3 ms·

This is with humans... using automated tools it's even less than 50% accurate.

• pclmulqdq 4y ago
  50% accurate is the worst thing possible on binary choices - it's equivalent to a random guess. If you are 25% accurate, inverting your answer makes you 75% accurate.
  • welder 4y ago
    But how do you know to invert your answer? You're assuming you know you're wrong.
    • sebzim4500 4y ago
      If you have an algorithm that is correct 30% of the time on some benchmark, then invert results and you have an algorithm that is correct 70% of the time. That's why 50% is the worst case result.
    • VincentEvans 4y ago
      You’ll know your bias if you’ve been tracking your success rate, and once you do - just keep doing the same thing, but use the opposite of your guess.
      • welder 4y ago
        So, if the average of your past results is under 50% then always invert your result? That makes sense, so you can never have less than 51% accuracy. That could still trend towards 50% though. Thanks for explaining it!