ECSS Paris 2023: OP-AP15
INTRODUCTION: Recognizing adversarial group activities in high- speed sports like table tennis remains a significant challenge due to the millisecond-scale interaction between opponents and the complex trajectory of the ball. Traditional methods primarily focus on collaborative semantics, which are unsuitable for captur- ing the zero-sum dynamics inherent in competitive duels. In this paper, we propose STAGE (Spatio-Temporal Adversarial Graph Encoder), a novel hierarchical framework designed to explicitly model two-player confrontational interactions. STAGE processes information across four distinct semantic scales: Skeleton Feature Extraction (Scale 1), Individual Action Modeling (Scale 2), Adversarial Interaction Coupling (Scale 3), and Tactical Situation Reasoning (Scale 4). By bridging the gap between fine-grained joint coordinates and high-level tactical intentions, the framework enables a panoramic analysis of match situations categorized into proactive control, balanced confrontation, and passive constraint. Furthermore, we introduce TTAdvantage-5K, the first dedicated skeleton-based dataset in this domain. Extensive experiments demonstrate that STAGE achieves state-of-the-art accuracies of 87.2% on TTAdvantage-5K and 94.3% on the VD Olympic split, significantly outperforming existing baselines while maintaining exceptional computational efficiency (297M FLOPs). METHODS: The proposed framework for analyzing two-player com- petitive situations is organized into a hierarchical, end-to- end pipeline. This section details the technical architecture, beginning with a global overview of the system, followed by the specific design of the feature acquisition modules, the multiscale Transformer backbone, and the specialized quantitative analysis dimensions. RESULTS: Results on TTAdvantage-5K. Table II presents the compar- ison with state-of-the-art methods on TTAdvantage-5K dataset. Our STAGE achieves the best performance with 87.2% ac- curacy and 85.6% F1 score, significantly outperforming all baseline methods.Results on VD Olympic Split. Table I shows the results on VD Olympic dataset. STAGE achieves 94.3% accuracy and 92.8% F1 score, establishing new state-of-the-art per- formance.Efficiency Comparison. Table III compares the compu- tational efficiency in terms of FLOPs. Traditional backbone networks such as VGG-19 and VGG-16 require 3.6T and 2.8T FLOPs respectively, making them less efficient for resource- constrained scenarios. CONCLUSION: In this study, we presented STAGE,a Spatio-Temporal Adversarial Graph Encoder specifically engineered to capture the tactical dynamics of table tennis competitions. By mov- ing beyond traditional individual action recognition, STAGE addresses the fragmented nature of current analytics through ahierarchical four-level semantic approach: Skeleton Feature Extraction, Individual Action Modeling, Adversarial Interac- tion Coupling, and Tactical Situation Reasoning.
Read CV Ma WenlongECSS Paris 2023: OP-AP15
INTRODUCTION: Tennis match outcomes depend on the stability of tactical decisions under various situational pressure, and the tactics of serve and subsequent first shot (Serve +1) have been prioritized within the scoring pattern of players since around 70% of points under 4 shots. Yet existing research relies heavily on aggregated match statistics, offering limited into how players adapt serve tactics in context-specific scenarios. This study quantifies serve tactical tendencies and scoring efficiency among men's singles players across multi-level pressure contexts to evaluate adaptive decision-making strategies. METHODS: A total of 247 men's singles matches from the 2023-2025 Roland Garros (French Open) were included. A logic-based screening identified four core serve tactics: Strong Serve, One-two Punch, Wrong Foot, and Drop Shot, using serve landing coordinates, rally length, and point outcomes. Validation included: 1) macro-validation: correlation (r) and Mean Absolute Error (MAE) against official statistics; 2) micro-validation: accuracy and recall against manual annotation. Situational pressure was defined as three levels: Macro (set status: regular and deciding sets); Meso (game status: leading, regular, trailing, critical, and tie-break games); and Micro (point importance: regular points, critical points including game, deuce, and break points, and tie-break points). A Generalized Linear Mixed Model (GLMM) assessed situational effects on tactic selection and scoring efficiency. RESULTS: The classifier demonstrated high validity: macro-validation showed strong agreement with official data (r = 0.961, MAE = 1.97, p < 0.001); micro-validation yielded 96.15% precision and 57.47% recall. Situational pressure significantly influenced tactical behavior. At the Macro level, player reduced tactical diversity in deciding sets, favoring high-probability scoring patterns (One-two Punch usage rising from 21.9% in regular sets to 26.2%). At the meso-level, trailing players exhibited greater tactical flexibility, though with fluctuating tactic usage and efficiency (usage peaked in tie-break games at 26.0%, while scoring efficiency dropped to approximately 61% during critical games). At the Micro level, critical points significantly increased reliance on the "One-two Punch" tactic (reaching 26.0% in tie-break points), though its scoring conversion rate was more inconsistent compared to regular points. CONCLUSION: Data trends suggest high-pressure situations require superior precision in tactical execution. Serve tactical performance on clay courts is dynamically modulated by multi-dimensional situational pressure. Elite players consolidate their tactical repertoire during critical moments, even at the cost of reduced execution efficiency. These findings inform context-specific training drills simulating critical match scenarios to enhance tactical resilience and decision-making precision in elite tennis.
