Assessing COVID-19 impact on user opinion towards videogames - Sentiment analysis and structural break detection on steam data

Abstract

As we live in a world where the videogame industry grows day by day and new media is constantly emerging, user feedback can be widely found online. User reviews are a highly valuable data source when studying player perception of a videogame. They are also apparently volatile to updates released by developers and other external events, which may change user opinion over time. Here we assess whether the COVID-19 pandemic outbreak fell in this category, having or not a noticeable impact on the player view and popularity of videogames. In this research, we build and implement a method to collect active player data and user reviews, identifying the sentiment contained in the expressed opinions. Furthermore, we investigate the existence of structural breaks in the time series we target. For this purpose, we targeted user-reviews and active player data collected of Steam’s twenty most popular Massive Multiplayer Online Role- Playing Games. To collect sentiment polarity values, two Natural Language Processing Python libraries were used, TextBlob and VADER, and structural break detection was put into practice using strucchange R package. The results of this work show us that despite having a great effect on the number of active players, the COVID-19 pandemic did not produce the same impact on Steam user reviews. Nonetheless, we were able to identify one of the platform’s major reviewing related updates as a structural break. We believe this approach can be used for further assessments on public opinion towards a specific product, in the future.