2009 Code of Points did not affect AA gymnasts on the vault, but mostly on the parallel bars. Table 7. Differences between AA scores from OG2008 and D scores from EC2009 0G2008 EC2009 Diference FX 6015.91 5381.63 634.28 PH 5677.27 4997.96 679.31 RI 5943.18 5273.47 669.71 VT 6445.45 6012.24 433.21 PB 6090.91 5202.04 888.87 HB 5897.73 5269.39 628.34 At the beginning of the Olympic cycle with the adoption of the 2009 Code of Points lower start values (as the value of some elements were lowered, less bonus points on apparatus) were expected, however the drop in scores was more severe than expected (from 0.43 to 0.88 point). If we compare what AA gymnasts could achieve according to maximum scores using the2009 Code of Points, it is noted that they were already achieving 81.2% of maximum possible score on the vault, while on all other apparatus they are below 70% of the maximum score. If we take into consideration the best gymnast by D score on each apparatus, the percentage of Science of Gymnastics Journal 5 8 Science of Gymnastics Journal Čuk I., Forbes W. HOW APPARATUS DIFFICULTY SCORES AFFEST ALL AROUND... Vol. 2 Issue 3: 57-63 maximum D scores were higher, but vault the best gymnast on other apparatus was was still the apparatus where the best AA below 90%. gymnast was already reaching 94.6%, while Table 8. Theoretical maximum D scores by Code 2009 and achieved ones at EC2009 The best AA gymnast The best AA Max Dscore Average AA gymnasts Dscore Code 2009 from AA gymnast % max Code 2009 % of max Dscore_gymnast_Code 2009_ 7900 68.1 6400 81.0 7600 65.8 6800 89.5 7600 69.4 6500 85.5 7400 81.2 7000 94.6 8100 64.2 6500 80.2 8500 62.0 6800 80.0 The pair-wise z-test (Table 3) showed 10 significant different pairs out of 15 pairs; all pairs with pommel horse and vault were significant different, and floor exercise with parallel bars. The average D scores on the vault were the highest and were lowest on the pommel horse. Similar results were obtained at OG2008 (Cuk, Atikovic, 2009). Pearson's correlations (Table 4) between apparatus D scores were all statistically significant, medium high. Correlations between all AAFS and each apparatus D scores were slightly higher, the highest was with parallel bars D score (0.83 - 68.9% of common variance). Surprisingly the lowest correlation was with the vault D scores (0.61 - 36.3% of common variance). The descriptive statistics and z-tests showed that the vault had important differences to other apparatus, but correlations revealed that for AA gymnasts the vault score had the lowest impact on AA score. Coefficient of multiple correlations (Table 5) between dependent variables of the AA final score and independent variables of apparatus D scores were statistically significant and very high (0.92). D scores explained over 84% of the final AA score, in general more difficult exercises attained better results in the AA. Significant predictors of AA success are parallel bars, rings, pommel horse and floor exercise D scores. It was interesting to observe that the vault and high bar D scores were not significant predictors of AAFS. On the vault there was not enough discrimination among gymnast's D scores.. Cluster analyses identified 3 groups of gymnasts. ANOVA (Table 6) showed they differed significantly on D scores. The third group (22 gymnasts) was very good on all events and had significantly higher D scores on all apparatus compared to the other groups. The first (21 gymnasts) and the second group (6 gymnasts) were equal on floor exercise and vault (the second group exceeded the first); while on the other apparatus the first group had higher D scores. Only the third group had the quality (level of D scores) of winning medals, so the questions to be asked are: why do gymnasts from the first and the second group compete in AA at all? Are they just trying to enter AA finals or are they just young gymnasts with a better potential future? CONCLUSIONS Based on the results presented it can be concluded that: - with the 2009 Code of Points, for all-around results the six apparatus are not equal to obtain D scores; - with the2009 Code of Points, for all-around gymnasts, the vault and the pommel horse D scores significantly differ from other apparatus; - with the 2009 Code of Points, the vault D scores do not discriminate between all-around gymnasts; - all-around gymnasts have the lowest D scores on pommel horse; Science of Gymnastics Journal 5 8 Science of Gymnastics Journal Čuk I., Forbes W. HOW APPARATUS DIFFICULTY SCORES AFFEST ALL AROUND... Vol. 2 Issue 3: 57-63 - with D scores only we can predict 84% of all-around final score; - after the Code of Points changed in 2009, the all-around gymnast who attained the highest D score on parallel bars has the best chance of good all-around results; - D scores for the vault and high bar did not significantly predict all-around final scores; vault D scores did not discriminate sufficiently (to many gymnasts with same D score), while on the high bar the lack of discrimination could be due to an increased number of falls. It seems it is more important to perform a slightly less difficult exercise well than a difficult exercise with a fall; - three groups of all-around gymnasts were classified (with 21, 6, and 22 gymnasts), and only the third group had potential of winning an all-around medal, as their D scores on all apparatus are much higher. REFERENCES Bučar Pajek, M. (1998). Primerjalna analiza tekmovalnih pravil v moški in ženski športni gimnastiki [Comparative analyse of Men's and Women's Code of points]. Diplomsko delo [Unpublished diploma thesis]. Faculty of Sport. Ljubljana. Čuk, I. Atikovic A. (2009). Are Disciplines in All-around Men's Artistic Gymnastics Equal?. Sport Scientific & Practical Aspects. International Journal of Kinesiology. 6, 1&2:8-13. Code of Points Men Artistic Gymnastics. (2000). Moutier: FIG. Code of Points Men Artistic Gymnastics. (2006). Moutier: FIG. Code of Points Men Artistic Gymnastics. (2009). Lausanne: FIG. Fink, H., Fetzer, J. (1993). Proposals of the new Code of Points. International symposium on Men's Artistic Gymnastics. FIG. Lugano. Hadjijev, N. (1989). Some problems of training young male gymnasts: FIG Symposium during Artistic Gymnastics World Championship. Stuttgart. Štukelj, L. (1989). Mojih sedem svetovnih tekmovanj [My seven world competitions]. Dolenjska založba. Novo Mesto. Science of Gymnastics Journal 5 8 Science of Gymnastics Journal