AI for the 2048 game. This uses expectimax optimization, along with a highly-efficient bitboard representation to search upwards of 10 million moves per second on recent hardware. Heuristics used include bonuses for empty squares and bonuses for placing large values near edges and corners. Read more about the algorithm on the StackOverflow answer.
Execute
./configure
make
in a terminal. Any relatively recent C++ compiler should be able to build the output.
Note that you don't do make install; this program is meant to be run from this directory.
You have a few options, depending on what you have installed.
-
Pure Cygwin: follow the Unix/Linux/OS X instructions above. The resulting DLL can only be used with Cygwin programs, so to run the browser control version, you must use the Cygwin Python (not the python.org Python). For step-by-step instructions, courtesy Tamas Szell (@matukaa), see this document.
-
Cygwin with MinGW: run
CXX=x86_64-w64-mingw32-g++ CXXFLAGS='-static-libstdc++ -static-libgcc -D_WINDLL -D_GNU_SOURCE=1' ./configure ; makein a MinGW or Cygwin shell to build. The resultant DLL can be used with non-Cygwin programs.
-
Visual Studio: open a Visual Studio command prompt,
cdto the 2048-ai directory, and runmake-msvc.bat.
Run bin/2048 if you want to see the AI by itself in action.
You can use this 2048 AI to control the 2048 browser game. The browser control capability is meant as a proof of concept to show the performance of the AI; it will only work on the original 2048 browser game or any compatible clone, not all 2048 games.
Enable Firefox remote debugging by setting the about:config options "devtools.debugger.remote-enabled" and "devtools.chrome.enabled" to true, then quit Firefox and restart it with the --start-debugger-server 32000 command-line option.
Open the game in a new tab, then run 2048.py -b firefox and watch the game! The -p option can be used to set the port to connect to.
Enable Chrome remote debugging by quitting it and then restarting it with the remote-debugging-port command-line switch (e.g. google-chrome --remote-debugging-port=9222).
Open the game in a new tab, then run 2048.py -b chrome and watch the game! The -p option can be used to set the port to connect to.
You can also use 2048.py interactively using 2048.py -b manual. In this mode, you'll be asked to input the board, after which the AI will give its suggested move. This might be useful for getting hints while playing the game on a platform without autoplay (e.g. on a phone), or for getting the AI's analysis of a given situation.
After each recommendation, the AI will assume you make that recommended move, and then prompt you to make any necessary adjustments to the board (usually, the location and value of the newly spawned tile) before giving its next suggestion.
Sample run:
Enter board one row at a time, with entries separated by spaces
Row 1: 16 128 256 1024
Row 2: 16 8 2 0
Row 3: 8 2 0 0
Row 4: 0 4 0 0
Current board:
16 128 256 1024
16 8 2 0
8 2 0 0
0 4 0 0
Enter updates in the form r,c,n (1-indexed row/column), separated by spaces:
16 128 256 1024
16 8 2 0
8 2 0 0
0 4 0 0
005.030340: Score 0, Move 1: up
EXECUTE MOVE: up
Current board:
32 128 256 1024
8 8 2 0
0 2 0 0
0 4 0 0
Enter updates in the form r,c,n (1-indexed row/column), separated by spaces: 3,1,4
32 128 256 1024
8 8 2 0
4 2 0 0
0 4 0 0
035.648508: Score 0, Move 2: left
EXECUTE MOVE: left
Current board:
32 128 256 1024
16 2 0 0
4 2 0 0
4 0 0 0
Enter updates in the form r,c,n (1-indexed row/column), separated by spaces: 4,3,2
32 128 256 1024
16 2 0 0
4 2 0 0
4 0 2 0
058.927319: Score 0, Move 3: left
EXECUTE MOVE: left
This tells the bot that after the first move, a 4 spawned in the 3rd row, 1st column, and after the second move, a 2 spawned in the 4th row, 3rd column.