New Contest - game like Risk

[GreenTea]

There is another topic about this
viewtopic.php?f=21&t=1361&hilit=Risk
but it is old, better to create new one.

So, I want to reopen discussion about next context game.
I think game like Risk, will completely satisfy all criterias:
https://github.com/aichallenge/aichalle ... e-Criteria
Easy
- this is not scary problem.

Familiar
- many people played this game at least once, or at least hear about it.

Interesting
- As for me it is very interesting. I also know it has many fans.
- many possible orders which bot is able to do.
- prediction of oponent moves / learning during game.
- has big article on wikipedia http://en.wikipedia.org/wiki/Risk_(game).
Also this game has many modifications, we can chose more interesting or invent new one.
For example - generate custom maps (not only earth map).

Technically Feasible
- to determinate winner on medium map game need maximum 20 moves, on big maps it can be 100 moves - maxiumum.
So if let 1 second on move, game rate can be VERY big! Just imagine 10 times more games than in Ants..

Non-trivial
- for simple maps like clasic risk earth - game possibly can be trivial. But on big graphs, with many wide places - i don't think so.
- bot has many posibilities to move. From territory to any other near territory, and from 1 to N armies count, where N is armies on territory.
And sequence of moves also very important. This is huge branch factor, I can't say quick which algorithm can be used here.

Fun to Watch
- risk games are very visual by definition. In case of 2 even player it can be very interesting to watch.

Possible problem is random factor in battle resolution.
Solutions:
1) We can remove it - provide strict formula to calculation of battles.
2) It can be not so big problem because of large amount of games played. Soon, in spite of randomness, better players will be on top.

[delineate]

I also quite like the idea of a Risk-like game. I think the game should be simplified/changed in many ways to make it more clean (and general).

There can be many different "maps" (which really are just different planar graphs):

Here's one way the game could work:

Initialization: Every player is given the "map" of the world along with game conditions (starting number of armies, turn order, etc.)

Initial placement phase: Each player in turn starts claiming countries until all countries are claimed. Then reinforcements are placed in turn.

Main game loop:
Players do the following (in turn):
-Receive and place reinforcements (based on the countries they currently own)
-Attack neighboring countries (if they wish) repeatedly until they want to stop
-Rearrange their existing armies

Play continues until only one player remains (or the turn limit reached).

There are many details to be worked out. When rearranging armies, can players only move them 1 country at a time? How are reinforcements calculated? Is there a "bonus" if players own an entire "continent"? Perhaps instead territories can just be given different values (similar to planet wars). That should create plenty of complexity to work with.

The key differences from planet wars are:
1) there is an actual map so territories / boundaries are important
2) the battle resolution can be stochastic
3) it's turn based
4) there's the whole pre-game placement phase (which is a deep problem in itself)
5) it's multi-player

[Scryer]

I support this product and/or service. It does indeed score high on interesting visuals, multi-player, and reasonably low entry threshold. I like the idea of making custom maps and several of the other changes. Dealing out random "cards" to add armies as in the board game may be problematical -- perhaps that could be play-tested during alpha.

I think making combat deterministic would damage the game, though. If we're keeping the alternating turns of the board game, then that would mean you could know before you move what you'll be able to capture, and how many armies your opponent will have to transfer to capture it back, and so on. The dice-rolling adds a little fog of war to keep the minimaxing down a little. However, if there are good ideas for resolving combat differently, those could be play-tested also.

Perhaps if we worked out a way to do simultaneous moves, as with Ants, we could do combat resolution without dice too.

On the other hot ideas:

- The asteroids idea doesn't really grab me. Maybe the physics wouldn't be too much of a drain on the server, but I'm not sure the result will be visually very interesting. Is there a video of a similar multi-player Asteroids game lying around?

- I agree that the transport fleet idea doesn't look visually interesting enough.

- If Haxball is really inherently 2-player I think it wouldn't be as much fun as the ffa style of Ants.

[Riseven]

I agree that this is the most visually interesting idea so far. I also think that making all players play simultaneously can be a good idea since it removes the necessity for randomness in combats, and allows the game to move faster.

An example of a simultaneous combat system could be:

1) Every player decides to move from each of his territories to each of its neighboring territories (including enemy controlled ones).
2) Resolution of travelling conflicts: If there are units of one player travelling from A to B, and units of another player travelling from B to A, they fight in a 1:1 ratio. If player 1 was moving 5 units, and player 2 was moving 3 units, then 2 units from player 1 survive and continue their move to the destination territory.
3) Resolution of territory attacks: Those could be resolved by a different ratio like 1:2 or 2:3 that benefits the defending player.

I think this game is also interesting from a theoretical point of view because of the game theory implications. The relation between two neighbor players ressembles that of the Prisoner's dilema. Would be interesting to see bots that learn whose neighbors are peaceful to them and decide to maintain a peaceful relation to concentrate on a third enemy.

Another interesting strategy is where 2 players make a pact to swap a territory continuously. Each player conquers the territory and leaves it unprotected so that the other player can recapture it easily. This way both players can have easy access to cards. Would be interesting too to see if bots manage to arrive to these arrangements without an explicit communication channel.

[GreenTea]

I don't like idea of simultaneus moves because it very-very simplify the game, and make it flat.
Moves ordering is often 80% of success to win, and this is thing which distinguish good player from bad player.

[dimkadimon]

I love this idea. I think it is very visual, fun and the entry barrier is not too high - even a random player can do quite well. If we need to make the game more complex, we can also add a fog of war - so you don't know the whole graph only the subgraph that is visible to you. I think it's quite important to keep the battle resolution random, otherwise the game dynamics become quite different. The original game had missions and sometimes you can win a game easily by finishing a mission, so do we want to have missions? I think it is best to avoid missions for now.

[grc]

[stupidquestions]

I like the idea... Anyway, I think the battle resolution should be partly random, but depend on something. For example, troops can have experience and/or moral, depending on how many battles they survived or which territory they are fighting for or something like that... So the troops with higher experience and/or moral would have a higher chance of winning.

[Corniel]

I don't like this idea. When good players (or bots) play it, no one will win. Just because there is only ONE strategy: Can I win? Yes, do it, no? Attack strongest player.

Futhermore, this game heavily hangs on a randomizer. There are variants of Risk without dices, but there are not better, and with dices/randomizer, some people will reverse engineer the server code, and predict perfectly what will happen randomly.

[GreenTea]