/jp/ - Otaku Culture

Ok, so while looking further at the girls' numbers, I found some more data:
1: There is a likely consistent pattern. Whenever I track a chart with obvious spikes, there's a fluctuation every three minutes. Because the tool isn't as precise as I'd like, I can't be perfectly accurate in the timeframe, but approximately three minutes between every spike holds as a reliable data point.
2: The spikes exist regardless of branch. The picture I liked is from the offcollab between Kanata, Coco, Rushia and Haachama. It was harder to see at first, but by zooming my screen in to it's maximum, the chart crunched to fit within the monitor's range, and with such a tight chart, the spikes become more obvious.
3: If I were to guess, the spikes are not the actual result of the bug. Instead, it would seem like the dips are the active events, and the spikes are the result of a more passive event. I say this because I highly doubt that Youtube would have something that adds extra views to a streamer, as that would artificially give the streamer more money via ads. Instead, it's more likely that the culling of viewers is the action that leads to spikes.
If I were to make a hypothesis with the data I've gathered, then I would say that the spikes may be forming through a program they run that scans for whatever they determine to be "spam viewers." Every three minutes, they open up the gates to count viewers, but when it rises to a point above what they would consider natural, the bot triggers and activates a suppressant to the viewcount, possibly only allowing the most clear-cut viewers into the viewcount and eliminating all edge cases. That cycle repeats every three minutes.
If that is the case, then that could be paired with the waiting room bug to explain what is happening. If the waiting room bug culls a huge chunk of waiting viewers off of the count, then it may be misidentifying all of the waiting room viewers as edge case viewers. So whenever it runs the three minute detection program, it starts counting culled waiting room members, which then get classified as "spam views," activating the suppressant program and creating the spike pattern.
Keep in mind that I'm only hypothesizing based on what limited data I could collect. This may all be false, but it's also better than having nothing to go off on.