The Reformed Dev's September Challenge

The TRD challenge for the month of September is to create an AI that can win a game of minesweeprs. The winner will be whoever can create an AI that successfully completes all game boards given in the fastest average time. Any game boards that are not successfully completed will exclude the AI's times from the ranking. The person with the fastest AI at 23:59:59 on 2021-09-30 will be the winner of the September Challenge.

Each AI will be run five times, against five separate game boards. The same five game boards will be used for all entries.

Entry Requirements

PLEASE NOTE THAT I HAVE NOT YET FINALIZED THE KUBERNETES MANIFEST SPECIFICS. THIS SHOULD NOT IMPEDE YOUR ABILITY TO DEVELOP YOUR AI, MERELY YOUR ABILITY TO RUN OFFICIAL TIMES.

To submit an entry, create a PR against this repository. Copy the example.yaml File in the ./deploy/entries/ directory. Rename the file to your GitHub username and [TODO]. Your entry will need to be bundled in a Docker image. Please open an issue in this repository if you require assistance with either the Docker image creation or setting up the Kubernetes manifest for your entry.

Each entry should accept either the following ENV or arguments:

SERVER or --server
WIDTH or --width
HEIGHT or --height
DIFFICULTY or --difficulty
SEED or --seed

This input will be used to construct the full URI to get the current round's game board from the API server. Your AI will then play the game and then return the following JSON:

'elapsed_time_in_ms': 123456789,
'final_game_state': '<GAME STATE WITHOUT NEWLINES>'

Please notice that the final_game_state should not include any newlines.

When measuring time, please start your timing as the very first possible line in your program, and end your time as the very last line before returning. The only line other than your return statement which is allowed to follow the end of the timer is the creation of the JSON. Please do not make any function calls in your return statement. In python the above might look similar to:

def main():
    ...
    
    
if __name__ == "__main__":
    start = time.now()
    x = main()
    elapsed = time.now() - start
    print(json.dump({...}))

There will be a hard cap on run time, as well as CPU and mem limits. I have notice determined these yet, but I will update this when it was been determined. I would appreciate it if you did not knowingly submit entries which consistently run past the upper bounds of the time limit, as this is just wasted cycles on my cluster.

Tests

Every group of tests will be made up of five unique game boards, and each will be run five times. The game board configurations will be as follows:

• 1 run @ 8x8 w/ 10 mines
• 1 run @ 16x16 w/ 40 mines
• 3 run @ 24x24 w/ 99 mines

This means your AI will be run 25 times and those times will be averaged for the final result. The same five game boards -- including the same seeds -- will be used for the entire month.

Minesweeper API

For the minesweeper logic we will be using the mineswepttd server. This is a stateless, RESTful server that handles the logic for the game of minesweeper. The server will be bundled with your AI when it is tested, you do not need to worry about implementing the server.

If there are any issues with the server, please submit an issue to the repo for the mineswepttd server and I will address it there.

Community

In an effort to foster a fair and enjoyable time from both the TRD Slack and the TRD Facebook communities, please use the GitHub issues and the GitHub discussion areas to communicate regarding this challenge when applicable. For TRD members who only are on Facebook, the finer points of the monthly-challenge infrastructure management wil be conducted on the TRD slack instance, feel free to reach out for an invite if you would like to contribute to the overarching infrastructure needs.