Successful Stories
Helped John Milman (ranked #52) for his first-ever win over Roger Federer (ranked #1 best all time) in US Open in 2018.
Helped Novak Djokovic (current world ranked #1) to win Australia Open in 2020.
Video Analytics
We have developed world-leading algorithms for analyzing Fast, Frequent, and Fine-grained events and object-level detection & tracking from sports videos. For example:
Fast, Frequent, and Fine-grained Events Detection
Tennis Ball Tracking
Tennis Action Detection & Court Mapping
Soccer Ball Tracking
Soccer Movement Detection & Court Mapping
Explainable Sports Game Modeling
We model sports games using Markov Decision Processes (MDP) with play-by-play data, capturing complex behaviors and interactions. The model includes various match information, such as player characteristics and possible actions. The probability distributions and success rates of these actions are mined from historical data.
Match Outcome Prediction
Our model can realistically simulating sport games based on players' historical performances and accurately predict match outcomes.
Strategy Recommendation
Our model can quantitatively evaluate the efficacy of various strategies and recommend optimal ones that increase the winning chance.
Read more:
[1] Zhaoyu Liu, Kan Jiang, Zhe Hou, Yun Lin, and Jin Song Dong. Insight Analysis for Tennis Strategy and Tactics. IEEE International Conference on Data Mining 2023. Paper
[2] Zhaoyu Liu, Murong Ma, Kan Jiang, Zhe Hou, Ling Shi, and Jin Song Dong. PCSP# Denotational Semantics with an Application in Sports Analytics. The Application of Formal Methods, Springer Nature Switzerland 2024. Paper