This is a Rust implementation of the famous 2048 game which runs in the terminal.
In addition to the standard user inputs, an AI can be used to perform the moves. This AI leverages the expectiminimax algorithm to recommend the best next move at each step.
Here are the statistics of the AI with its default parameters:
| max tile reached | frequency | average time to reach max tile |
|---|---|---|
| 2048 | 100 % | 12s |
| 4096 | 100 % | 23s |
| 8192 | 96 % | 74s |
| 16384 | 70 % | 202s |
| 32768 | 10 % | 673s |
Average times have been computed on a macbook pro 3,5 GHz Dual-Core Intel Core i7
In particular, these statistics correspond to a minimum branch probability of 0.001.
Decreasing this value would lead to better performance, as more branches would be explored, but this would also take more time.
Some other hardcoded parameters can be tweaked in order to further improve the algorithm. I have not performed an exhaustive grid search, thus the parameter set is probably sub-optimal.
Rust and cargo must be installed to run this game:
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | shcargo install play-2048Then, in order to play:
play-2048> git clone https://github.com/adrienball/2048-rs.git
> cd 2048-rs
> cargo run --releaseYou can change the probability of drawing a 4 tile:
> play-2048 --proba-4 0.5Or adjust the minimum branch probability of the expectiminimax search:
> play-2048 --min-branch-proba 0.0001To get the list of available options, simply run:
> play-2048 --helpAll original work licensed under either of
- Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT) at your option.
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.