Projects


NBA Player Interaction Network Analysis

  • Analyzed NBA player interactions (passes) to model team chemistry and playstyle across seasons.
  • Created dynamic networks representing players as nodes and passes as edges, measuring team performance, playmaking roles, and chemistry using centrality and assortativity metrics.
  • Explored network evolution across different NBA eras and its impact on team performance.

Nano Type Inference

  • Implemented Algorithm W for type inference in `Nano’, a miniature Haskell-based language
  • Designed type environment, unification, and substitution logic according to Hindley-Milner type systems
  • Collaborated in a 4-person team to integrate independent Nano modules (pattern matching, parser, REPL)

Predicting Game Length of 2022 League of Legends Esports Matches using Early-Game Data

  • Developed a regression model to predict the total length of 2022 League of Legends matches using early-game data (gold difference, comeback status).
  • Improved model performance from linear regression (R²: 0.185) to Random Forest Regressor (R²: 0.741) by incorporating additional features and hyperparameter tuning.
  • Conducted a fairness analysis comparing World Championship matches to other leagues, confirming model fairness with a permutation test (p-value: 0.4).

Competitiveness of 2022 League of Legends World Championship Matches

  • Analyzed 2022 League of Legends match data to assess if World Championship (WCS) games are more competitive than regular season games.
  • Investigated the correlation between gold difference at 15 minutes and match outcomes, finding that WCS games had similar gold differences to other competitive leagues.
  • Cleaned and processed data from multiple leagues and visualized key patterns in match dynamics using EDA techniques.
  • Performed hypothesis testing to compare gold differences in WCS matches vs. other leagues, concluding that WCS games are not statistically more competitive in terms of early-game gold difference.