… We focus in this paper on Twitch both because of its scale and popularity and because of its wide distribution of types of communities; while Twitch is still primarily a game streaming platform, it has branched out significantly in recent years to encourage creative streaming and In-Real-Life (IRL) streaming. There is vast HCI research on audience and spectator engagement, including within gaming \[8\], but our understanding of how audiences and streamers collectively participate in live game streaming platforms is still limited \[8,14\]. Most existing studies have focused on small scale audience-gamer interactions, for instance how small groups of gamers and audiences collaborate at home or in arcades \[18,26,47\]. …

… Most existing studies have focused on small scale audience-gamer interactions, for instance how small groups of gamers and audiences collaborate at home or in arcades \[18,26,47\]. But, very recent late breaking work \[14\] has showed the significance of this limitation, identifying critical relationships between audience size and audience behavior. …

… Although stream content types vary widely, studies indicate viewers gravitate to the most popular streams \[19\], and overall watch a limited number of games \[12\]. While the number of Twitch viewers is steadily growing \[48,49\], the number of streamers and the average number of hours streamed is growing even faster \[14,48,49\]. Additionally, the Twitch interface promotes streams with large audiences, creating a feedback loop in which large streams can grow faster than small ones \[20\]. …