Possession-level live win probability model for football using LSTM and GRU on StatsBomb La Liga data, with Poisson goal conversion and proper probabilistic evaluation.
Live cricket scores and win probability dashboard. Built using Django and CricketData API, featuring dynamic polling, key rotation, and a custom WASP algorithm.
Mechabellum replay and live-match analytics toolkit for data collection, model training, and win probability prediction.
Live win probability, blunder detection, and hidden-team inference for competitive Pokémon Showdown battles — LightGBM on 19k+ ladder replays.
Interactive NFL fourth-down decision audit using nfl4th and nflverse data. Compare coach decisions with modeled win-probability recommendations, identify conservative and aggressive misses, and explore coach decision fingerprints.
NBA win probability, momentum, and turning-point analytics engine.
NFL in-game win probability model on 225k plays from six seasons of nflfastR. XGBoost, log loss 0.485 / AUC 0.841 on the held-out 2023 season. Python + DuckDB.
Multi-level probabilistic framework linking risk-adjusted soccer action values to match-outcome context.
Win-probability curve calibrated to human play — held-out log-loss 0.575 vs Lichess's 0.738, re-validated on 4M off-corpus positions. Static review app runs Stockfish in a Web Worker, no backend.
🏏 IPL T20 Win Probability Predictor - ML-powered live match analytics built with Scikit-learn & Streamlit
AI-powered Brawl Stars ranked draft assistant: win-probability model, seat-aware draft engine, and a web draft board.
Machine learning dashboard for NBA win probability, covering pre-game predictions, live replay, model evaluation, and automated Finals refreshes with Google Cloud.
CS2 win-probability & Player Impact (WPA) — end-to-end ML pipeline from parsed pro demos
NFL 4th-down decision audit. Win-probability model + coach scorecards with bootstrap confidence intervals.
My play-by-play ML win-probability model for a Bruin Sports Analytics NBA project: game-state feature engineering + MLP (PyTorch), Random Forest & XGBoost, evaluated by Brier score.
Open, pre-registered cricket win-probability and WAR from 5.9M deliveries. 22 predictions committed before any model existed; 6 held. A power check retracted the headline.
Bayesian player ability model for predicting golf win probabilities — time-varying ratings, Bradley-Terry/Glicko, historical PGA Tour data
MCP server for cricket: calibrated win-probability model, 22k-match ball-by-ball archive, career and matchup records, live scores. No API key required.
Add a description, image, and links to the win-probability topic page so that developers can more easily learn about it.
To associate your repository with the win-probability topic, visit your repo's landing page and select "manage topics."