Quantitative Evaluation of Cross in Esports Soccer: Modeling Based on Offense/Defense Positioning

The term “esports” is used to refer to game-based competitions considered as sports events. Esports currently involves a qualitative instruction that relies on individual experience and intuition, such as instruction from professional gamers. By contrast, sports observe a shift from qualitative to high-quality instruction using quantitative evaluation indices. Therefore, in esports, evaluation indices must be quantified to improve the training quality. This study focuses on crossing situations in soccer games. Crossing is a play in which a player passes from the side to the goal. It is important in soccer because of its high possibility of leading to a goal. Therefore, the quality of cross training in soccer must be improved to increase the scoring probability by crossing. As a feasibility study, this work aims at a quantitative evaluation of crossing in soccer games. The important factors for a successful cross are the positional relationship and the speed of each player at the time of the cross. A crossing scene is modeled to obtain a cross score based on these factors. First, when the velocity and trajectory of the cross ball are determined, the area is defined in which the offensive and defensive players can touch the ball. This area is calculated, and the value of how close to the center the ball trajectory passes in relation to the area is determined for each player. Using these values, the Cross Score (CS) was obtained by constructing a cross evaluation formula. To confirm the validity of the obtained evaluation values, a system that can obtain the evaluation values using image processing was developed. Then, 264 videos of crossing scenes in a soccer game were obtained, and CS was calculated for each of them. The correlation coefficient between the cross score and the percentage of successful crosses is 0.923, showing a strong correlation and confirming CS validity. As an example of application of the CS, it is possible to judge whether a cross is good or bad by displaying the CS on a heatmap using quantitative values. This enables visual and quantitative feedback using CS, suggesting the possibility of improving the quality of training.