Read CV Xuehan ZhangECSS Paris 2023: OP-AP15
INTRODUCTION: Techniques and tactics, holding fundamental positions in tennis, are the dominant factors in competitive performance and significantly influence players’ performances and match results. Prior studies mostly focused on hard courts without considering venue effects and remained at the descriptive and comparative levels, with relatively lacking of specialized and targeted research on shot placements by using complex mathematical models. This study integrates venue factors from shot placements and studies the evaluation metrics of techniques and tactics. After formulating the evaluation criterion of shot placements in different types of courts, this study determines and analyses the impact of shot placements scoring rates in different types of courts on match results, which is valuable and innovative. METHODS: The Percentile Method is used to formulate evaluation criteria and the Grey Relational Grade(GRG) assesses the correlation between the scoring rates of shot placements and the total scoring rates. By ranking the impact of each tactical and technical metrics on scoring and losing points, this study determines the influence of metrics from shot placements scoring rates on match results. Recorded game videos includes 70 hard-court matches (2,520 games, 14,480 points), 70 clay-court matches (2,180 games, 13,252 points), and 70 grass-court matches (2,512 games, 14,022 points). Six metrics of shot placements include forehand straight(FH-ST), backhand straight(BH-ST), forehand diagonal(FH-DL), backhand diagonal(BH-DL), forehand side(FH-S), and backhand side(BH-S). All data observation, statistical analysis, and data entry were independently conducted by the researcher. To validate inter-rater reliability, 10 recorded game videos were randomly selected by a master’s student specializing in tennis for parallel observation and analysis. The Kappa coefficient between the two datasets reached 0.92, indicating high consistency and suitability for research. RESULTS: Scoring rate evaluation criteria for shot placements formulated by the Percentile Method are court-specific. On hard courts, the correlation between the scoring rates of shot placements and the total scoring rates are ranked as follows: FH-DL, BH-ST, BH-DL, FH-ST, FH-S and BH-S, with coefficients of 0.884, 0.872, 0.853, 0.848, 0.791, and 0.743. On clay courts, the order is BH-DL, FH-DL, BH-S, BH-ST, FH-ST and BH-S, with coefficients of 0.879, 0.875, 0.861, 0.857, 0.848, and 0.718. On grass courts, the sequence is FH-DL, BH-DL, FH-ST, BH-ST, FH-S and BH-S, with coefficients of 0.849, 0.826, 0.808, 0.803, 0.780, and 0.666. CONCLUSION: Evaluation criteria for shot placements on hard, grass, and clay courts enrich the tennis technical-tactical metrics system. The ranking of evaluation metrics for shot placements on hard courts, from highest to lowest, is: FH-DL, BH-ST, BH-DL, FH-ST, FH-S, and BH-S. On clay courts, is: BH-DL, FH-DL, FH-S, BH-ST, FH-ST, and BH-S. For grass courts, is: FH-DL, BH-DL, FH-ST, BH-ST, FH-S, and BH-S.
Read CV Chang LiuECSS Paris 2023: OP-AP15