It's not ********, it's AI
- Thread starter CornPlanter
- Start date
r_rolo1
King of myself
Sorry, haven't seen this when posted
To add, the Civ V big lines are clearly geared towards a gamer AI, and you can see that by the diplo enviroment in general. So it needs a cunning and scheming AI far more than civ IV, that is far more based in inter-player cooperation. Delivering a game like this with a AI that has no more tricks in the sleeve than civ II ( or it was I ? ) had, the "gang-banging on the advanced guy" , is beyond stupid ...
That is why i said the civ V Ai developers needed someone to remind them those basic facts of game theory. The strategy they employed is fine in 1:1, but it goes down the drain as soon as you put 3 or more people on the map ... and civ games rarely have only 2 players
It might be even trying to be smart, but the point is that the current strat of "get furious with advanced people" does give worse results than the strategy "simply stay quiet and mind your own business" . So , as it is not, that strategy is actually making them dumber ...
Maybe, but the fact is that even the civ IV AI, that even in it's better BtS AI version is not a portent of wisdom, does that better than civ V AI ... and that is to say a lot.
To add, the Civ V big lines are clearly geared towards a gamer AI, and you can see that by the diplo enviroment in general. So it needs a cunning and scheming AI far more than civ IV, that is far more based in inter-player cooperation. Delivering a game like this with a AI that has no more tricks in the sleeve than civ II ( or it was I ? ) had, the "gang-banging on the advanced guy" , is beyond stupid ...
That is the main point. They are not sound ... in fact they are stupid beyond reason. The main base of this AI is to assume that it can never count with a cooperative partner and it has to solve the problems of the world by themselfes, even if they lack the ability to do so. In other words, using the good ol'prisioners dilemma jargon , they consistenly choose to betray instead of cooperating , when even in a system that has limited memory, cooperating until backstabbed is the better solution.
That is why i said the civ V Ai developers needed someone to remind them those basic facts of game theory. The strategy they employed is fine in 1:1, but it goes down the drain as soon as you put 3 or more people on the map ... and civ games rarely have only 2 players
Nunya
Warlord
- Joined
- Sep 30, 2010
- Messages
- 270
Oh, lol. I think I just realised why the high-difficulty AIs were not attacking me in the last few games I played. It might be because my empires were small and my armies pathetically weak. If the AI is more likely to become hostile towards you if it perceives you a threat, does that similarly make it more likely to be friendly with you if you aren't a threat?
I think some of you have already seen that screenshot of the China blob of death that never bothered to annihilate me. I had like 2 samurais and a couple of musketmen or something pathetic like that. I was never attacked by any of Greece, Rome or China. None of them even went hostile towards me.
They like it if you have a small number of cities, but a weak military should be bad for foreign relations since it's basically an invitation to get conqured. However, I have read somewhere that the AI uses open border agreements to determine your military strength. So if you weren't allowing open borders and they didn't know how pathetic your military was that could have prevented them from attacking. then again I'm not sure how reliable the source was so it may or may not be true.
Nope I had open borders with all three of the super powers in that game. I did have significantly fewer cities though. About 6 IIRC when each of the other AIs probably had around 15-20.
Nunya
Warlord
- Joined
- Sep 30, 2010
- Messages
- 270
What worse result? There is only one winner, everybody else is a loser. And it doesn't automatically get furious with advanced people. Corn Planter mentioned that Harun previously lost most of his empire to him so that was a bad initial relation before even factoring likelyhood of victory.
Civ IV AI manipulates other AIs? Wow, I never noticed that.
Can you elaborate what you would like to see the AI doing which it isn't doing? I mean appart from better combat AI.
That is the main point. They are not sound ... in fact they are stupid beyond reason. The main base of this AI is to assume that it can never count with a cooperative partner and it has to solve the problems of the world by themselfes, even if they lack the ability to do so. In other words, using the good ol'prisioners dilemma jargon , they consistenly choose to betray instead of cooperating , when even in a system that has limited memory, cooperating until backstabbed is the better solution.
That is why i said the civ V Ai developers needed someone to remind them those basic facts of game theory. The strategy they employed is fine in 1:1, but it goes down the drain as soon as you put 3 or more people on the map ... and civ games rarely have only 2 players
The AI actually is willing to cooperate for quite a time, until factors such as too close borders, limited space for expansion, or your aggresive warmongering make it turn against you. And of course the leaders have different personalities as well. Montezuma doesn't get upset if you're a bloodthirty warmonger, but Gandhi does and Ramses hates it if you build too many wonders.
Now I agree that the AI in general is bad, but that is mostly a problem of the tactical combat AI. The grand strategy AI is reasonably good.
r_rolo1
King of myself
Ok, let's break it out:
-If they gang on the guy and can't stop them all together, they would not stand a chance alone anyway. So in here , the end result is equal for both gangbanging or staying quiet at home.
- Now if the coalition of the willing actually stops the menace, the thing gets murky
The menace is stopped so it can't win for the time being , but most likely the guys that stopped him will be in no conditions of winning either ( if they could stop the menace without major effort, it wasn't such a big menace after all ) ... so who is in better condition to win? The guy that stayed home In other words, gangbanging the guy that is close to win in average favours to not participate in the gangbang ....
