Absolutely no balance at all !?!?

[ScottyPip0]

I am a long time player of civ3 and previous incarnations. I love the game, it has a lot of potential to be a great game. But it lacks in certain ways that I just cannot get past.

Obviously multiplayer would fix this problem, but has anyone else except me noticed how when playing the computer AI, depending on which difficulty you play, you and the AI are never balanced. In every difficulty neither the AI nor the computer player have the exact same potential for playing.

What I mean by this, is that, if you play on the easier setting, your production and growth facters are normal, but the AI has to wait over 2 times as long to build anything. On the hardest difficulty it's the other way, giving the AI a 1/3 of the time to expand and build.

I understand that this is a variance of difficulty, but It's unrealistic. I'd like to play the game, and know that my oppenent has the exact same potential to expand, not kick the computers ass because i've got a clear cut advantage over them, or the computer kick my ass because of the sheer logistics of having to spend 3 times as long to do anything.

I don't want a hard game or an easy game, I want a realistic game. A game where if I beat my opponent it's from smart tactics and decisions rather then a total imbalance.

Maybe in the next version of civ, the makers should consider this in their making? It's silly giving a person a false sence of greatness by allowing them to have advantages of the computer, and it's completely nonsence to give the computer advantages over the player, to make it harder. A game can be hard or easy without this. Perhaps if they develop AI that can make smarter decisions, rather then using an imbalance? Those are my thoughts. No matter what difficulty you play on, it's very apparent that the AI lacks.

Does anyone have any thoughts on this?

 

[chunkymonkey]

The game can be beaten at any difficulty. Sure, the AI will have some starting advantages compared to you on the higher levels, but does not go so far as to make up for the fact that the AI is inept. The AI is terrible at fighting wars, and cannot micromanage citizens to the degree that a human player can.

Why not beat the game at each progressive difficulty, and work your way up until you feel challenged? The challenging games are where you learn most about your own strategies and why you are employing them. Then, have a read of some of the strategy articles here. Go back and play the same difficulty, and hopefully you might find yourself doing better.

I used to be a Warlord/Regent player, but since I've started to think more about my games, and since I joined this forum, I consider myself to be an Emperor/Demi-God player.

The key is just to slightly exceed your comfort level. That's what makes life fun!

EDIT: Oh, and welcome to the gang!

 

[Scuffer]

I'd love to see the difficulty levels scalable on AI tactics, rather than production advantages. It would be great to see the AI mounting a proper naval invasion force for example.
Perhaps something like chess programs, with less time to consider possible moves at easier levels, would work. That said, I know nothing about AI programming, so it may not be feasible.

 

[ScottyPip0]

I've played the game a lot, I have a large understanding of the game, and I haven't found my tactics on forums. I've been playing for a very long time.

Civ3 has a better interface then call to power, but I would consider call to power to have a larger amount of management.

To be honest, one of the many things that piss me off about civ3 is the AI, I mean, it's good enough to give a noob a challenge the first 2 times around, but it just sucks.

But even saying that, remember that half the difficulties the AI has to content with the sheer fact of requiring double the time to complete tasks and grow. Heck, it doesn't matter if the AI was stupid or smart, no one can work with that. The difficulty system is terrible.

Because the AI is terrible, so is the management. Micromanagement is fun at first, but it gets too time consuming and the whole management of the game is set up pretty ****. This is how most games lack. Unless they invent some thought control device where your brain can send off a million messages a second to tell the computer exactly WTH you want it to do, it just takes too long.

You shouldn't have to spend all day making silly decisions. As the leader of a civ, you have very little control of how your society forms. The game misses the randomness of how cities actually work. A societies growth and production aren't mathematical certainties. So I hope if they ever do make another civ game they start to think more about making it in-depth, rather then designing a nice looking engine. During the entire time civ has been around, nothing has really changed in all these years. They've changed the way it looks, reshuffled a few of the elements, added a few simple ideals. But the game itself hasn't evolved past it's boundaries.

That is the thing about any sim style game, once you find out where the boundaries are there's no longer any point in playing it. The discovery phase is over.

 

[kokoras]

ScottyPipo has a point.Indeed the level of dificulty is based on the way AI is expanding and researcing tecks.Not on an Inteligence of another type,the way AI makes war or even declares war has nothing to do with Inteligence.In a hight level the difficulty is that you have to go against a greater quantity of units and not against a more complex strategy.Learning to play is easy,depends on how much time you dedicate on the game.Once you know the ''tricks'' then it becomes more easy,that's why you go to a higher level,because the level has to do with the time that AI needs for production\research\expansion.

By the way,welcome to the forum.

 

[kb2tvl]

I agree with what you say BUT the difference between a good player and a poor player is how fast and efficient they are. A good player can connect cities and improve just the number of tiles needed by that city. Where a new player will place cities for a size 20 and have unworked tiles and large amounts of corruption due to city spacing...

