Attempt to Quantify Momentum in Sports Competitions

This study aims to investigate the impact of momentum in tennis matches and develop methods for its prediction. Momentum is defined as the deviation of an athlete's scoring trend from a random process, representing a critical non-technical factor in competition. We employed a fuzzy comprehensive evaluation model to assess athlete performance during matches and quantified momentum by calculating the deviation between actual scoring probabilities and theoretical binomial distributions. To predict shifts in momentum, we utilized Long Short-Term Memory (LSTM) network models, which effectively address issues related to short-term memory and analyze various pieces of information at critical game moments, such as score differences, consecutive scoring, and errors, to determine the factors influencing momentum changes. Additionally, we tested the generalizability of our model in table tennis and volleyball matches, demonstrating that the model can reflect the impact of momentum across different sports. Finally, we compiled a one-page memo summarizing our findings to assist coaches in devising strategies or providing timely reminders during matches to ensure athletes gain a momentum advantage.