Final Data Analysis Report

Data Analysis Project Nathan Missen

1

Introduction and Purpose of study What characteristics make for a winning rikishi? Sumo is a unique sport in its own right; no other event that has its combination of explosive timing, brute force, use of strategy, and luck. It is also one of the few sports that requires extreme physiological characteristics of its participants (other comparable sports include female gymnastics & jockeying, with the possible inclusion of female ballet). Thus, it stands to reason that there would be unique characteristics for a successful sumo wrestler. There are many different factors involved in determining the outcome of a single sumo match. Timing, strength, the crowd, and many more aggregate together to eventuate in either a win or a loss. Of these many aspects, three wrestler-explicit variables have been isolated and chosen for analysis in this study: weight, height and favorite technique type. Each will be explained in more detail in the following paragraphs.

First, why include weight in this analysis? Weight would certainly seem to be important factor; any sport involving the use of leverage, pivot points, etc places a heavy emphasis on the individual’s body mass. In Greco-Roman wrestling, which can be compared, at least superficially to sumo, weight control and weight loss are important morphological variables (Martin & Margherita, 1999). Similar conclusions were also made for elite judo athletes (Callister, Callister, Staron, Fleck, Tesch P & Dudley, 1991). Weight gain has also been linked to rank progression in sumo (S.W.A, 1993). Also, we can apply basic physics principles, such as Newton’s Second law and the formulas demonstrating its effect. In particular, Force = Mass x Acceleration, and Momentum = Mass x Velocity. The first formula is of importance in Sumo, as when the match begins, the contestants are only 70cm apart. Therefore, generating maximum force in the first few seconds of a bout is essential to winning; the more weight a wrestler has, the more force they can produce, thereby increasing the certitude of a win from the initial charge (taichiai). However, if a rikishi weighs too much, they may not be able to accelerate very quickly from their starting position, thereby reducing the potential match-winning force. Weight needs to be carefully

Data Analysis Project Nathan Missen

2

monitored and regulated to have a positive effect upon a sumo combatant’s success.

Secondly, why has height been included in this study? There is practically no research (again, in English) concerning height in sumo; nor is there any literature relating the importance of height in other combative arts (judo, Greco-roman wrestling, jiu-jitsu, etc). However, we can utilize basic physics principles to provide reasoning for the inclusion of height in this research project; according to mechanics of torque (Torque = force x length), a wrestler who is a taller (and, by extension, has a longer reach) will be able to produce a higher degree of rotational force than a shorter opponent, given the situation that both wrestlers are in an equidistant, stable position (i.e. a “clinch”). This is a regular occurrence in sumo, as wrestlers tend to absorb the impact of the taichiai through their sheer mass. The rikishi are then in ideal position to throw each other; thus, being able to produce and utilize torque is of extreme importance in these situations. However, Sang-Hyop (2006) do mention that a wrestler of greater than average height would be at a disadvantage; since their center of balance is higher, taller wrestlers can succumb easier to throws, especially if their lower body is targeted. Height, it would seem, is a double-edged sword.

Choice of technique can also be of paramount significance in determining who is victorious in the doyho (sumo ring). Of the forty-six official techniques (or kimarite as they are known), two types are pre-eminent among rikishi: yori (“push” type) and nage (“throw” type). Both of these types encompass a range of techniques that are aligned upon a spectrum of complexity; however, there is a fundamental difference between these two. Yori category moves rely primarily upon momentum to (literally) propel rikishi to victory; techniques that fall in the nage classification are dependent upon utilization of rotational and perpendicular force. Thus, even though yori and nage are used quite commonly used in the sumo tournaments by all wrestlers, the principles and application behind each are subtly different, and validate their inclusion in this study.

Data Analysis Project Nathan Missen

3

Finally, it must be noted that there is a limited amount of material in this area. For certain, there are papers covering a rikishi’s risk of diabetes (Manyu et al, 2005, Ekoe, 1989), how sumo participants challenge the limits of the human body (Kondo, Abe, Ikegawa, Kawakami & Fukunaga, 1994) and even jump speed (Alegre, Anzar, Jiménex & Aguado, 2005). However, it is believed that there are currently no papers in English covering this particular aspect of sumo physical characteristics (in combination with technique preference) and their influence on separate and specific measures of success.

Thus, the overarching purpose of this study is to evaluate what makes for a successful rikishi. By examining and assessing how three factors (height, weight and favorite technique type) operate as single sources of influence as well as in combination, insight can hopefully be determined into what distinguishes a consistently above-average performing sumo from the rest.

