Skip to content

Bayesian in-game win probability model for K-league(Korean football league) based on Pieter Robberechts's model

Notifications You must be signed in to change notification settings

jmlee8939/K_league_in_game_win_probability_model

Repository files navigation

in_game_win_probability_model

Bayesian in-game win probability model for K-league based on Pieter Robberechts's model

K-league 실시간 승리확률 예측 모형

실시간으로 제공되는 이벤트 데이터를 활용하여, 경기 시점마다의 각 팀의 승리확률을 예측하는 모형입니다.

  • Premier league win probability model. alt text

K-league LIVE EVENT DATA

  • K-league 데이터 포털 에서 제공하는 실시간 이벤트 데이터
    • 슈팅
    • 오프사이드
    • 반칙
    • 카드(경고, 퇴장)

alt text

활용한 데이터

  • 활용 가능한 K-league 데이터가 없어 Statsbomb opendata 활용.
    • PL, La Liga, Ligue1 814경기의 event data 활용
  • 전 시즌 순위 및 해당 시즌 경기 결과로 elo-rating 값 산출.

모델링

  • Bayesian poisson model
  • 베이지안 추론 pymc package 활용.

모델 구조

K-league 실시간 숭리확률 예측 예시

  • 하나은행 K-League 4round
  • 전북(홈) vs 울산(어웨이)
  • 경기일 : 2024/03/30
  • 경기결과 : 2:2

alt text

Reference

  1. Statsbomb opendata
  2. K-League 데이터 포탈
  3. pymc
  4. Explaining Live Win Probability (LWP) - opta
  5. Americal Soccer analysis
  6. Win probability model - Abhshek sharma
  7. A Bayesian Approach to In-Game Win probability in soccer(2019) P. Robberechts, Jan Van Harren, J. Davis
  8. A Bayesian approach to predict football matches with changed home advantage in spectator-free matches after the COVID-19 break(2022) J Lee, J Kim, H Kim, JS Lee

About

Bayesian in-game win probability model for K-league(Korean football league) based on Pieter Robberechts's model

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published