Prototype & Proposal: Multiplayer Asteroids

[aerique]

Or "space rocks challenge" if Atari complains.

o http://youtu.be/9VeWz24_VHk
o Extensive description: http://github.com/aerique/aichallenge/t ... ids#readme

Inspired by this post: viewtopic.php?f=21&t=2122 I spent a couple of evenings hacking the ants code base to turn it into a multiplayer asteroids engine: http://github.com/aerique/aichallenge/tree/asteroids

There's enough issues with the code, the engine and the sample bots but it is possible to start writing your own bot. Sample and starter bots from the ants challenge in any programming language should be easy enough to adapt for the asteroids setting.

The main point now is to see whether the idea has any merit. This should be possible with the current state of the engine.

Taking into account the idea and the expected complexities my proposal is for this challenge to be a smaller one spanning four to six weeks and to be held in the late spring or early summer so a more complex and longer challenge can be planned and worked on for the end of this year.

The README at the GitHub page (linked at the top of this post) has more specific information about the game's parameters, client server communication and possible extras once the beta has ended.

[GreenTea]

As I can see from video visualizatoin of game is continuous. But we need stepped game right? What will be time interval between bot moves? And second question which comes from previous, how much steps (game time) approximately need to find winner?

[cowtippervirus]

Taking into account the idea and the expected complexities my proposal is for this challenge to be a smaller one spanning four to six weeks and to be held in the late spring or early summer so a more complex and longer challenge can be planned and worked on for the end of this year.

This seems really surprising to me. What would prevent this from being a full challenge? I think this has enough variety (different maps with fixed obstacles, different numbers and compositions of asteroids, different numbers of players, etc) to make for an interesting competition. Maybe the contest organizers have a different opinion than me, but I don't think subsequent contests should necessarily be harder than the last.

[aerique]

As I can see from video visualizatoin of game is continuous. But we need stepped game right? What will be time interval between bot moves? And second question which comes from previous, how much steps (game time) approximately need to find winner?

The game is stepped, the steps are just small enough that it looks continuous when visualized at 25 turns per second

To put it another way: the turns works the same as in Ants or Planet Wars which would look continuous as well if visualized fast enough.

The answer to your second question: I don't know. When developing the prototype and using the totally stupid sample bots it depended on the map size and was anywhere between 50 turns and the square root of map_width x map_height.

[aerique]

This seems really surprising to me. What would prevent this from being a full challenge? I think this has enough variety (different maps with fixed obstacles, different numbers and compositions of asteroids, different numbers of players, etc) to make for an interesting competition. Maybe the contest organizers have a different opinion than me, but I don't think subsequent contests should necessarily be harder than the last.

You might very well be right, it is just my uninformed, personal opinion.

[aerique]

I just merged some patches from berak which add proper replay output, a first step at a visualizer and scoring.

[cobracom]

I would be interested to explore the idea of having two ships for each player. I think it will be add an interesting twist to the game. A few years back I was following JRobots (http://jrobots.sourceforge.net/jjr_match.shtml) and the double matches were usually the most interesting ones (they also have team matches and single matches).

[MagnumPU]

I would be interested to explore the idea of having two ships for each player. I think it will be add an interesting twist to the game. A few years back I was following JRobots (http://jrobots.sourceforge.net/jjr_match.shtml) and the double matches were usually the most interesting ones (they also have team matches and single matches).

What if the game had levels, like maybe a maximum of 3. You get one ship at the first level, and if you're still alive at the end of the first level game, you and every other player that is still around will 'level up' by getting one additional ship each. Hopefully, a good bot would have to be good at coordinating multiple ships against various numbers of enemies. I think each bot would still get a single core.

[akarukov]

I can't visualize a game. Can I hope for any changes for this issue in nearest future?

[smiley1983]

The visualize-game.sh script is no longer correct. Try this (under linux) :

visualizer/visualize_locally.py < game_logs/0.replay

or

cat game_logs/0.replay | visualizer/visualize_locally.py

You may need this first:

chmod u+x visualizer/visualize_locally.py

I like this game idea, would be very happy to see a contest based on it at some point.