macroxue / 2048-ai Goto Github PK

Is it possible to fine-tune the AI ​​to increase the average score in a game with a different % chance of getting the tile 4 ? Let's say, for 20% chance of getting 4? Maybe some startup parameters or code changes?

Can I increase tuple table size to get more variants of moves? Now tuple_moves.10a have size 1315mb , tuple_moves.11a have 1215mb. On my machine I can run only small runs by command ./2048 because runs by command ./2048b fully load my ram(8gb) and after that process is killed. For example when I run small runs by command ./2048 -d7 2833 I get a score of 1556664, but in real games, the maximum result that I get from dozens of runs is around 400k. Can I improve my result somehow?

P.s (I use server mode to got right moves for real 2048 game)

I think if you split the lookup tables into each individual layer by the sum of tiles, you will need much less memory to create the large tables. This could allow 12 space lookup tables with two free rows and columns, which would allow for some much better strategies to appear

Hi,
Im completly new to ubuntu and i was wondering if you could help me.
I dont know what im doing wrong after "building".
How do i find/use the 2048 and 2048b you are talking about ?
Thanks you !

I wrote a Endgame Table Generator that generates DPDF-512 (L3-512) lookup tables in 4 minutes and only takes up 2GB . It can also take more than 10 hours to generate a 12-space lookup table (like LL-2048), which takes up about 200GB of storage under compression. It also provides a variety of 12-space lookup tables that can support more formations（like 444-2048 in about 60 GB playing snake）, and free lookup tables to make a dash in the final stage. If your AI can utilize these results, it should be a big boost!
link: https://github.com/game-difficulty/2048EndgameTablebase

Hi MacroXue, I am quite new to ubuntu and using github! I am just wondering if I can setp-up a pre-existing board, and ask the app what's the next best move? If yes which mode should I be using? Is that the AI mode? Thanks

Hi, Macroxue. THX for this amazing work. I have question for you. Can i run this script in mode when i send to console command right with massive like this [16,512,2,0,64,256,0,2,16,128,0,0,8,4,0,0] and got best direction? This is very needing function for me.

I'm trying to win the game on a device other than my PC.
Is there a way to input a specific state of the game into the solver and get a move suggestion?

To tell the truth, I don't fully understand all those clever words in description, but I used this repository many times to get first place in different games and game bots.

This solution probably the best one among other github repos.

So I just wanted to let you know that there is at least one person who very thankful for your work, best wishes!