Author Topic: perfect equilibrium  (Read 2667 times)

Offline ikke

  • Bot Destroyer
  • ***
  • Posts: 300
    • View Profile
perfect equilibrium
« on: December 27, 2009, 12:32:44 PM »
Absent other f2 & f3 bots that can compete I run matches between variants of Gimmick, usually differing only in one line of code, or even just in parameters. Two days ago I started to see which variant of the two I just made was better. First competition came in about 30-60 for one. For some reason I decided to run a second. It came back 60-30. So I ran a third series. At 29000 still no winner

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
perfect equilibrium
« Reply #1 on: December 27, 2009, 04:04:28 PM »
That's really weird.  But I guess it tells you that the two bots really are identically matched.  Although I did some simulations with coin tosses and a winner was still eventually declared.  I'd have to dig it back up to find how long it would take.

Basically, to win, a bot needs sqrt(number of matches) + number of matches / 2.  Asymptotically it approaches just number of matches / 2.