The comparison between a player and the AI is similiar. Production rates. The default AI is tuned to the average player and made easier for the learner and much more difficult for the advanced player. This removes excess computation time and allows the game to get sold to a larger audiance.

The AI does have weaknesses like becoming distracted or confused or obsessed. These tendencies can and do get exploited by players.

 

[nullspace]

Do you realize how impossible it is to write an AI that can beat a smart, experienced player without cheating? It's just not going to happen. The AI programmer would have to know all the good strategies (before the game is even completed), be able to program the AI to make long-term plans and able to implement them down to the last detail, make it able to react to any possible action of the player, and make it unpredictable enough that it can't be exploited. We've had several years of experience and discussion between many people to figure out how to beat the Civ3 AI, the programmers would have to figure out how to beat us before the game is released, and teach the computer how to do it. It's impossible.

The civ3 AI is far from perfect, but it does a good job of expanding, improving its lands, building units when it needs them, trading intelligently, and attacking in big groups. But it's never going to compete with a good player. If you want to play against an AI that doesn't cheat, that's Regent level, if you want a challenge, you have to play higher than that. Beating a high level does take good strategy and tactics. I think emperor is about where the game is challenging, but doesn't feel like the deck is totally stacked against you.

As for micromanagement, I agree with you. Fun for me is making international treaties, invading other civs, choosing my government, or deciding if I should be building banks or universities or cavalry in all my cities. Not fun is moving individual units, improving my empire one tile at a time, telling which person which square to work, haggling for pennies in negotiations, or fiddling with the sliders so I don't overspend on a tech. I wouldn't mind a bit if all that micromanagement was gone in the next game.

 

[magritte]

I'd have to say I'm in between on this. Yes, it's really difficult to get a good A.I. and I doubt you'd ever be able to design an A.I. smart enough that highly skilled, experienced players can't beat it easily if it starts with no special advantages. Computers can beat human beings at games like chess, where the number of choices is limited enough that their shear numbercrunching skills can overwhelm all but the best players.

On the other hand, there are dumb things the A.I. does that would be fairly easy to program out of it. For example, it wouldn't be that hard to program it to wait until a galley/galleon/transport is filled to capacity before moving troops across the ocean. Nor would it be difficult to program it to only declare war on reasonably nearby opponents, or to have it build courthouses only in cities with corruption problems.

 

[Admiral Armada]

As for micromanagement, I agree with you. Fun for me is making international treaties, invading other civs, choosing my government, or deciding if I should be building banks or universities or cavalry in all my cities. Not fun is moving individual units, improving my empire one tile at a time, telling which person which square to work, haggling for pennies in negotiations, or fiddling with the sliders so I don't overspend on a tech. I wouldn't mind a bit if all that micromanagement was gone in the next game.

I agree, the fun lies with invading people and conquering, not microing every litttle detail.

 

[TimBentley]

On regent the AI has no production advantage or disadvantage. IIRC, the only difference between you and the AI is their discount on trading with other AI's.

 

[Tomoyo]

Nope, on Regent, everything is equal.

 

[MSTK]

Tomoyo is correct, IIRC.

And, there is no way to make a smart AI that can beat an expert player that knows tactics not even the creaters comprehend.

The only way to do that is to have true AI, with a learning computer in every game that learns from your tactics and uses them against you.

 

[crazyluke2001]

I think that what he maybe means is that even on regent, where things are theoretically equal, the computer makes such bad decisions that it is still to the player's advantage.

 

[dreamvirus]

You can play on an even playing field on Regent level as Tomoyo points out. As you improve you'll find this gets boring and you don't mind giving the AI bonuses to give yourself more of a challenge. Everyone finds their level. For example, I love playing on Deity, but I never play on sid because I feel th AI bonuses are too large - even if it can be beaten - and I've beaten it once - it's not the same game for me. I don't find it enjoyable to just spend boring ages and ages using tactical tricks to carve up legions of AI units.

As for micromanagement, though, if you removed this feature you would remove a large part of what makes this game so great - the fact that attention to detail can get you almost infinite rewards, expecially in the early part of the game. Remove this and you might as well be playing an entirely different form of war game, like maybe Age of Empires, Command and Conquer...whatever. The ability to consciously control almost every aspect of how your empire develops, to an obsessive degree, is what allows humans to be so good at Civ3 compared to the AI. It wouldn't be the same game without it.

 

[Tomoyo]

If you play on Regent Level with the Governors on and the workers automated, the AI's knowledge of the world and resource locations would offset your military tactics, so that is about as equal as you can get.

 

[ScottyPip0]

Okay, thanks for the responces.

