A Pirate’s Discretion: Modeling Differences Between Legal and Illicit Anime-Streaming Communities

Abstract

Market segmentation allows businesses to divide their target market into distinct sections and fine-tune their commercial approach to each one. In the anime-streaming market, however, these distinctions are often blurred, particularly over the distinction of the legality of streaming means—either of legal or illicit streaming sites—attenuating market insights that stakeholders might expect to successfully guide internal and customer-facing decisions. This paper proposes and tests the hypothesis that differences between the two kinds of streaming sites and their user bases are statistically and practically significant, refactoring the way entertainment leaders select and support shows for broadcast and streaming in the anime market. To do this, the paper takes three data sets representing the legal anime-streaming community (Crunchyroll), the illegal anime-streaming community (AniWave), and the whole, non-sectioned anime community (MyAnimeList), and studies the intersection of all three sources across key feature variables. Using LASSO regression and random forest trees, two powerful machine learning models, it then predicts the commercial success of shows on both the legal and illegal streaming sites and highlights the different variables significant to each platform. The best-performing model on the Crunchyroll data, LASSO, achieved an r-squared of 0.494. Random forest trees, the best-performing model on the AniWave data, achieved an r-squared of 0.596. The most significant variables for the Crunchyroll data, moreover, regardless of model, are controversy (the extent to which AniWave and Crunchyroll ratings disagree about any show) and the number of episodes in a show; for AniWave, the show’s success according to MyAnimeList, and again, the number of episodes.