Análise da influência das avaliações de críticos e usuários nos resultados do The Game Awards

Abstract

This study investigates the correlation between critic and user reviews recorded on the review aggregator Metacritic and the probability of a video game winning the main category at The Game Awards (TGA). To this end, a methodology previously applied to film awards—where no clear relationship was identified between aggregated scores and Oscar or Golden Globe winners—was adapted to the context of video games. Individual reviews of games nominated between 2014 and 2025 were collected through web scraping, utilizing a temporal filtering protocol that discards scores published after each ceremony to prevent data leakage. Based on the consolidated scores across all platforms, twelve statistical features were extracted. Three classifiers—Naive Bayes, K-Nearest Neighbors, and Random Forest—were trained on the 2014 to 2023 editions and subsequently tested on the 2024 and 2025 editions, across four source combination scenarios. The Random Forest model achieved a 91.67% accuracy in the combined sources scenario and 100% accuracy in the 90/10 weighted scenario, which replicates the weights assigned by the TGA to the jury of critics and the popular vote, yielding perfect precision and recall. The results indicate a strong and measurable correlation between aggregated reviews and the award outcomes, contrasting with observations in the cinematic landscape. Furthermore, the findings suggest that the predictability of an award is positively correlated with the transparency of its voting criteria.