A Survey on Machine Learning Approach for Momentum Shift Detection and Win Prediction in Cricket Matches

Abstract: Cricket is among the most strategically complex and data-intensive sports in the world, producing extensive real-time performance data across multiple formats. Despite significant advances in sports analytics, existing research predominantly focuses on pre-match outcome prediction or score estimation, with limited investigation into in-match momentum dynamics and their effect on win probability. This survey critically examines five state-of-the-art contributions: momentum shift modeling in sports [1], DoE and regression-based cricket analytics [2], ensemble ML- based outcome classification [3], T20 dangerous-ball impact analysis [4], and ODI score prediction using regression [5]. Through systematic comparative analysis employing three reference tables, seven critical research gaps are identified— most notably the complete absence of real-time cricket-specific momentum shift detection frameworks. In response, this paper proposes an intelligent system integrating LSTM-based temporal momentum detection, XGBoost ensemble win probability prediction, a novel Cricket Momentum Index (CMI), and an interactive Real time analytical dashboard covering all cricket formats

Keywords: Cricket Analytics; Momentum Shift Detection; Win Probability Prediction; Machine Learning; LSTM; XGBoost; LightGBM; Random Forest; Gradient Boosting; Real-Time Sports Analytics; Match Outcome Prediction; IPL; ODI; T20;Feature engineering