Dumb AI

[Solver]

Originally posted by gdgrimm
They're capable of doing it with Chess games. Admittedly, that's a far simpler and much more studied game than Civ, but I would think Civ could come at least a little closer than it currently does.

Not even anywhere near close. The current computational powers are about enough to iterate through all possible chess moves. There are relatively few, and thus relatively little "thinking" complexity is required.

Consider this, for your first move in chess, you can make 20 different moves, most of which are stupid anyway. For Civ, if you have a Settler and a Warrior, you can start the game in 513 ways, without considering that you can move the Settler 1 tile and found a city then and without considering any options on what to do with the city, what to research, etc.

Even in mid-game, chess doesn't quite reach the amount of possible piece moves compared to how many ways you can move your units in Civ4. And again, this is only considering pieces/units, without even mentioning the other Civ factors.

[padillah]

Originally posted by gdgrimm


They're capable of doing it with Chess games. Admittedly, that's a far simpler and much more studied game than Civ, but I would think Civ could come at least a little closer than it currently does.

Can you imagine playing a Chess game where, at the highest level, the computer started with an extra row of pawns?!?

This has been debated (and soundly thrashed, by none other than myself and Solver) in several other posts.

The upshot of which is, chess has a limited number of tokens on a limited playing area, Civ outpaces the playing area, number of pieces, complexity of movement, and complexity of capture (combat) by several orders of mangnitude. Throw in a changeable playing area (improvements), production as well as destruction of pieces, culture, aliances, tech trading... and you get a game whos mechanics, if not the game theory defined by them, is not even remotely coincidental to chess.

For the rest of the argument search for some of the other threads.

Tom P.

[Stickyman]

Do you play chess or just know how to move the pieces?

Chess is a far more complex game then civ.

How many mainframes have they taught to play civ in order to defeat a human opponent?

[Arrian]

Chess is more complex than Civ?

It may be a more elegant game, yes. But there are sooooo many more things to be considered in Civ.

As for the mainframe argument... I don't follow you. Are you asserting that the fact that it took a supercomputer to beat a human makes Chess more complex than Civ? The computer was playing the (or one of the) very best Chess players in the world. A comparably skilled Civ player will make mincemeat of an AI in a duel game (balanced starts). Mincemeat. Could a supercomputer do better than the CivIV AI? Certainly. That's kinda the point, though: in order to have the AI act smarter, it needs a ****load of processing power/memory. The average home computer doesn't have nearly enough.

-Arrian

[Solver]

Chess is no way more complex than Civ. Complexity is a mathematical concept. And anyone familiar with very basic combinatorics can conclude that Civ is more complex, even only taking in account unit movements and actions, without any regard for city management, economics, research, diplomacy and other stuff.

What it takes to beat a human at chess varies. Computing power grows real, real fast. 10 years ago beating the top players was a challenge. Now, the best chess computers can absolutely trash 99.99% of chess players. There are maybe 5 players who are still able to play on par with computers. And it's safe to say that in 2010 even they won't stand a chance against computers.

Thus, actually, maybe in 2030 or so standard home computers will have the power of current top supercomputers, and everyone's home machine will easily beat any player .

[MightyTiny]

Complexity, and difficulty for AIs are different concepts from what is complex and difficult for humans; for a human, chess can seem very complex, simply because it requires deep calculation, something that humans aren't that well equipped for.

On the other hand, Civ4 seems much simpler to humans because it more or less directly simulates "real life"; an environment where our intuitions work. Moreover, Civ4 presents a "wide" problem instead of a "deep" one like chess; raw calculating power counts for much less, because you do not have perfect knowledge of your opponent's position, nor do you have perfect knowledege of what their options are. That is why it is impossible to calculate very deep, but instead, success is achieved through anticipation of realistic possibilities, and the ability to at each turn prune the list of sensible actions from the ENORMOUS list of possible actions.

AI's are much, much easier to make for fully defined problems with a relatively narrow range of options at every stage such as chess. For problems that are not fully defined, where the consequences of actions can only be anticipated, never fully known, and where there are innumerable possible actions to take at any moment, making an AI to handle this is a task not at all comparable to making an AI for a game like chess. It's orders of magnitude more difficult.

