Skip to main content

2021-05-31

  • Daily Log:
    • Memorial day, going to try to use this day to do some self-learning
    • Read some books tonight before bed and get to sleep at 10
    • Go through another chapter in [[Hands On Machine Learning]]
    • Dream's cheating controversy
      • Dream is a popular minecraft Youtuber/Streamer who has been caught in a cheating controversy from October of 2020
      • He submitted a several minecraft speedruns and the mods investigated the runs and claimed Dream must have used a modded version of the game in order to encounter the scenarios that he did
      • Dream hired his own mathematician shortly after and released a report that refuted the mods' claims
      • Both side released very "mathematical" LaTeX documents arguing using statistics to prove their points
      • Some key facts
        • Minecraft speedrun on v1.6 is to fight and kill the ender-dragon. This is predicated mainly on two important chance events
          • Ender pearl drops and blaze rod drops
        • The combined probability of both things happening at the rate it was shown in Dream's submitted runs had the odds of 1 in 2×10221 \text{ in } 2\times10^{22}
          • Using binomial probability formula for both events and multiplying the probabilities together
      • Mod's argument was that this was extremely unlikely but they did mention how various biases can decrease this odd
      • Dream's counter-argument was using these sampling biases to recalculate the odds given that we are only seeing the runs he submitted
      • However, all of this doesn't matter
        • The one fact both sides agree on is that the simple probability of the item drops in Dream's submitted run had the odds of 1 in 2×10221 \text{ in } 2\times10^{22}
        • Regardless of how this sample was taken, it is an extremely unlikely event
          • The most unlikely confirmed event from chance based game was 154 rolls of non-seven in a craps game and that had the odds of 1 in 1.6×10121 \text{ in } 1.6 \times 10^{12}
          • If 10 billion people for every second in a century had a Minecraft run the chance of an observed event happening has the odds of 1 in 3.1×10191 \text{ in } 3.1 \times 10^{19}
        • So even if there was biases from sampling, the likelihood of this happening is still 1000 times less likely to happen than an 10 billion people century second event

      • Takeaway: The biggest takeaway I have is our inability to grasp exponential past a certain point. We need to ground the extremely large (or small) numbers in events that we can explain with words

  • Retrospective::
    • One week ago: [[May 24th, 2021]]
    • One month ago:
    • One quarter ago:
    • One year ago: [[May 31st, 2020]]
  • Daily Stoic::
    • The stoics believed that our job on earth is to be a good human being
      • "Do your job" - Bill Belichick [[quotes]]