Research Questions In this study, three key research queries are of interest. The first question addresses if height, weight, or technique in isolation have a significant influence upon the three designated measures of success (i.e. percentage of matches won, number of tournaments/prizes won, and time between initial debut and makuuichi debut). Research and scientific principles mentioned in the preceding paragraphs outline the importance of height, weight and technique. Consequently, each one will be appraised separately. The second goal is to evaluate if the above factors, when aggregated, have a cogent impact upon the outlined measures of success. This is to assay if a single factor source of variation exerts a larger effect upon the sumo’s performance than a multi-factor source of variation. Also, the results may prove valuable to other researchers or experts in the sumo field wishing to perform multiple regression analyses. The final inquiry for this study is to explore what exactly defines “success” in sumo by examining and comparing the data yielded on the three measures of success. The rationale behind this revolves around the core concept of multiple

Data Analysis Project Nathan Missen

4

methods of data collection; having just a single measure of success may well be misrepresentative as there could be external, uncontrolled influences that can bias certain outcomes. By utilizing several gauges of accomplishment, a more complete and comprehensive profile of that sumo wrestler’s performance can be produced. In particular, the most common measure (that is, win percentage) can often be deceiving at a superficial glance. Thus, the number of months between initial debut and makuuichi debut (or, the number of months between attaining a professional rank and an elite rank) and number of tournaments/special prizes won (which could prove to be a direct measure of a rikishi’s skill, as prizes are usually awarded for technical demonstrations, rather than winning per se). In addition, these results may serve to inform and prompt discussion amongst followers of sumo, or even lead to a broader discussion about what constitutes success in sports as a whole.

Based upon the aforementioned reasoning and research and the author’s (albeit limited) knowledge of the mechanics of sumo, it was projected that of the three quasi-independent variables in this study, none of them in and of themselves would have an effect upon any of the three measures of success. This is because the author believes that there is no single, readily isolated factor (such as weight or height) can solely account for success. In contrast to this, at least one interaction is predicted to occur and have an effect upon one (or more) of the specified measures of success. In particular, sumo wrestlers of slightly- above average to above average weight were expected to have more success with yori-based maneuvers than nage-division skills. The reverse was expected for those rikishi of slightly-above average to above average height. Finally, it is anticipated that at least one of the estimates of success will be consistent and conclusive.

Methodology The current forty-two makuuichi (top division, or literally, “inside the curtain”) ranked wrestlers

Data Analysis Project Nathan Missen

5

were selected for this study. Of the forty-two, fifteen are of non-Japanese origin (ten Mongolians, two Georgians, one Bulgarian, one Estonian and one Russian). Of the twenty eight remaining, fourteen are from south-west Japan, six are from northern Japan, five are from central Japan, two are from western Japan and one is from southern Japan. For age, the youngest subject is 22 years, 1 month, the oldest is 37 years, 2 months. All information was gathered from the International Sumo Association website.

The quasi-independent variables used in this study were: height, weight and favorite technique type. Height and weight were divided into three nominal categories: Average (within +1 standard deviation from the mean), Average (within -1 standard deviation from the mean) and NonAverage (outside +/- 1 standard deviation). Technique type was divided into two nominal categories: yori and nage. There were three measures of success: number of wins (expressed by a percentage), number of tournaments/special prizes won (i.e. the “fighting spirit” prize, the “technique” prize, etc) and the time between initial debut and makuuichi debut (in months).

Statistical procedures used included one-way ANOVA’s, post-hoc tests, interaction comparisons and simple main effects analyses.

Null hypotheses for the one-way ANOVA’s are:

Hwin percentage: µheight = µweight = µtechnique type Htournaments/prizes: µheight = µweight = µtechnique type Hmonths: µheight = µweight = µtechnique type

Null hypotheses for the interactions are:

Data Analysis Project Nathan Missen

6

Hwin percentage: We predict that there will be no significant interaction between the factors Htournaments/prizes: We predict that there will be no significant interaction between the factors Hmonths: We predict that there will be no significant interaction between the factors

