Add a variable associated with each planet which increases each time a planet exchanges hands and decreases slowly overtime. Adjust your algorithm to make planets with high values on this variable less desirable. You might also consider to add some way of estimating score/how well other players are doing and make AI somewhat favour attacking planets of a player who is ahead to stop players steamrolling ahead easily.
AI conquest deadlock
Author
I think, going into direction of "even more smart" is pointless... Even if I manage to fix it and make it work, is it really fun? This Spock like strategy :)
Maybe go somewhere into realm of emotions? AI being afraid, hating someone, wanting to take revenge...
As a quick fix I guess I will make some sort of "tired" counter, it increases each turn you fight over that planet (and decreases slowly over time). If you are too tired fighting over a certain planet (reach certain thereshold) it gets desirablity halved/nullified. With exceptions, like you never will get tired defending your homeworld :D
Smart & cowardly races would have that thereshold lower (quicker to back off), which adds personality to aliens I guess.
Maybe your AI is perfect indeed, but the game mechanic needs some tweaking.
1.Usually the defender has advantage and keeping a position is easier than winning it.
2.if the attacker won the planet, then he should have even more advantage, because he had more power and now owns more resources to extend the power.
how can this even happen?So the effect is 2 AIs fighting over like 2-5 super important planets (where ownership changes all the time
1.Usually the defender has advantage and keeping a position is easier than winning it.
2.if the attacker won the planet, then he should have even more advantage, because he had more power and now owns more resources to extend the power.
Author
Maybe your AI is perfect indeed, but the game mechanic needs some tweaking.
how can this even happen?So the effect is 2 AIs fighting over like 2-5 super important planets (where ownership changes all the time
1.Usually the defender has advantage and keeping a position is easier than winning it.
2.if the attacker won the planet, then he should have even more advantage, because he had more power and now owns more resources to extend the power.
Long term, possibly. But when a planet changes hands every single turn it's irrelevant (both resources gain and ability to build defences).
I think, going into direction of "even more smart" is pointless... Even if I manage to fix it and make it work, is it really fun? This Spock like strategy
Maybe go somewhere into realm of emotions? AI being afraid, hating someone, wanting to take revenge...
As a quick fix I guess I will make some sort of "tired" counter, it increases each turn you fight over that planet (and decreases slowly over time). If you are too tired fighting over a certain planet (reach certain thereshold) it gets desirablity halved/nullified. With exceptions, like you never will get tired defending your homeworld
Smart & cowardly races would have that thereshold lower (quicker to back off), which adds personality to aliens I guess.
An alternative approach, which depends strongly on game rules (fleet movement has to take enough turns): send a reasonably sized fleet to the very good planet, but reroute it to another nearby objective (or to retreat in extreme cases) if the good planet is too well defended, or expected to be when the fleet arrives. This way every fleet fights against appropriately small defenses, or retreats with little harm, instead of suffering unusual losses.
In case two large invasion forces meet, instead of mutually annihilating for the benefit of all other factions they would stop, conquer what they can from neutral parties and other sides, and gradually send away excess ships from places they don't want to fight at.
I think, going into direction of "even more smart" is pointless...
Well, your AI is very smart within the parameters it evaluates...
But is the end result really that smart?
How rational is it really to grind billions into a never-ending war that gain nothing?
I'd say... not really.
It sounds a bit like your "perfect AI" need to read "The Art of War" to start with 
My point is just that your AI need to take more parameters into consideration, and that would mean being "smarter", not trying to make it less smart. Lorenzo again gives good examples of what these parameters could be.
Other parameters could be that which we call "emotions". A never-ending war will for example likely run your troops morale into the ground, which makes your war even more costly and pointless (unless everyone is robots, then I guess a never-ending war is plausible, but still sounds more like a programmer bug on the part of the creators of said robot civilization). Technically, this is just a float somewhere that get reduced every turn the war continues and modify the "importance" of this planet, but it is more fun and useful to think of it as "morale"
Another way to say it, I guess, is: The game should be balanced such that stupid and boring strategies simply isn't perceived as advantageous. Then the AI won't pick them.
THX1138 solution : Once the objective costs more than what might be recovered by achieving it you give up.
Recalculating the choice of target to see if 'resistance' was encountered and having the new decision reflect that (as it should have the original cost in time and other resources).
Distance to the target might have been part of the calculation (depending on your game mechanics and whether it actually amounts to 'cost'). Most duplicated targets likely WONT be the same distance from either side...
Having a fuzzy calculation that classifies 'best' targets with a little more slop/roundoff (the outcomes calculation cant be wholey mechanical so there is an 'unknown' potential that might have the different side picking different targets -- a 'quantum' judgement 'set' of targets that is then randomly selected from...)
Also so you dont allow for a variability of your 'strategies' (again game mechanic dependant) where slightly different emphasises in the calculations dont result in somewhat different choices ???
Randomness can do a lot, but like others have suggested, there's also a lot to be said for a variety of personalities, emotions, and strategies. There's also the subtle differences of layout that can lead two civilizations to put high priorities on different locations, like proximity to the home world or other governing planet. I imagine you can still give that high degree of variety without a lot of randomness. Randomness can help simulate the rest of that in a simpler way, though.
Author
Actually, most are... It was surprising to me too.
Most duplicated targets likely WONT be the same distance from either side...
For example the degenerated case when 2 homeworlds are generated next to each other (I know I can fix it on the homeworld placement stage but still...). They fight each other attacking their homeworlds forever (taking all planets surrounding your homeworld are extremelly important for security, and no one wants to back down here...)
shouldn't there still be the powerful forces that won the fight over the planet? and wouldn't a smart AI value to keep an existing planet over conquering a new one, and therefore send those forces that would otherwise invade a new planet to defend the one it has?Long term, possibly. But when a planet changes hands every single turn it's irrelevant (both resources gain and ability to build defences).Maybe your AI is perfect indeed, but the game mechanic needs some tweaking.
how can this even happen?So the effect is 2 AIs fighting over like 2-5 super important planets (where ownership changes all the time
1.Usually the defender has advantage and keeping a position is easier than winning it.
2.if the attacker won the planet, then he should have even more advantage, because he had more power and now owns more resources to extend the power.
I think that's what humans would include into decision making.
maybe you could let some human vs human play the game and observe whether they run into the same problem?
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement