[Feature: Sports and Science] Seiji Hirosawa: The Development of Figure Skating Brought by Technology

2024/07/05

In figure skating, performances are conducted to music within a specified time, and rankings are determined by the scores of judges. Such competitions are called scored sports (judged sports). You are likely familiar with this classification. When it comes to technology utilization in scored sports, the introduction of automated scoring systems is a topic that often attracts attention. Since it is a competitive sport, many people believe it is desirable to minimize variations in judgment caused by the subjective decisions of judges as much as possible.

When thinking of automated scoring in scored sports, the case of gymnastics likely comes to mind first. Under the leadership of the International Gymnastics Federation (FIG), an AI scoring system has been introduced with the aim of achieving automated scoring to ensure accurate and fair judgment while reducing the burden on judges. Development began in 2017, and since 2019, it has been used in some events at international competitions, including the World Championships. At the 52nd FIG Artistic Gymnastics World Championships held in Antwerp, Belgium, in 2023, its application to all events began. In introducing this system, the FIG worked with the development company to establish standards so that non-quantified scoring rules such as "straight" or "slightly bent" could be judged mechanically, such as "the knee angle is greater than 170 degrees." In the future, they plan to sequentially release guidelines for judging techniques more specifically using numerical values such as joint angles and limb positions, as well as provide specialized commentary and footage to viewers worldwide.

Following these efforts in gymnastics, I feel that voices calling for the introduction of automated scoring systems in figure skating, which is also a scored sport, are becoming stronger. On the other hand, I also hear opinions that "figure skating should be evaluated by humans because it competes on artistry." How should technology be utilized in the judging of figure skating?

One hint lies in the concept of a new sports genre called "Artistic Sports" proposed by Associate Professor Tatsuki Machida of Kokugakuin University *1. Mr. Machida, who was also a figure skater representing Japan at the Sochi Olympics, proposed subdividing scored sports to newly include "competitions where the competition rules require artistic physical expression that necessitates subjective interpretation" as Artistic Sports. In the competition rules for figure skating and rhythmic gymnastics, expressive acts and originality developed alongside music are required. On the other hand, what is required in men's artistic gymnastics is the "difficulty and quality of techniques," and since it is scored by comparison with the ideal form defined by the competition rules, the originality of the competitor is not required.

In this way, it can be said that Artistic Sports, where artistry requiring subjective interpretation is questioned, and Formalistic Sports, which seek the ideal form defined by competition rules, have different characteristics despite being the same scored sports.

Therefore, the scoring rules for Artistic Sports that require artistry must separate "Technical Scores" that evaluate technical aspects and "Artistic Scores" that evaluate artistic aspects. Regarding technical evaluation, it is necessary to make objective evaluations without leaving room for interpretation. On the other hand, regarding artistic evaluation, it is said that room for interpretation should be allowed for both the competitor and the evaluator.

In other words, in the case of figure skating, there may be room for the introduction of technology for objective evaluation regarding the difficulty certification and execution evaluation of technical elements such as jumps, spins, and steps. However, if technology is introduced to perform objective evaluations for artistic evaluations such as musical expression, there is a possibility of undermining the unique competitive characteristics of Artistic Sports.

Recently, tracking systems have been introduced at various sports competitions, making it possible to obtain athlete data without experimental intervention.

In figure skating, which is highly popular as a spectator sport, although an automated scoring system like that in artistic gymnastics has not been introduced, tracking systems have been introduced for the purpose of enhancing the viewing experience for spectators. Fuji Television Network, Inc., which holds the broadcasting rights, and Qoncept, Inc., which specializes in image processing, have developed two systems to date. One is "I-Scope" *2 for visualizing jumps, and the other is "Ice-Stats," which visualizes the trajectory of an athlete's skating during a performance. The introduction of these systems has made it possible to visualize the performance of figure skaters during competition, which was previously difficult to understand quantitatively (Figure 1).

I focused on I-Scope. In current figure skating scoring rules, the weight of jumps is extremely high. To win in competitions, it is important to successfully perform high-quality, high-difficulty jumps, and athletes are going through trial and error not just to avoid falling, but to perform jumps that judges will consider high in quality. I-Scope films the entire rink with two 4K cameras and calculates the height, distance, and skating speed after landing. The takeoff and landing points of the jump are defined by the operator. Since the actual size of one pixel on the filmed screen can be calculated from the information in the skating rink's drawings, the values can be measured. In broadcasts, a separate video with the trajectory of the jump's parabola is prepared to display the measurement results. The measurement and creation of the trajectory video are completed in about one and a half minutes and delivered to viewers. After the competition, distance and height rankings are released, providing a different way to enjoy the sport than the competition scoring. It might be interesting to have formalistic competitions that compete only on such numerical values of jumps.