Results To examine the influence of height, weight and technique, nine one-way ANOVA’s were conducted, three for each measure of success. Descriptive statistics are provided below (Table 1). All results were non-significant at the α =.05 level, with the exception of height on number of months between initial debut and makuuichi debut, F(2, 39) = 3.882, p = .029. A graph is included (Figure 1). Variance was also examined using Levene’s test, F(2, 39) = 3.558, p = .038, and the result was significant. Since the homogeneity of variance assumption was violated, a non-parametric ANOVA alternative was selected; in this case, the Kruskal-Wallis test. The procedure was significant, Χ2 (2, N =42) = 7.964, p = .019. Thus, we can safely conclude there was significant difference among the means of the three levels within “height”. Follow-up tests were conducted to evaluate pairwise differences among the three levels. Type I error was controlled for using Bonferroni’s sequential procedure. The results of these tests indicated no significant difference between the levels of the “height” factor. In summary thus far, the results yielded indicate that height has significant effect upon how quickly a rikishi rises through the ranks of sumo. However, even with the use of post-hoc tests, it is unclear what height range is required for such rapid progression.

Table 1. Descriptive statistics for Height, Weight and Technique type

Data Analysis Project Nathan Missen Variable Height (cm)

Weight (kg)

Technique type

Variable Level Average height (within +1 SD) Average height (within -1 SD) Non-Average height (> -/+ 1 SD) Total Average weight (within +1 SD) Average weight (within -1 SD) Non-Average weight (> -/+ 1 SD) Total Throw type (nage)

7 Mean 188.313

SD 1.740

N 18

182.777

1.996

16

186.375

96.234

8

185.286 163.67

6.06 3.71

42 20

148

4.739

14

175.25

36.648

8

156.262

18.895

42 25

Push type (yori) Total

17 42

Figure 1. Height vs. number of months (DV = Number of months)

To ascertain the effect of the three quasi independent variables upon the three measures of success, three 3x3 ANOVA’s and six 3x2 ANOVA’s were utilized. All results were non significant, except one significant main effect for height in the height by favorite technique type

Data Analysis Project Nathan Missen

8

with number of months as the dependent variable, F(2, 36) = 3.419, p = .044 and a significant interaction between height and favorite technique type with win percentage as the dependent variable, F(2, 36) = 5.329, p = .009. Pairwise comparisons were then employed to examine the main effect for height. The Tukey HSD procedure was used to control for Type I error. No significant differences were found between the three levels of height. To examine the interaction, custom hypothesis tests were used. Results yielded indicate that there was a meaningful difference between Average height (within +1 standard deviation) and Non-Average Height for “Throw” type wrestlers, using win percentage as a dependent variable, as well as the same result for “Push” type wrestlers. A similar pattern was demonstrated between Average height (within -1 standard deviation) and Non-Average height. Further custom hypothesis tests also evinced the change in win percentage across technique type was significantly different between Average height (within +1 std. deviation) and Non Average height. Again, a similar pattern was observed with Average (within -1 std deviation). A graph has been included (See Figure 2) Regarding the interaction, the information yielded demonstrates that rikishi of nonaverage height enjoyed a significantly higher degree of success using throw type maneuvers than their shorter fellow competitors. Conversely, when using push type techniques, taller wrestlers had a significantly lower rate of winning, and that this shift in performance, when compared to the other groups, was of a noteworthy magnitude.

Data Analysis Project Nathan Missen

9

Figure 2 A significant interaction between height & favorite technique type

Discussion This study was conducted as an exploration into what characteristics of a rikishi are associated with elite-level performance. The findings extracted from the data analysis are inconclusive at best; in regards to the original research expectations, two of the three were disproved, while minimal evidence was found for the remaining one. The first expectation was concerned with single factors accounting for variation within measures of success, and it was predicted that no single factor would influence any of the measures significantly. However, it was found that height did impact upon the number of months; those of Non-Average height progressed quicker through the sumo ranks then their shorter peers. The author believes this can be explained through three premises: 1) the principles of torque explained earlier 2) inexperience in dealing with taller opponents may account for some success as well. However, since this result was not repeated across the other two measures of success, the third premise seems more likely; that is, this study had some limitations that biased the results

