Predicting Game Performance Via YouTube Sentiment

This analysis investigates whether sentiment expressed in YouTube comments on game trailers can serve as a reliable predictor of a video game’s commercial performance. The study explores the hypothesis that aggregated audience reactions—ranging from enthusiasm to disdain—provide actionable signals regarding a title's potential success or failure upon launch.

The methodology involved two distinct phases. An initial, smaller-scale test analyzed tens of thousands of comments for four high-profile titles, including both critical successes and commercial flops. To improve rigor, a second, broader analysis examined 59 independent games released in late 2024. This phase utilized the YouTube API to collect over 18,000 comments, which were then processed using OpenAI’s GPT-4o model to assign sentiment scores of positive, neutral, or negative. The resulting sentiment data was compared against estimated sales figures provided by the GameDiscoverCo database, filtering for non-free-to-play, non-adult titles that sold between 1,000 and 250,000 copies.

The findings indicate that while sentiment analysis appears to align with the reception of major AAA titles, it fails to predict the commercial performance of independent games. In the broader sample of indie titles, there was no meaningful correlation between the mean sentiment of trailer comments and subsequent game revenue. However, the study identified a robust relationship between the total volume of comments and game performance. This suggests that for independent developers operating in a crowded marketplace, the sheer level of audience engagement—regardless of whether the sentiment is positive or negative—is a more significant indicator of market awareness and potential reach than the qualitative nature of the feedback itself. Ultimately, the research concludes that predicting game success remains an elusive challenge, with raw engagement metrics currently offering more utility than sentiment-based analysis.