Looking into artificial intelligence a bit can really increase your appreciation of what a marvelous piece of machinery we carry around in our skulls - things that we achieve without real effort, the "wide and messy" problems common to real life situations, are inordinately difficult feats to duplicate with an AI.

[Rook]

The difference between chess and Civ is the number of variables. Chess is rigid. Pawns always move like pawns, the board is always the same, the queen is the most powerful of the pieces. It's actually quite simple to get a computer to "understand" those concepts and develop a strategy. In Civ, there is just so many more levels of complexity involved.

With that said. I didn't mean for this thread to be made into an argument about hard-core gamers vs standard gamers. I would certainly consider myself closer to the latter. I just think that blatent AI goofs should be addressed. The AI, shouldn't build 800 cannons when he needs to defend his city. More thought should go into where an assault should land its units and not stack them in one big stack. these are just a couple of things that even a "standard" player could counter too easily.

[xxFlukexx]

Originally posted by padillah


You are confusing "Hard" with "difficult", they are not the same. Yes, Ninja Gaiden is hard, but is it really that difficult? As difficult as say...Civ? With hundreds of units and resources and leaders and culture and buildings and improvements... It's a difficult game.

I'm playing "Need For Speed: Most Wanted" right now and I can tell you: it's hard. You have to drive almost perfect to beat the last few racers and if the cops get to level 4 (or, heaven help me, level 5) you are pretty much guarenteed to get busted. It's hard. Difficult, no. Not in the least. It's driving a car... actually, it's not even that. There's no damage to your car, you never break it, you can jump off a cliff and it's fine. You can't leave the track and get lost, you can't loose (in as much as you just keep racing until you win or quit).

So, hard - yes, difficult - NO.

Tom P.

Yeah, you are correct in that, I am getting hard and difficulty mixed up.

So you guys are talkingh about the complexity and learning curve of a game, ok. How complex should a game be? Well I think Civ4 certainly isn't too complex for most people. I think there is a market for a more complex game. Not a blockbuster huge market, but a good hardcore audience. For example I think SMAC's unit workshop is complex but not overly complex to take out the casual gamer market.

The games that do it best kind of hide their complexity so casual gamers can play fine but those who wish to tinker and get really in depth can. An example is Gran Turismo, I play the game with the cars it gives me, but you can customize your car with parts and ever further get in deep and change the specifications of the cars and I have no clue to what any of that stuff does, but can still have fun.

Hehe and if I'm off-topic again oh well I give up, ignore me.

[Solver]

So you guys are talkingh about the complexity and learning curve of a game, ok. How complex should a game be? Well I think Civ4 certainly isn't too complex for most people. I think there is a market for a more complex game. Not a blockbuster huge market, but a good hardcore audience. For example I think SMAC's unit workshop is complex but not overly complex to take out the casual gamer market.

Civ4 strikes an excellent balance of complexity and fun, IMO. It doesn't require you to do things like micromanage production not to waste one shield, but it's a strategically complex game - there are many good strategies, and no one right way to go. In this, Civ4 is probably the best Civ game to date.

SMAC complexity was great. I love the game for the sheer depth of concepts, yes, truly great. However, SMAC is only half as fun as it could be because it features an AI that absolutely can not handle that complexity and has no idea what half the features are.

There's definitely market for more complex games - that's what Paradox Games produce. Hearts of Irons or Europa Universalis don't top any lists of best selling games, but do have their own market. And that's a seriously complex game, one that is actually somewhat too complex for my liking.

[potoroo]

Originally posted by CerberusIV
That an AI that is a serious military challenge is hard to program I can understand. However my current game is in the final stages of a spaceship race and one of the AI players has a windmill on his only aluminium - so he's not getting any benefit in building spaceship components. That seems a basic flaw.

Yes, the AI sometimes seems to be unable to sensibly prioritise. I've also seen other resources, like oil, go undeveloped by the AI if there's an existing improvement on the tile. You'd think that where a resource was a prerequisite for other things, like aluminium or oil, the AI would develop it even at the cost of destroying a previous improvement.

