1

I am building a 2048 AI, and it is leading to a rather peculiar observation (peculiar enough to me).
The optimizations are not up to the mark right now (coupled with the fact that the code is written in python), which is letting me reach till only a depth of 3 moves (plies).

Summary after playing some games using both the algorithms

As evident from the results, Expectimax is quite dominant over minimax (similar results can be seen without alpha-beta pruning in minimax) in terms of results produced. Both use the same evaluation function and do not proceed any further than 3 moves. AFAIK, minimax should work optimally in such games, but that doesn't seem to be the case here. My question is, this observation is due to the fact that:

  1. I am not going deep enough into the search tree?
  2. 2048 is a stochastic game, and that is hampering the performance of minimax (or boosting the performance of expectimax)?
  3. The opponent (the 2048 game logic) is not playing optimally (90-10 % chance of putting a 2-4 tile, random adversary) (if yes, then why should this affect the performance of minimax)?
  4. Anything else that is not apparent to me?
Tanmay Garg
  • 801
  • 11
  • 20
  • 1
    (1) Not traversing the tree completely results in losing all optimality-guarantees for minimax. (2) (Classic) Minimax is designed to minimize the possible loss for a worst case. So if you want to optimize play for the expected score, it's probably not what you want. (3) It's questionable to use Minimax on some game which is not 2-player-zero-sum. Maybe it's legit with a special-design but i can't give any info on that (4) I assume there are tons of 2048 AI questions here on SO which can provide a lot of info! – sascha Apr 27 '17 at 14:14

0 Answers0