Kick Analysis Post


Working Titles

  • How Kick defrauds sponsors by inflating live viewer counts: A data-driven analysis.
  • Is Kick (causing creators to) defraud(ing) sponsors? A data-driven analysis.
  • Kick inflates live viewer numbers, causing streamers to defraud their sponsors


  1. Noticed with sethdrums that he has 10x viewer count when doing special event, look for patterns in this
    1. Check for viewer patterns across Twitch viewers and Kick viewers
    2. Check if anyone streaming has a significant portion of the total live viewers at any given time across the platform, any anomalous spikes that cause the total live viewers to be way above what they normally are on the platform
    3. Compare live viewers to chat, if the views went up 10x then chat activity should go up 10x
    4. Twitch can’t do this because they have advertisers
  1. How we investigate
    1. How many streamers have inflated streams, and how many inflated streams do they have?
    2. There are certain things that should happen when real viewers increase massively
      1. Increased chat messages
      2. Increased self-subs
      3. They are featured
      4. … see questions below
    3. There are things that shouldn’t happen
      1. People who have never chatted/gifting subs before being especially generous. Especially if these account do this across streamers or have sort of random names
      2. Viewers should not spawn out of thin-air
      3. They shouldn’t pop above the data collection threshold then fall below
    4. Did these inflated streams cause an equivalent jump in viewers on the platform? e.g. did they come out of thin air?
  1. By inflating viewers numbers, creators are mislead in their metrics, and unknowingly charge sponsors too much info
    1. Proves there need to be better metrics than ACV
    2. ACV is specifically bad, anyone in DS will know this, because anomalies destroy the value, need MCV (median) or some specific like AVG p75 (AP75)
    3. Creators are unaware, so they can’t really be to blame
  1. Kick has a history of shady tactics, and comes from a brand of shady tactics
    1. These tactics are going to ruin creators because they are pulling them away from stable truthful platforms
    2. any platform that has opt-in bot protection should be seriously criticized for trustworthiness
    3. the 95% split is unsustainable, used as marketing channel for gambling (they pump numbers there to make it on the front page), and it will get rug-pulled once they have enough ownership. Say they take over twitch, nothing stops them from rug-pulling that split.
    4. The fact they have not implemented other tools like an API, extensions, etc. shows they don’t really care about streamers, they just care about market share. Same with the fact that they use AWS and pusher, which has to be mind-boggly expensive

Questions and Charts

  • How many streamers have inflated streams? — How many streamers have inflated peak viewers, who had a stream where the peak viewers was more than 5x their median views within +- 15 days, charts:
    • Distribution of the median of inflated streams as a multiple of their median
    • Number of streamers with at least one stream at some multiple (bar, x-axis multiple above median, y-axis count of streamers — maybe could do funnel with integer multiples?)
    • Compare to the % of streamers this happens to on kick vs twitch?
  • Is there increased chat activity during inflated streams? — Chart ratio of chat messages to viewer count (overlay chat ratio on top of viewers, distribution). For streamers that have 5x streams:
    • stream peak views as a multiple of their median (bar, count, y-axis being multiple, x-axis median number of streams per creator at this multiple)
    • chat messages/min as multiple of the median (line
  • Let’s dig in to one a seemindly highly inflated stream, does chat look different? — Look at an individual stream, plot viewers over time against chat message velocity (sethdrums with %12% in stream name)
    • each pulse of viewers (line)
    • count of chat messages (bar between prev and next point in line)
    • Another Chart: Compare this to the avg chat messages across streamers where this is the median viewer count
  • What kinds of streams get viewers inflated? — Look at what kinds of titles cause inflated numbers (sentiment/LLM analysis on whether title is a special event compared to the rest)
    • Find streamers who had a stream where the peak viewers was more than 5x their median views
    • get stream names sorted by peak viewers, take the title with the largest peak during the day
    • Run this against LLM to see which it determines it would consider “special” events without giving viewer numbers?
    • Compare to twitch?
  • Does the number of self-subscriptions change during inflated streams?
    • Peak stream live viewers (bar) more than 5x above median
    • Number of gifted subs (line)
    • Number of self-subs (line)
  • Do gifted subs go up? If so are they repeat gifters on the channel? Might be a bot sending gifted subs because they don’t need it for themself
    • Find streams where viewers are above 5x median
    • count sub gifts by gifters
    • find which of them have gifted before, how many of them are new gifters
  • Did these viewers come out of thin air? Compare to the count of viewers across the platform and see if it seems inflated (or did they get subtracted from somewhere else)
  • Based on the 8 viewer limit of the data, are there people that temporarily popped above this line, that have not re-appeared? Can find people that had few streams above 30 live viewers, and then we don’t have any more data. Then manually check their pages for missed streams in our data.
  • What is the totaly platform to chat ration on kick vs twitch? Compare same 16 hour time frame for chat data we have, see ratio of live viewers to chat messages.