And as for the AI's fixation with windmills - maybe Firaxis should stop hiring Dutch programmers.

[Urban Ranger]

Originally posted by Solver
Consider this, for your first move in chess, you can make 20 different moves, most of which are stupid anyway. For Civ, if you have a Settler and a Warrior, you can start the game in 513 ways, without considering that you can move the Settler 1 tile and found a city then and without considering any options on what to do with the city, what to research, etc.

To be fair, that's at least two "moves" in the first turn, maybe three.

One thing is, while the total possible moves in a chess game decreases as the game progresses, the same can't be said of a Civ game.

[Urban Ranger]

Originally posted by padillah
You are confusing "Hard" with "difficult", they are not the same. Yes, Ninja Gaiden is hard, but is it really that difficult? As difficult as say...Civ? With hundreds of units and resources and leaders and culture and buildings and improvements... It's a difficult game.

I think you are trying to draw a distinction where there is none. I don't think you can use "difficult" and "complex" synonymously, while OTOH, "hard" and "difficult" are usually interchangable when used to refer to things that aren't easy to do/perform/accomplish/etc.

[padillah]

Originally posted by Urban Ranger


I think you are trying to draw a distinction where there is none. I don't think you can use "difficult" and "complex" synonymously, while OTOH, "hard" and "difficult" are usually interchangable when used to refer to things that aren't easy to do/perform/accomplish/etc.

Right, "complex". (I knew there was a better word)

Syntactically you could draw a difference between "hard" and "difficult" but you are right that the difference between "hard" and "complex" is more noticable and requires less semantic "reaching".

Thanks.

Tom P.

[padillah]

Originally posted by Stickyman
Do you play chess or just know how to move the pieces?

Chess is a far more complex game then civ.

How many mainframes have they taught to play civ in order to defeat a human opponent?

Yes, I know how to play chess, not just move the pieces. ( I have a penchant for the King's Pawn/Ruy Lopez openings)

I also know that a chess board has 64 x 64 = 4,096 squares and 16 + 16 = 32 tokens. If every token could move to every square at any time (clearly not possible in chess but I am looking for an extreme example) that would be 64 x 64 x 16 = 65,536 possible moves for each side.

On a standard CIV map (84 x 52 = 4,368) and 8 teams there are 2 x 8 = 16 tokens on the first move, some of which can move twice. That's 69,888 moves for the first move posibilities. Now, consider that ten moves in every team builds a new unit the number of moves to be considered now grows to 4,368 * (3*8) = 104,832. Now consider that ten moves after that each team created a new unit again, that's 139,776...

That could go on for quite some time. An average game is ~400 turns long so it could go on, at one unit ever ten turns, until all the teams have 320 units at 1,397,760 possible moves. On a standard map. with only one city. Not including production bonuses that make unit production faster.

This doesn't address the esoteric decisions of "What tech to research?" A not straightforward topic at all (as evidenced by the number of different "bee-line" strategies around). This, above all else really puts Civ in a different place than chess. Chess has no esoteric decisions. Every move can be calculated to be good bad or indifferent. Nothing in chess comes close to the ethereal nature of "Did I use the tech tree well?"

As for the mainframe argument - I'm with Arrian, I don't follow the logic. First off we are arguing the fundimentals of the game, not the AI (which is a false term as applied here anyway). In fact, as Arrian points out, this goes to the heart of our argument, you would need a supercomputer, several if you ask me, to make an AI that could beat a human at Civ. The processing power of a simple PC is not enough to support an AI that must take into account the numerous moves and intricacies of Civ. On the other hand, I have a chess program on my Palm T3 that consistently beats the crap out of me.

That being said, chess has been around for hundreds (dare I say, thousands) of years and still is one of the best known and most challenging games in history. Civ has been a really cool game for the last 25+ years... it's got some catching up to do.

Tom P.

[Stickyman]

I'm afraid that the point I was trying to make is lost in communication - my fault.

Would you say it is easier to become a "good" player in Civ or in Chess?

I personally believe that it is much easier to become "good" or "proficient" in Civ then it would be in chess. As result of that belief, I think it is fair to say that chess is more complex than civ for humans.