Data Analysis Project 10 Nathan Missen (see below). The second expectation for this study was upheld, though in a limited fashion. Height and technique type did produce a significant interaction. More specifically, those wrestlers designated as Non-Average height enjoyed more success with nage, rather than yori techniques. Of the eight wrestlers within the Non-Average category, six were more than 1 standard deviation above the mean. From this, we can tentatively conclude that taller-than-average wrestlers are more successful with throw type techniques. However, this trend was not repeated across any of the other measures of success. Furthermore, there were no significant interactions for weight, so any conclusions made are tenuous at best. The third and final prediction for this study was also rejected. None of the measures could reliably differentiate between more or less successful elite sumo. While this may be due to characteristics inherent in the sample, a methodological error could account for these results; since the author is not an expert on the subject, these measures may have been inappropriate choices. Further investigation into this is warranted. A major limitation of this study was the small sample size. While the sample was representative of elite sumo wrestlers, it was not characteristic of sumo wrestlers as a group. The elite sumo wrestlers diverge in a number of ways from lower-ranked rikishi; elites receive a higher salary, have more time and opportunities to train, are excluded from many menial tasks and receive more food (S.W.A, 1993). In contrast, elites are also subject to a variety of unique adversities, ranging from a higher chance of public criticism to rapid rank loss or even expulsion. These factors can directly or indirectly impact the success of an individual; as such, if the study were redone, or a similar study initiated, then the sample would definitely have to include a broader range of ranks. A second limitation related to the first is the limited power of this study. Posthoc power analysis revealed a large range of power values, ranging from .053 to 1.00. However, the majority

Data Analysis Project 11 Nathan Missen of power values revealed were below .70, which is the standard for adequate power in research. This is quite meaningful, as the data produced would lead us to reject almost all of the null hypotheses within this study. However, it may be that there is a significant difference, but there is no enough power present to detect it. This quandary is another reason to increase sample size in the future. Another limitation is that the cognitive component of the rikishi has not been addressed. Sang-Hyop (2006) and Liu (2003) argue that beliefs in self-efficacy and the efficacy of others play a vital role in how athletes perform when in the field. In addition, Liu mentions that the spectators at a given sport event can also influence the athletes perceptions. This data would most likely be a valuable source of information to any study concerning sumo performance success; by not examining this characteristic, valuable influences may have been not been accounted for. However, gathering this data would require a lot of time and effort that may be beyond the resources available to some researchers. A final limitation to note is that the sport of sumo is not without its imperfections. Using econometrics, Duggan and Levitt (2002) concluded that match rigging occurred regularly within sumo, and had been occurring for some time. If this is the case, this study may not be taking the correct approach; rather than analyzing bodily characteristics and techniques, other constants should be investigated (rank progression, increases in tenure, overturned win/loss decisions, etc), as well as modifying or completely restructuring, incorporating different research questions as needed. In conclusion, this study aimed to shed insight into what influences performance amongst elite rikishi, as well as asking what defines success in the sumo realm. Even though a minimally significant data was found, the results are a may be helpful in guiding and informing future investigators of this area. Research in the future should focus on obtaining a larger, more representative sample and acquiring as much relevant data on individual sumo

Data Analysis Project 12 Nathan Missen (physiological/cognitive) and sumo culture as possible before analysis. References

Alegre, L.M., Anzar, D., Jiménex, F. & Aguado, X. (2005). Architectural characteristics of vastus lateralis muscle and jump performance in young men. Journal of Human movement studies, 48, 109-123.

Callister, R., Callister R.J., Staron, R.S., Fleck, S.J., Tesch, P. & Dudley, G.A. (1991). Physiological characteristics of judo athletes. International journal of sports medicine, 12, (2), 196 -103.

Duggan, M. & Levitt, S. (2002). Winning isn’t everything: Corruption in Sumo. The American Economic Review, 92, (5), 1594 – 1605.

Ekoe, J. (1989). Overview of diabetes mellitus and exercise. Medicine & Science in Sports & Exercise, 21, (4), 353-355.

Liu, W. (2003). Field Dependence-Independence and Sports with a Preponderance of closed or Open Skill. Journal of Sports Behavior, 26, (3), 285-297

Manyu, L., Dehuang, G., Carlos, M.I., Decio, L.E., Atkinsin, M., Jin-Xiong, S. & CongYi, W. (2005). SUMO wrestling with type 1 diabetes. Journal of molecular medicine, 83, (7), 504-513.

Masakatsu, K., Takashi, A., Shigeki, I., Yasuo K., Tetsuo, F. (1994). Upper limit of fat-

Data Analysis Project 13 Nathan Missen free mass in humans: A study on Japanese Sumo wrestlers. American Journal of Human Biology, 6, (5), 613-618.

Sang-Hyop, L. (2006). Belief, Skills, and Winning: Evidence from Sumo Wrestling. Paer presented at the Society of Labor Economists, 2006.

S.W.A. (1993). Sumo Watching. Tokyo, Japan: Yohan Publications.