This is actually a 2-in-one question.

I am a big chess fan, and I have been following the FIDE candidates tournament closely. I was watching an older interview where IM Levi Rozman interviews GM Ian Nepomniachtchi on youtube.

If you hover over the duration bar, you can see a time-dependent histogram of the number of viewers for each timestamp (not sure how the time is discretized for this sampling, but that's not important for this question).

![enter image description here

I have 2 questions, the second being the more important one:

  1. How can I capture this data from a screenshot of the viewer distribution?

This is the viewer distribution data as a function of time by itself (using ImageTake and I ColorConverted to grayscale): ![enter image description here

I quickly tried using Binarize but had trouble teasing the grey distribution out of the background (maybe ImageSaliencyFilter could work here?)

Binarize[%,0.42] seems to work reasonably well but I'm struggling to find the baseline: enter image description here

  1. (and the bigger question): What are these distributions for the "spikes" in viewerships? It seems like they are non-Gaussian: the "front" part of a lot of the spikes rise steadily, but then they drop off very quickly once whatever interesting thing has finished being said . Note also the vertical bars in the time-dependent viewership distribution indicate a change in subject. I am struggling to understand what kind of distribution could describe this. Is there any way to describe this behavior with some distribution? An event occurs that causes people to pay more attention, and then when it ends, the attention quickly falls off.

I know I have a reputation for being good at answering questions, but not asking them... I hope this was well-written enough that I could get some insightful responses.

  • $\begingroup$ "The bigger question" is outside of this forum. $\endgroup$
    – user64494
    Commented Apr 18 at 14:36
  • $\begingroup$ If you open up a browser and inspect the video player you'll find there are several divs under class .ytp-heat-map-container with multiple chapters. Each chapter has an svg element with class .ytp-heat-map-path containing a filled curve. I don't use Firefox / Chrome so I can't test this right now as my WebExecute isn't working, but you may be able to try: session = StartWebSession["Chrome"]; elements = WebExecute[ session, {"OpenPage" -> "https://www.youtube.com/watch?v=tHvU3Qky8gU", "LocateElements" -> "CSSSelector" -> ".ytp-heat-map-path"}]; DeleteObject[session]; $\endgroup$
    – flinty
    Commented Apr 19 at 12:33
  • $\begingroup$ @flinty thanks I will try to get this to work. $\endgroup$
    – ydd
    Commented Apr 19 at 15:24


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.