I myself entered the Faculty of Environment and Information Studies in 2010 and joined the figure skating division of the Keio University Athletic Association Skating Team to begin competing in figure skating. While competing, I wanted to research the question of "what is necessary to jump better," but obtaining figure skating data in an experimental environment was difficult in many ways. In such a situation, "real-world" jump measurement data during competitions was very attractive to me. Rather than letting the data end with its use as media content, I thought, "Can it be utilized for research to support the improvement of competitive ability?" Under the consultation of Professor Yoshimitsu Aoki, my supervising professor in the Ph.D. program at the Keio University Graduate School of Science and Technology, I entered into a joint research agreement with Qoncept, Inc., which develops the system, and began my research. The target was "Grade of Execution (GOE) evaluation of jumps" within the technical scores where objective evaluation is required. The score for a figure skating jump is determined by the sum of the "Base Value," which is determined by difficulty such as type and number of rotations, and the "Grade of Execution," which represents the quality of the performed technique. Unlike the E-score in artistic gymnastics, a characteristic of figure skating is that the GOE is not a deduction from a perfect score but includes additive evaluation. Judges look at a jump and immediately evaluate it on an 11-step scale from -5 to +5, including 0. Evaluations regarding the base value, such as under-rotation of jumps or edge use at takeoff, are judged for deductions after integrating the opinions of multiple judges (Technical Officials). On the other hand, for the GOE, a trimmed mean is adopted, which averages the scores after excluding the highest and lowest scores from the nine judges. For this reason, a certain amount of variation among judges is permitted. It can be said that the judging criteria also have room for interpretation. Table 1 shows the evaluation criteria for positive aspects. For example, regarding height and distance, it is written as "very good," and not as "the value is large."

This time, to examine "what characteristics of jumps are receiving high GOE evaluations from judges," I analyzed jump data obtained at the 2019 World Championships (Saitama) and the 2023 World Championships (Saitama). Figure 2 shows data for a total of 66 double axel jumps performed by female athletes at both competitions, which were judged to have a GOE of 0 or higher by all nine judges. The horizontal axis of each graph shows the values obtained from I-Scope, and the vertical axis shows the GOE after the trimmed mean (out of 5 points). Gray dots and dash-dotted lines are 2019 data, black dots and dash-dotted lines are 2023 data, and the black solid line is the overall trend. Looking at this, it can be seen that the height of the jump has almost no relationship with the GOE evaluation. On the other hand, regarding distance, there seems to be a tendency (positive correlation) where the larger the value, the higher the GOE. Both the 2019 and 2023 tournaments showed similar trends, and a characteristic is that the relationship between distance and GOE does not change significantly depending on the match. Regarding skating speed after landing, although not as large as distance, a positive correlation is seen in both 2019 and 2023. In other words, although the criteria for jump GOE evaluation include the item "height and distance are very good," it can be said that judges tend to give high GOE scores to jumps with large distances that flow smoothly after landing—that is, jumps with a large horizontal component.

As data acquisition during competitions has become possible in this way, signs have emerged that we can quantitatively understand the tendencies of judges' subjective judgments. However, there are few competitions where the system is introduced, and because the sample size is small depending on the type of jump and number of rotations, quantitative evaluation is difficult. Also, because the takeoff methods differ, it is unknown whether the same trend would be seen for different types of jumps. If such a system is continuously introduced and analysis continues, it will serve as a benchmark for quantitatively determining "what is the best jump as a sport" when performing digital adaptation of judging criteria in the future, as in gymnastics. To introduce the system in sports competitions held with various stakeholders involved, many challenges beyond technical aspects such as measurement accuracy must be overcome, but for the development of the sport, it is strongly desired that data acquisition be conducted continuously in more matches.

For previous related results using tracking data, please refer to published papers *3.

In addition to those related to scoring, I will briefly introduce the use of technology in coaching, mainly in Europe and the United States. One is the introduction of wearable devices worn around the waist. Many skaters do not know how many times they have jumped during practice. In modern figure skating, where higher difficulty jumps are required, the impact on athletes when landing jumps is also large, and it has been reported that cases of stress fractures are increasing due to athletes practicing too much. It is said that the U.S. Figure Skating Association uses devices to manage practice volume and provide guidance based on quantitative jump data. This is expected to contribute to injury prevention and performance improvement.

Another is the emergence of analysts. Recently, the placement of staff specializing in information analysis has been progressing in various sports. The Japan Sports Analysts Association defines an analyst as a "specialist who supports athletes and teams with a high level of expertise in terms of information strategy to lead them to achieve their goals." In figure skating, Mr. Arnaud Muccini from France is a representative figure. He uses the image processing technology of Dartfish, a dedicated software used by sports analysts for analysis, to quantify jump data from video and use it for coaching. Toward the 2022 Beijing Olympics, he served as an analyst for the Chinese national team and led the athletes to a gold medal in the pairs event. His skills are highly evaluated by Dartfish, and he has been awarded the title of certified expert, making him one of the world's leading analysts beyond the framework of the sport.

The development of technology can protect the health of athletes and contribute to the improvement of performance. In Japan as well, the cultivation of personnel and the development of environments where guidance can be provided using the latest technologies mentioned above will be required.

Figure skating is a sport where both artistry and technique are emphasized, and by utilizing technology after identifying the points to introduce it, its appeal can be further brought out. Along with the evolution of technology, I will pay attention to how figure skating develops in the future, and as a researcher, I want to work to contribute to the development of the sport as much as possible.

*1 Tatsuki Machida (2020) "Introduction to Artistic Sports Studies: A Cultural Theory of Creation and Enjoyment Based on Figure Skating" (Hakusuisha)

*2 "Figure Skating Life Extra Professionals: People Who Support Figure Skating" (2020, Fusosha Mook)

*3 Hirosawa, S., Watanabe, M., & Aoki, Y. (2022). Determinant analysis and developing evaluation indicators of grade of execution score of double axel jump in figure skating. Journal of sports sciences, 40(4), 470-481.