Suspicious Cheering with Bits

This research analyzed irregularities in payment through monthly tips from users on a well-known live streaming platform by analyzing a public dataset containing over one million entries from January 2022 to October 2024. The platform records revenue through its Bits virtual tipping system while providing channel-level reporting about monthly earnings together with average Bits per message and cumulative revenue. The research method included first cleaning the data across months, then using Z-score thresholding and rolling average analysis as statistical detection tools to identify channels with abnormal revenue rises. A process of manual verification examined flagged anomalies by matching them to external metrics, including subscriber counts, streaming activity, viewership trends, and event timelines for the purpose of detecting legitimate spikes from potentially suspicious irregularities. Most flagged revenue spikes corresponded to actual popularity increases due to audience growth or occurred during special events. However, a distinct subset of these spikes lacked any external explanation, raising suspicion. These research findings contribute to the development of fraud detection capabilities for digital content platforms such as Twitch, YouTube Live, and Facebook Gaming, and help to create platforms that are more transparent regarding revenue and maintain higher integrity.