hasyimibhar / tetris-ai

Various Tetris AI implemented in C++. The code has been fine-tuned to avoid unnecessary copying and memory allocations.

Tetris rules used

Tetris varies a lot between implementations. The following rules are used:

• 10x20 board
• 1 piece at a time, 0 piece lookahead, no holding piece
• Hard drops only
• Piece starts from row 0 (some implementations use hidden row 21-23)
• Game is over when piece overflows row 0

Randomizers

• uniform: Randomly pick any of the 7 piece using a uniform PRNG
• 7bag: The Random Generator
• nes: NES randomizer, but LFSR is only scrambled on piece generation (in original NES Tetris, LFSR is scrambled many times in between spawns)
• nesApprox: NES randomizer, but approximated as first-order Markov process

AI

• eltetris: El-Tetris
• yiyuan: The (Near) Perfect Bot

Implementing your own AI

Each AI implements the following function signature:

DropMove yourOwnAI(const Board &board, PieceType piece) {
  // do stuff
}

Usage

Compile:

$ make

Usage: tetris [options]

Optional arguments:
-h --help       	shows help message and exits [default: false]
-v --version    	prints version information and exits [default: false]
-a --ai         	AI to use [default: "eltetris"]
-r --randomizer 	randomizer to use [default: "7bag"]
-s --seed       	RNG seed [default: 0]
-p --pieces     	Number of pieces to generate [default: 1000000]

Example:

$ bin/tetris -a eltetris -randomizer 7bag -s 123 -p 100000
ai=eltetris
randomizer=7bag
seed=123
pieces=100000

10000
20000
30000
40000
50000
60000
70000
80000
90000
pieces=100000, lines cleared=39999
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
..........
.O..OO....
OOOOOO.O..