MomentumShiftAI is a machine learning project dedicated to analyzing the inner dynamics of tennis and table tennis matches.
- 📊 Dataset construction: build reproducible datasets for tennis (point → game progression) and table tennis (point → set progression).
- 🔢 Theoretical probabilistic formulas: estimate the probability of winning from any intermediate score state.
- 🤖 Machine learning models: capture recurring patterns and sequential dynamics that theoretical models cannot fully explain.
- 🔍 Scoring sequence analysis: identify key moments such as momentum shifts and comebacks.
- 🧠 Psychological factors study: understand how competitive pressure influences performance.
- ✅ In tennis, the theoretical formulas showed excellent adherence to real data (maximum deviation ≈ 3%) and allowed the analysis of comeback patterns across different rankings and individual players.
- ✅ In table tennis, LSTM networks proved to be more effective than closed-form formulas, thanks to their ability to model the sequential nature of the data.
MomentumShiftAI is a solid and reusable foundation for those who want to explore new ideas or develop future applications.