Special Issue Information Sciences

Research for Supporting Tactical Analysis Concerning Pass Skeleton in American Football

The authors have been conducting researches on the measurement and analysis of athletes, things and events in field sports. In particular, we focused on American Football which is the most intelligent and complicated sports, and carried out the possibility of matchup analysis of its pass play. Based on the past researches, it was found that it was possible to estimate success or failure for each play by applying to the time-series trajectory image data of pass matchup players using the CNN（Convolutional Neural Network）. However, it is necessary to improve versatility by supporting pass skeletons considering not only one-on-one defense but also formation of zone defense that are close to the match format. Therefore, we aim to support the analysis of pass skeletons, which had to consider huge parameters such as the position and movement of each player from the start of play to the success or failure of the pass. The research was carried out based on a hypothesis that the success or failure of a pass from the position several seconds before the origin of the QB pass throw could be determined by taking into consideration the skill that is the compatibility of each player, the positions and the movement trajectory of the receiver team（WR, TE, and RB）, and the defenders（LB，DB，S）that mark them. As a result, the success rate and failure rate of the assumed pass player could be predicted by using the position and its trajectory image of each player from 3 seconds before the pass pitch. And then, by determining that the pass to the maximum likelihood receiver is optimal, we confirmed that useful information can be provided to support strategy planning during the game and guidance during practice.

View more >>