As for the AI, it is terribly flawed not because of programming, but because it's filled with boundaries and it cannot learn. True AI would start off just randomly doing anything at all with absolutely no concept, and would assess the results and slowly build up an idea of the best way to achieve a set goal. Because the programmers have made the AI with set patterns of code already, there is no learning at all.

You wouldn't even have to do that, the programmers could start by implementing true AI into civ3, and let it run constantly learning for a very long time, and then somehow when the time comes move it over to civ4 when it is ready and give it a little bit of time learning on that.

I will make this very strong point of how a very smart AI program was devised. Has anyone played Chessmaster? The chessmaster program has an AI that constantly thinks ahead and devises the best way to win. It starts off with a large set of openings, which guide it to the middle game. During the middle game it designs the best way to win by looking heaps of turns into the future of all possible scenarios. The only limit the AI has is the time limit placed on it's head. If you gave it an infinite time to think it would spend countless days playing out every possible scenario. Chessmaster is only as good as the computer it plays on. At the moment the chessmaster AI has never been beaten.

Is this same formula not possible with civ? Civ is just one very huge chess board, with a very huge playing field of almost infinite squares and infinite possibilities. But, the main objectives are still almost the same. That is why I say the makers should spend less time making the game look good, and more time making the AI proper AI.

 

[Alexnm]

Is this same formula not possible with civ? Civ is just one very huge chess board, with a very huge playing field of almost infinite squares and infinite possibilities. But, the main objectives are still almost the same. That is why I say the makers should spend less time making the game look good, and more time making the AI proper AI.

Civ is much more complex than chess, it's not just "one very huge chess board". In chess, you don't have to decide when and where to irrigate a tile or to mine it. And this is just one example.

 

[dreamvirus]

Scotty, I suggest that you don't know as much about "AI" as you are trying to make out, otherwise you wouldn't be asking these questions. The simple answers:

1) The programmers didn't implement the "true AI" you are talking about - which starts off knowing nothing and learns as it goes along - because that kind of program is at the very highest and most difficult level of programming. You can try to make games seem "smart" but the kind of learning you're talking about is so far beyond the scope of a standard store-bought PC game, it's absurd.

2) Chess programs, as you point out, work by crunching possibilities for many moves ahead until they can work out the best sequences of moves. Chess is played with an extremely limited number of pieces on an extremely small board, and it still takes an enormous amount of computing power to calculate even a few moves ahead. It's only in the last decade or so that computers have started to be able to give human grandmasters serious trouble, and that's not because the chess programs are any better, it's purely because processors are now faster. Try to implement the same method in Civ3 and your computer would simply grind to a halt as the computer tried in vain to calculate every possible move of huge amounts of pieces on an enormous board.

In my opinion your idea that the AI is "fatally flawed" makes no sense. I think they actually did a pretty good job. What you're seeing is not a failure of programming or game design. What you're seeing, and underestimating, is the amazing power of the human brain.

 

[Mewtarthio]

I am a long time player of civ3 and previous incarnations. I love the game, it has a lot of potential to be a great game. But it lacks in certain ways that I just cannot get past.

Obviously multiplayer would fix this problem, but has anyone else except me noticed how when playing the computer AI, depending on which difficulty you play, you and the AI are never balanced. In every difficulty neither the AI nor the computer player have the exact same potential for playing.

What I mean by this, is that, if you play on the easier setting, your production and growth facters are normal, but the AI has to wait over 2 times as long to build anything. On the hardest difficulty it's the other way, giving the AI a 1/3 of the time to expand and build.

Actually, if you play on Regent, you and the AI are completely even.

 

[Uiler]

For example, they have not yet been able to make a computer that can beat humans at go and poker. These are games that are more complex than chess but still less complex than Civ 3. As dreamvirus points out, the only reason why AI in computers can win in chess is because there are only a limited number of possible moves and the computer brute forces it. There is no to little true AI involved. Chess programs are not "smart". Introduce a more complex board (which is what go does) or human psychology of lying, bluffing and deceit and preying on weaknesses like poker does and the computer becomes incapable of winning. Civ 3 does both - the chess board becomes much immeasurably more complex and human psychology comes into play through the human player.

As for your suggested methods of AI Scotty. It's hard enough to program a computer using those methods just to recognise patterns in data (I did a bit of it in uni and I've read about the problems that big defense companies have with face recognition programs) with an acceptable degree of accuracy let alone use those methods to program a computer game as complex as Civ3. When we learnt about the history of neural networks the message was "It was met with great enthusiasm when it was first proposed and received lots of publicitiy, but people became disillusioned with it as a method for AI. It has been relegated to pattern recognition (and can still take a lot of effort for it to work)." When people think of AI they generally think of neural networks and from your message I get the impression that it is what you are talking about. Not sure what the new ideas about AI programming are though. Could be more successful, but that's still in the realm of top-tier computer science done by academics with PhDs and long lists of publications.