This is a fragmentation chart of how Lowco split her time while streaming in October 2015: pre-recorded material, live content, and when the stream is idling. It also shows stream length and number of streams. A fragmentation chart is band-stroke chart similar to when you defragment your hard drive or use certain "download managers." It is meant to express the frequency and distribution of one variable versus one or more other variables. Specifically, this chart is supposed to illustrate the amount of times that Lowco leaves her chair empty and abandons the stream because she doesn't like us any more... or some other reason that sounds less needy and less like an older sibling complaining about a newborn. I know. Let's go with the chart is supposed to illustrate the amount of screen time that Charlie gets. The amount of time that an "Empty Chair" gets to shine in the spotlight. Yeah, yeah. That's more positive. Here is the data in a spreadsheet: horizontal and in stalactite form. And here is the Infographic: Data Gathering Methodology Watching 190 hours straight for when Lowco gets out of a chair is implausible and boring -- like watching security camera footage for for a week straight. Even if you could scrub through a video on a Twitch VOD (like YouTube), that would take hours of time. Hours that I don't have even on 72 hour days, meaning that every two of my days syncs up with every six of yours. Plus almost all of that time is spent actually seriously working and not making puckish, slightly confusing infographics. So, I made the computer watch all of October's casts and note when Lowco was gone for a period longer than 300 frames (approximately ten seconds), which eliminated when she got up to turn on/off air conditioning, took off a jacket, or leaned out of frame. Luckily, Lowco's shot composition is relatively high contrast: white walls, dark hair, mostly black chair. The computer sampled one frame in every 300, averaging blocks of pixels in the camera frame as either black or white. When the frame had a change from the last sample that met a threshold of light pixels in certain regions, the computer would go back and read each frame in a more granular fashion to find when Lowco first stood up and to check if the series of frames were more-or-less static. Then it would search for an end point when Lowco sits back down (majority dark frames). The parameters are kind of loose but should have worked well enough to give a broad overview of Lowco's time spent on stream. Here is a demonstration of what you see and what the computer was looking for:
