Draft

Lets play Guess Who

1LosBlobbos
Description
Loading description...
Games
View
AllIssues4QuestionsSuggestionsShow Resolved
  • Please sign in or sign up to leave a comment.
  • Voile Avatar

    If we're using the same set of characters and characteristics from Preloaded anyway, is there any reason to pass in persons into GuessWho? It'd imply that it's possible to only have a subset of all characters for guessing, which does not occur in the tests.

    • LosBlobbos Avatar

      Aye I did start with using the preloaded then I had an idea of some different tests where I needed to be able to change it. But those tests were insanely bad so...

  • Voile Avatar

    In all the tests the opponent character always have Male characteristic for some reasons. Is this a dev exit?

    • LosBlobbos Avatar

      It's because there are more male cards so it gives the player a bit of an advantage.

    • Voile Avatar

      It doesn't change the optimal amount of maximum guesses needed anyway, so there's no point in doing that.

      What's worse, this is a hidden knowledge that artificially changes the kata's difficulty from what's stated, so it's definitely a no-go.

    • LosBlobbos Avatar

      Lol I made a very naive assumption it seems and didnt do my calculations correctly. Turns out the difference between male and female charcteristics is 17 to 15 so a useless advantage.

  • Voile Avatar

    The tests as they are designed right now is unsolvable, as it is unclear what a solution should optimize for: Average guesses? Minimum guesses? Maximum guesses? Number of guesses for specific match-ups?

    The tests current does this:

    • Implements a not strictly globally optimal player
    • You have to be better than this player on almost all inputs to pass

    This basically forces a solution to be implemented the same way as the reference solution, or be strictly superior to it.

    • Voile Avatar

      Also, in any case, there exists a solution with at most 5 guesses across all possible characters (it's in fact very easy to derive it by hand), and the lower bound is 5 for obvious reasons. So this kata is basically another "optimal solution search" problem with a fixed input, like this.

    • LosBlobbos Avatar

      Tbh that is basically what GuessWho is.
      I wanted to make the computer (good one, not the dunce) actually play the game and ask valid questions but without using any strategy.
      unfortunatly if not using strategy it is a game of chance and I couldnt find a way to garantuee that the Kata solver would win everytime unless they also use a rather good strategy

    • LosBlobbos Avatar

      Any suggestions how I could allow less optimized solutions but still require more than just chance?

    • Voile Avatar

      You can choose one of the metrics above (usually it's either average guesses or maximum guesses), depending on what you want.

      You'll need to test every option to calculate these metrics anyway, so randomizing the test order is better than just choosing an element with replacement randomly every time.

    • LosBlobbos Avatar

      I'll see if I can find a solution that still holds with my intentions with the Kata.

  • Voile Avatar

    Kata design problem: We're required to implement the player logic correctly, running games and maintain game states, and play optimally the same time. Why are we doing 3 unrelated tasks together in one? The kata would sound like we only need to do the last one, so having to do the other two as well is very off-putting. Not to mention the current interface is ill-designed for doing this, as game, player and controller logic should belong to different objects.