Ai game competition




















How would you create a single agent that is able to play any game it is given? Could you program an agent that is able to play a wide variety of games, without knowing which games are to be played?

Can you create an automatic level generation that designs levels for any game is given? You can now download the starter kit for the competition and submit your controller to be included in the rankings. For any question contact us. Join our Google Group here for the latest updates. If you are interested in studying for a PhD in this area, you can have a look at our Centre for Doctoral Training. Also, consider joining our MSc in Computer Games. Getting started Get the Code. Participants write bots using the programming language of their choice to compete in an original online multiplayer game.

Since , more than 15, people around the world have participated in a Halite challenge, applying advanced algorithms in a dynamic, open source game setting. The strategic depth and immersive, interactive nature of Halite games make each challenge a unique learning environment.

Halite IV builds on the core game design of Halite III with a number of key changes that shift the focus of the game towards tighter competition on a smaller board. Submitting Agents We require entrants to submit the source code of their agents -- not just compiled agents. Any instructions for compilation if any non-standard steps are required if compilation steps are extraordinarily difficult and we fail to compile your agent, it will be disqualified -- try to make sure to submit early such that we have time to test!

Which class to instantiate for your agent if your source code includes multiple classes that extend the AI abstract class. Name of your agent. Competition Rules 20 games will be selected by the competition organisers, and not revealed to any participants.

Submitted agents will be required to play each of these games against every other agent, as both the first and second player. Depending on the number of submissions we receive, this process may be run multiple times to give a more robust result.

The win-rate for each agent across all matchups that it plays will determine its final ranking. Draws count as half a win for each of the two players. For each of the 20 games, the competition organisers will pick an amount of thinking time between 1 and 5 seconds. All entries in all matchups using this game are allowed to use this many seconds per move. Agents will have seconds five minutes to initialise themselves between when a new game starts, and when they must make their first move setup time.

All entries must function on Linux. Entries are permitted to write files to their current working directory, but these will not be preserved between match-ups i. We will enforce a turn limit, after which matchups are declared a draw, to avoid infinitely-long matchups. This limit will be sufficiently high that it shouldn't be reached by "reasonable" play. The forward model provided to agents will be aware of this limit, which means that forward planning agents will automatically be aware of the existence of this limit without requiring any extra implementation effort for entrants.

Each entrant may only win a single prize, and must outperform a baseline UCT agent. Releases No releases published. Packages 0 No packages published. Contributors 3. You signed in with another tab or window. Reload to refresh your session.



0コメント

  • 1000 / 1000