AI and "Raging Barbarians": impressions...

DarkLizener

Chieftain
Joined
Apr 24, 2006
Messages
24
Hi
I just wanted to ask around how the AI performs with raging barbs on. In my recent games I enabled this option to have a little more early struggling and fighting, but all the games went down the drain. The reason: the AI-players got eliminated one after another (didn't even last long enough to see orthus joining)

I won two games in a row by just surviving long enough. It seems the AI does not respond to the enabled option at all and just guards their citys with one or 2 warriors which are overrun sooner or later. This games were on a pangea map.
 
What difficulty are you on?

On prince difficulty and no building requirements the AI usually fights back against the barbarians very well in my games...
 
If you want an harder game you shouldn't turn on Raging Barbarians, becuase lot of AI are just destroyed by barbarians, while even an average player can get a great early military advantage through XP farming.
 
While one or two civs sometimes get anniahlated by barbs early on in my games, all of them usually make it to the midgame. But then, I play on pretty dense maps (room for maybe four or five cities for each civ), so I probably don't run into as many barbs as most people would.
 
I usually play on prince
 
Surviving of the fittest. I play on Emperor with AI No Building Requirements and Raging Barbs, usually on a standard sized map. Sometime one AI becomes eliminated early but that's an rare event. More common is kind of a repression for some AI while others flourish. This leads to quite some gaps between them early in the game, which makes the whole thing more interesting for me as a human player.
 
Immortal/raging barbs/barb world: Suprisingly it is very rare for them to do die. The only time they died in mass was when I played Clan of Embers with 5 civs on huge. 3 of them were crushed by middle game.
 
Try Creation Mapscript by Cephalo if you dont do allready.

Helps AI at the start, offers lots of free wilderness areas until late midgame and to a smaller degree reduces AI-expension frenzy a bit. Also the Barbs dont just go after one Player then it seems to. :) (So most of the things that are critisized widely are partly fixed by using that script. And its very flavorful as well since most civs start at their prefered Terrain. Can be found in the Scenarios & Mods subforum.)

Since i started to play with that script i always play with raging barbs + barbarian world (and wildlands most of the time which gives some respite to you and the Ai before the hordes start to come. Also some Animals to catch if you like and still a lot of opposition for explorers) so much more fun when those settings really work out.

Whould also advise you try Ai-No-building requirements for without it AI sucks vs. barbs as vs. players. (If afterwards AI is to strong for you, rather reduce the difficulty a notch than put it as high that the AI can compete with the requirements. For then it can become really nasty if you are used to a bit lower difficulties.)

So raging Barbs and no Ai-building requirements seems to be more or less tied to work in the game.
 
at the higher difficulty levels where the AI gets obscene bonuses against barbarians, I've *never* seen an AI civ eliminated on raging barbs since the BTS upgrade. Personally I don't like the AI's anti-barb bonus, it's just a cheat in place of actually keeping enough units in their cities.
 
Personally I find that playing on any difficulty which isnt noble is rather pointless. The AI should have the same build times and research times as a player. If anything, the "effectiveness" of the AI should be improved so that they can compete with players without having to handicap human opponents.

If this was done, I'm sure the AI could deal with raging barbarians in the same way as a veteran player.

Al
 
Improving the AI is not easy. It requires a lot of testing and coding, and with all the new features and feature changes from one version to the next, it's just not possible in FFH until the mod is "complete." The bonuses are (hopefully) a temporary measure.
 
Personally I find that playing on any difficulty which isnt noble is rather pointless. The AI should have the same build times and research times as a player. If anything, the "effectiveness" of the AI should be improved so that they can compete with players without having to handicap human opponents.

If this was done, I'm sure the AI could deal with raging barbarians in the same way as a veteran player.

Al

well, that would take a veteran player who manages to program every single evaluation, action and strategy into an AI code. Not only would that take months/years of time, it would also make load times atleast 10x as long as they are now. Remember, the human brain is still way more advanced than the average pc ;).
 
I didn't mean create a human-like conciousness in a computer. I mean't just upgrade the AI. At the moment the AI makes some very obvious mistakes, and there is certainly room for improvement.

Al
 
Remember, the human brain is still way more advanced than the average pc ;).

We aren't that far away from the human brain power, which approximately achieves 20 Teraflops. Intel's experimental Polaris multicore chip has a performance of up to 2 Teraflops.

So let's see how long it takes until the AI is unbeatable (at least for humans) on settler difficulty. ;)
 
Personally I find that playing on any difficulty which isnt noble is rather pointless. The AI should have the same build times and research times as a player. If anything, the "effectiveness" of the AI should be improved so that they can compete with players without having to handicap human opponents.

If this was done, I'm sure the AI could deal with raging barbarians in the same way as a veteran player.

Al

... says the guy who has obviously never programmed a strategic AI. It is *really* hard to make a fun AI to play against. The fact the CIV 4 AI can AT ALL work with the FFH mod is a testament to the tweaks the modders have made to it, as well as the very solid original AI. Tweaking little bonus' into them allows the player to test their skills more, without years of complicated inflexible AI research.
 
We aren't that far away from the human brain power, which approximately achieves 20 Teraflops. Intel's experimental Polaris multicore chip has a performance of up to 2 Teraflops.

So let's see how long it takes until the AI is unbeatable (at least for humans) on settler difficulty. ;)

it's already possible for computers to become "unbeatable", because of the level of micromanagement they've got, and the extent of turns they can think ahead (if given the oppertunity). Just look at a chess computer ;).

The problem is the variety and multitude of tasks a computer game like Civ has to offer. Humans can learn, AI (at the moment) hardly can. Humans usually get a bit better with (almost) every game they play, the AI stays exactly the same. That's why there are difficulty levels, to try and mimic a "learning" AI which stays challenging.
 
So I take it you have programmed strategic AI Zechnophobe?

I can actually program now you ask, but not using Python or C/C++.

If you cannot program stragetic AI then I don't think you should be making such comments as you did about my suggestions.

Al
 
Main problem that tends to be encountered with AI was already mentioned, though possibly on accident. It needs to be good, but it also needs to be bad. You have to balance the abilities to achieve something which is an enjoyable opponent. One could code the AI such that it always plays an expert game, and is impossible to beat fairly easily. But to make it play well, yet make mistakes, but ones which are not crippling... It's a pain.
 
The palace should have 10-25% of defense rating + health recovery (like herbalist) to help the AI.

Or an extreme solution, create an ai imobile unit at the capital, with 100%-200% bonus against barbarians.

EDIT: Or a conditional promotion to any unit in a tile with (palace, winter palace or forbidden palace) which would grant the bonus vs. barbarians.
 
Back
Top Bottom