Game Idea: RoboRally

[Karlimero]

A game, based on the idea of this board game
http://en.wikipedia.org/wiki/RoboRally

This game is great fun. Program your robot 5 move-steps in advance to achieve the checkpoints and to disturb your enemies route.
Benefit from factory facilities like conveyor belts or destroy other bots with a laser.

I think with a few changes in rules it can become a great online bot game.

[EDIT]
Short Video how the Game looks like.
http://www.youtube.com/watch?v=8qpLEt9j0wM
With a good visualizer (lasers, bombs, emp, traps, beamer, ...) i think it could be nice looking

[delt0r]

My only problem with these types of games is they just become a min/max type tree problem. In other words throw enough CPU at it and you have close enough to optimal.

Of course there is also a lot of randomness in a game like this. This means more matches before you get a stable ranking order.

[Karlimero]

Could be possible, yes.
The thing I found interesting is to plan always 5 moves in advance without the chance to revert. So I think you have to learn and cater? to the others strategy to guess his next steps.

[EDIT]
Make weapons strong enough to really hurt your enemy if you guess his position in 5 steps and optimize your min/max-tree is not the only thing you have to do.

[delt0r]

Its still a min/max tree problem with weapons, can't see how that changes the basic problem. Sure its high branching ratio, but again pruning and plenty of cpu will still beat anything else. The randomness makes it hard to rank and hence requires more servers and stuff.

[Karlimero]

The change in my opinion ist, that with the planning in advance you have a Paper, Scissors, Stone problem.

[delt0r]

I don't understand what you are saying. Adding weapons does not make anything a RPS problem, its still a min/max tree pruning problem. With a lot of randomness.

[Karlimero]

Weapons are not the point.

In this game you mainly have 2 things to do.
1. Reach your goal (Simply a min/max problem or even only a pathfinder)
2. Guess _how_ your enemy reaches his goal

And: Be unpredictable in point 1, so I think its not a simple min/max problem anymore

[Memetix]

Great idea - I haven't played this game in ages but I do recall it was a lot of fun and there are a lot of unknowns. Writing an AI would be really interesting.
I believe there is enough complexity in the game to ensure a simple minimax algorithm wouldn't "solve" the game.

One thing that isn't obvious is how to manage risk in the game. Being out in front means less risk of other robots interfering with your movement plans but have greater risk of getting hit and therefore having less movement options on the next turn.

[delt0r]

That is what i mean by min/max tree. The max part is you maximize your gain, while minimizing your opponents assuming opponents attempt to maximize their gain. It works for all parts of the game. Anywhere where you have a discrete number of options with a branching factor not too high this method applies and typically can work well.

But its less interesting since really he who can write the most efficient code wins rather than the smartest code. In other words it becomes an optimization problem. This is what happened with tron more or less.

[Karlimero]

while minimizing your opponents assuming opponents attempt to maximize their gain.

And here I think is one of the challenges in the game. By default this is a endless recursion.