Page 4 of 4
Re: A.I. behavior: Realism or Performance?
Posted: 17 Sep 2009, 09:43
by hughperkins
.... you have a gf that plays RTS games with you???
Unfortunately she prefers AOE3, but she quite likes Tiberium Wars.
She's not too keen on Spring actually, though if there's a balancing mechanism that might help. The issue with Spring is that it's intuitive like walking backwards holding a mirror in front of you is intuitive... I mean you can do it, but it takes a bit of practice, and I bet you walk into a load of lampposts first before you get the hang of it...
Even simple things like showing los by default. I mean, how do you explain to her that my units appear from nowhere? In AOE and so on, it is simple because that bit is totally black. ctrl-L works ok, but that means I have to tell her to hit ctrl-L at the start 'just because it will work better that way'.
Anyway, getting off topic I think :-DDDD
Re: A.I. behavior: Realism or Performance?
Posted: 17 Sep 2009, 10:01
by hoijui
hmm... i see.
I had the same "problem" with a friend of mine. It took about half a year of playing spring about every second week for some hours, till the awesomness he expierienced playing it > walking into lampposts issues.
about your gf ...
i dont have a car, nor a house, a job, friends, a live ...
but you can tell her i am a spring dev, and i can change eeeverything i/she want(s)!

"everything?"
eeeeeverything!
sorry, i just had to do this! :D
only work and no joy, you know

Re: A.I. behavior: Realism or Performance?
Posted: 17 Sep 2009, 12:38
by AF
You can improve scouting performance by guessing that the enemy will be on the opposite side of the map as you, so why bother scouting all the bits in between right at the start?
I thought there was a scouting algorithm I forgot about lol, I actually implemented this and set a time limit so scouters would randomly visit starting locations if built within the first few minutes of the game, IIRC it worked out alright until something changed in the interface ages back.
Re: A.I. behavior: Realism or Performance?
Posted: 18 Sep 2009, 10:59
by 1v0ry_k1ng
im against resource hacks. if i want an AI to have a resource advantage, I give it a handicap in the lobby.
frankly, while scouting etc might be an interesting idea, its just not possible to make a no-resource-hack AI that can match up to a human without the maphack. KAI at its best playing EE was a fantastic opponent for advanced players. without the maphack it wouldve been as stupid as other AIs and easily defeated.
the one currency of an AIs quality imo, is how competetively it plays the game in an even contest with a human. since no AI can think and predict as a human does, they more or less require the maphack to play effectively, and I dont even consider the hack cheating since the difficulty for the AI is using that information effectively.
also, worth pointing out, while KAI has maphacks, it cannot see cloaked units
Re: A.I. behavior: Realism or Performance?
Posted: 18 Sep 2009, 11:10
by DJ
You may just have something there IK. What about a level of map hack the can't see units under a radar jammer? That way the AI could be competitve but still allow the covert sides of the mods.
Obviously you could just code the ai to ignore anything under a radar jammer....
Re: A.I. behavior: Realism or Performance?
Posted: 18 Sep 2009, 15:04
by AF
hmmm, the cheating version fo the callback doesnt differentiate between cloaked and noncloaked units, is this something specifically checked for by KAI?
Re: A.I. behavior: Realism or Performance?
Posted: 19 Sep 2009, 12:30
by 1v0ry_k1ng
unique to KAI, it was designed specificly for EE and since the URC faction had large numbers of units that stay permamently cloaked, krog had to make it play properly w/ cloaked units
Re: A.I. behavior: Realism or Performance?
Posted: 19 Sep 2009, 16:52
by JohannesH
DJ wrote:You may just have something there IK. What about a level of map hack the can't see units under a radar jammer? That way the AI could be competitve but still allow the covert sides of the mods.
Obviously you could just code the ai to ignore anything under a radar jammer....
That would lead to a use of radar thatd be totally useless in a real game.
Re: A.I. behavior: Realism or Performance?
Posted: 19 Sep 2009, 20:08
by DJ
That would lead to a use of radar thatd be totally useless in a real game.
?
In a real game radar jammers jam radar - the clue is in the name... I'm suggesting making a map hack AI behave more realistically by knowing when a unit is under cover of a radar jammer and not including those units in its enemy list - until it gains LOS.
Re: A.I. behavior: Realism or Performance?
Posted: 20 Sep 2009, 00:05
by JohannesH
Thatd lead to you jamming your main base etc. cause you dont want the AI to see it. Thatd be totally worthless in a multiplayer game, when the enemy likely doesnt have radar on it.
Unless you expect liche attack but thats not relevant here.
Re: A.I. behavior: Realism or Performance?
Posted: 20 Sep 2009, 01:54
by 1v0ry_k1ng
I think he means it ignores units it dosnt have LoS on that are jammed, but from what I understand about how threatmaps etc work, its probably not a good idea.
Re: A.I. behavior: Realism or Performance?
Posted: 20 Sep 2009, 02:37
by AF
1v0ry_k1ng wrote:I think he means it ignores units it dosnt have LoS on that are jammed, but from what I understand about how threatmaps etc work, its probably not a good idea.
Actually, if the area is jammed, and thus being ignored, then the relevant threat map sections will not be as high, and thus the area will not pose as great a threat as it actually does, thus presenting a much less attractive attack target.