How We Used Machine Learning to win at HQ Trivia

Mux, a company obsessed with HQ Trivia, has developed a machine learning model to predict the correct answers in the game. The model uses Optical Character Recognition (OCR) and Google search queries to extract feature values from the text, which are then passed into an XGBoost model for ranking and scoring. Despite achieving 80-90% accuracy, the model does not guarantee consistent wins due to its inability to predict when it will be wrong. The team has won four times in two weeks using this model but has only made $76.13 from playing games with it.