Predicting Football Matches Using Data

on YouTube ↗14m

" summary

Builds logistic‑regression models on Google Cloud (Dataflow for ingestion, BigQuery for features, Compute Engine for training, Docker, pandas, StatsModels, scikit‑learn and IPython) that predict match outcomes from event‑level Opta data, a constructed expected goal feature and a power statistic for team strength. Models correctly predict 14 of 16 World Cup games (8 of 8 round of 16), hit by a duplicated‑ingest bug for one quarterfinal, and overall show ~88% success with limits set by Poisson scoring variance.

AI-written summary · about

" chapters (3)

1. 0:00Introduction
2. 4:50The Data
3. 7:00Machine Learning

" related talks

• 39mPredicting sports winners using data analytics with pandas and scikit-learnRobert Layton · PyCon Australia 2015 Science/Data Miniconf
• 46mLooking Into the Crystal Ball: Using Google's Prediction APIJustin Grammens · DevFestMN 2016
• 32mCutting Edge Football Analytics in PythonJoris Bekkers · PyData London 2025
• 18mis delighting fans with game-changing insightsAndreas Heyden · AWS Machine Learning Summit 2021
• 31mMachine Learning Driven Injury PredictionThomas Huijskens · Data Science Conference 5.0
• 59mPredictability Check: Demand ForecastingHillary Farmer · DSC DACH 24
• 22mPredict the future (or don't) with machine learning in applicationsMark Hornick · Oracle DatabaseWorld 2022
• 33mYour best Bet: Effortless MLOps with Python Models inDorian Van den Heede · PyData Eindhoven 2023