I haven't said that either
i said that they were actually closer of being able of manipulating the other AI than civ V AI , even if because the game gave them more tools for that ( espionage, AP, UN, the diplo tools in general ... ) and they actually undersood the value of cooperation.I would like to see the AI:
- first, to determine what is a menace to their win with far more accuracy. IMHO they only realize too late who is menacing the win and because of that their measures are late.
- be able to determine if it has a actual chance of stopping someone to win and act accordingly. This implies a minimally good estimate of who is going to fight or not
- be able to consider if it is not a better idea to side with someone strong instead of fighting it
- if weak , manouver to a balance of power scheme ( chop the tally poppy, like the brits say ). This in general , not only where there is a menace to a win.
- be wary of making research pacts with people clearly more advanced than them. Research pacts as they are now are grave diggers to the least advanced civs and they should act accordingly
- consider the possibility of robbing Ally CS of other people just for balancing proposes.
All of this ( and it is not by far a extensive list ) has heuristics published in literature , is easy enough to program ( surely easier than making a decent tactical AI for a TBS with 1upt ) and would make the game feel a lot more like actual thinking heads below the animated 3D LH
Now you touched another point: the fact that civ V is as good in calculating who is a threat to a win as it is in knowing what friendships are important at the moment
Or what grudges are worth forgetting ... Civ IV AI didn't even tried to guess that , Civ V tries to guess and goes wrong is most of the times , to the point that probably civ IV is even on par with civ V in good calls by the AI in this regard On grand strategy ... the fact is that the AI has a grand strategy to win as much as civ IV had in general : none. Ok they have more strategies coded than civ IV has , even on BtS ( not on BB AI though ), but they don't live by it. ramesses will hate you for building too much wonders, even if they decided to win by domination, while it will probably give a pass with some ease to a Alex that is gobbling CS after CS ... in the same way that a AI that had decided for a culture win BtS would still eagerly make a war that diverted them of their planned win just because they had more units than a guy on the other side of the globe. If they don't care of their own interests , they don't have a grand strategy , period. And going to a war against someone more powerful just because he is too powerful is simply a glowing example of that ... any strategy for a win pases by being alive in the first place
But that also means you can forget diplomatic victory if no one will even vote on you if you have a chance to win with it. Then whats the function, you could just disable it.
And Ive also noticed that if you are small and weak everyone all of sudden want to have pacts of cooperation and constant open borders with you and happily trades anything. Its very rare that they declare war on you.
Drago Askani
Warlord
- Joined
- Jun 19, 2007
- Messages
- 293
I say write diplomacy on the side of a Nuclear Missile and drop it in his capital... I bet he will be little nicer to you while you negotiate his surrender
lawl so true.
From what PoM is saying, it sounds like this may be more of a parameter tweak issue than a real fundamental AI flaw. It makes sense that the backwards, nonthreatening Civs (player included) wouldn't be an immediate target when there's bigger, much more threatening opponents out there for the AIs to due it out with. So that half of the equation makes sense. (It also makes 3-city/OCC culture wins possible I suppose, and makes it possible for some AIs to be programmed to do less expansion without being entirely screwed.) That's all good.
However, that implicitly means that the larger/more powerful an AI or player is, the more likely others are to attack it. That makes sense... up to a point. But when that doesn't get adjusted back down against empires that are way freakin' bigger, it's a problem. It doesn't really make much sense, and it's really bad for balance. But, like I said, that may be more of a "we'll just adjust this number" or "just have to rewrite this string of code" than an "aw crap guys the AI's totally screwed."
However, that implicitly means that the larger/more powerful an AI or player is, the more likely others are to attack it. That makes sense... up to a point. But when that doesn't get adjusted back down against empires that are way freakin' bigger, it's a problem. It doesn't really make much sense, and it's really bad for balance. But, like I said, that may be more of a "we'll just adjust this number" or "just have to rewrite this string of code" than an "aw crap guys the AI's totally screwed."
Infiltrator
Warlord
- Joined
- Sep 13, 2009
- Messages
- 295
The AI only acts on these rules:
1. When fighting against human opponents, always place your units in front of his first, even if they are ranged or have enough MP to attack.
2. If the human does not respond to your units, attack anything closest, regardless of STR ratio, fortification or terrain.
3. Always trade for Luxury resources, even if you have 20+ happiness.
4. Always trade for Strategic resources, even if you don't plan on using them.
5. If you lose 2-3 cities, offer 4-5 more, all your money and Luxury resources. This will surprise the human and he will accept, and your plan to survive and win the game will be set in motion. I think.
1. When fighting against human opponents, always place your units in front of his first, even if they are ranged or have enough MP to attack.
2. If the human does not respond to your units, attack anything closest, regardless of STR ratio, fortification or terrain.
3. Always trade for Luxury resources, even if you have 20+ happiness.
4. Always trade for Strategic resources, even if you don't plan on using them.
5. If you lose 2-3 cities, offer 4-5 more, all your money and Luxury resources. This will surprise the human and he will accept, and your plan to survive and win the game will be set in motion. I think.
Similar threads
