Dota 2 Bayesian Tracker

A performance tracking system that leverages Bayesian methods to analyze and predict Dota 2 gameplay metrics. The system treats unknown parameters (player skill levels and hero selection impacts) as random variables with prior distributions that reflect initial beliefs or known constraints. As new match data is collected, these priors are updated to posterior distributions, capturing both updated estimates and uncertainty levels.

The implementation uses a Naive Bayes classifier approach where:

• Prior probabilities are established for match outcomes P(Win)
• Likelihood of hero combinations given outcomes P(Heroes|Win) is estimated
• Posterior probabilities P(Win|Heroes) are computed for new lineups using Bayes' theorem

This probabilistic framework allows for continuous model updates as matches complete, providing not just point estimates but full probability distributions that reflect genuine uncertainty in predictions.

Interactive visualizations of player performance metrics, Bayesian updating in action, and predictive modeling results.