You mean like why are there spikes between events? I'm not exactly sure.
I suspect it's because the labeling is based on the schedule, which they didn't always perfectly adhere to. So like many events would go a little long, and so the peaks right around the turn of the hour are actually at the ending of the event rather than right after they finished.

I think at some point I'll have to rewatch everything and make a note of when stuff actually happens to match up against the data.

I kept track of the number of twitch viewers, the donations, and the chat comments during Tip of the Hats 2016 and graphed some of the results.

Graphs
Github Repo
Reddit Discussion
Stats from Last Year

Data collection done with rust, curl, and irssi; processing done with python; graphing done with R and gnuplot.

Pretty sure they were trying out phorofor