ZOMG - chess?!

I am guilty of being one of the people who has mentioned chess in comparison to Civ. I am not saying Civ AI and chess AI can use the exact same techniques, but they are both 1UPT games and I believe most competent a chess AI programmer wouldn't program his units to perform suicide moves (unless it gave him a positioning advantage of course). Some parts of Civ's AI has little resemblance to chess, but the part when the AI moves his artillery to the grassland in front of your melee units certainly does, that's basically the same as moving your queen in front of his row of pawns.
 
I don't think you quite understood his post. an "approach using different levels of AI to consider moves" would be an example of "high level abstractions and shortcuts". The number of possible moves on any given turn of civ is many many times more than for chess.

But is it? I know that with Civ you can play on some pretty enormous maps and can obviously have quite a few civilizations to play against in a game, but the AI like us doesn't have to consider the whole map, at least in the beginning. Civ V's approach is suppose to be one of:

"The artificial intelligence (AI) in Civilization V is designed to operate on four levels: the tactical AI controls individual units; the operational AI oversees the entire war front; the strategic AI manages the entire empire; and the grand strategic AI sets long-term goals and determines how to win the game. The four levels of AI complement each other to allow for complex and flexible AI behaviors."

Surely if they are able to do that with the AI you could assume it could be broken down even further by breaking down game play into regions of the map. After all the AI is considering things by what we see, but as one big spreadsheet for which computers accelerate at.

This is an offhand reference to GalCiv2, but I think that game was actually made backwards. What I mean by this is that Wardell is careful not to put elements in his games that are difficult for AI's to handle, something that he owns up to more than many of his fans.

A cursory glance at the building list and tech tree for GalCiv2 is enough to show anyone that it is a much flatter, more computable type of game. Not on the same order as Chess certainly, but this kind of brute force AI really is hopeless for Civ.

Making an AI that could play Civ4 well enough to pose a Deity level challenge at Noble handicaps would be a larger project than developing the game itself.

Fair enough, but the AI has to consider things like customized ships which allows for a pretty large amount of different units. I think as you are saying some of the best AIs are ones in which the designers only allow x amount of units and not the wide array that Civ has.
 
Chess sucks due to 1UPT! Bring back SoDs!

Also, the horse rush in chess is absolutely awful!
 
It seems like every second thread I read on here ends up with somebody mentioning chess ... and I'm like WTF? What possible, conceivable similarity is there between an old and boring glorified game of checkers and a PC game loosely themed on human history?

Am I missing something here?

Oh and I really hate chess, but I absolutely love Civ ... :p

I am one of the many that compare chess to Civ. Mainly for military reasons.

1.) Civ like Chess is a board game, but a much larger board.

2.) Civ has to calculate not only different tactics, but combine them into a strategy. easier to do in Chess because of the very small 8 x 8 tiled board, 6 unit types, and only one other player.

3.) Civ has many unit types like chess with vastly different abilities. Civ has many more unit types and most types can evolve to yet another type.

4.) The different unit types can interact with each other to make them more powerful.

so programming wise, Civ is similar to chess for resolving combat, planning moves, figuring tactics & strategies.

So Civ combat is a lot like Chess, just a more advanced version. It is so complicated, I don't think you could feasibly make civ into a board game. Besides, you'd need a whole room just to lay out the map, and that would be the small map.

-=Mark=-
p.s..Chess roxors checkers by a huge margin. Checkers only have one main unit, with the possibility to gain a second one (king) while chess has 6 different unit types with an ability to a possible upgrade(queen, which is already a unit type). Chess can use all 64 tiles, while checkers can only use half the tiles.
 
Chess sucks due to 1UPT! Bring back SoDs!

Also, the horse rush in chess is absolutely awful!

I opened with 1. Nf3 2. Nc3 and the computer just suicided pawns in the open ground... SURELY this should have been tested before release?!?
 
Interesting that no one mentions the fact that all units can move in the same turn in Civ, possibly the thing that complicates it more than any other in comparison to chess when combined with 1UPT.
 
The reason chess AIs are so good is that they are brute force, a luxury that no Civ game has.

By brute force, I mean that the AI evaluates every possible move that it can make, and every possible counter, and look ahead 2, 3, 4, or even more turns, and then weights each possible move and picks the best, based on its knowledge of future outcomes.

This is possible in chess for several reasons, such as, only 6 unit types, very limited board size, no fog of war, etc.

Civ could never have this type of AI simply due to its complexity and the size of the play area. So Civ must work with some high level abstractions and employ a number of shortcuts, or the AI would take hours simply to evaluate a single move.

And that brings up another favorite topic of mine, that of game complexity. Which is more complex, Civ, or chess? Or Go, for that matter?

I am firmly convinced, that for all of their simplicity, that chess and Go are miles ahead of any Civ game in the complexity department. Go to your public library and see how many books there are on chess strategy and analysis. The only people who think that these games are simple are the ones who haven't taken the time to understand them.

I think that one of the problems with modern computer games, and I use Civ games as a prime example, they have confused having a plethora of options with crating a more complex game. Sometimes this works, but most often, not.

For myself, I don't see any justification to rating Civ 3 or 4 as any more complex that Civ 1 or 2. They have many more options, more choices for the player to make, but in my mind, they are no more complex, and in many ways, not even as good as their predecessors.

Sorry if this seems flame worthy, it is not intended in any was as flame bait. These are just my opinions, and now and them I get the idea to express them. You're free to ignore them if you dislike them. :)

This is because you are tempted to brute force chess because you can, while you (like the AI) are forced to use rules based approximations and short cuts because you cannot brute force Civ. I wonder how many possible moves (including variable combat results) there are on turn 100 of the average Civ game. :)
 
If you don't know the difference between chess and checkers, I imagine you think Civ V is a fantastic game.



Hmm ... :confused: Well, in checkers, it's a bit like all your pieces are a hybrid of bishops and knights, because they can only go diagonally, but they hop over other pieces on the way. :D
 
chess is a turn based strategy game. with multiplayer options

civ is a turn based strategy game. with multiplayer options.

there you have it.


as of AI which was mentioned above, that guy is totaly right.

you can't compute all moves the way computers compute now.

unless you instal a self-learning AI, which are currently still prototypes and probably very expensive (and quite hard to make for game designers i'd guess). you will need to use workarounds for a lot of situations.

ALSO, if you want the AI to play well, you should be able to play well yourself (or at least know what exactly is playing well, and wheat is not playing well). usually when game designers actually play their game online, they'll get beaten sooner or later by the hardcore gamers.

the AI they created will get beaten even harder.
 
Interesting that no one mentions the fact that all units can move in the same turn in Civ, possibly the thing that complicates it more than any other in comparison to chess when combined with 1UPT.
I was considering mentioning it earlier, but I decided to make a snarky joke instead. A better comparison would be a game like Arimaa or something which allows moves of multiple pieces per turn.

And of course, this ignores that the battle AI is a minor piece compared to the huge, over-arching strategic AI which makes even more important decisions which are even more difficult to evaluate.
 
I don't think chess AI is a true brute force AI. Games like chessmaster have a dictionary of opening moves that matches a high level players knowledge of chess games; in addition they use lots of heuristics to prune paths of inquiry so that the computer can move in 10-20 seconds instead of 10 minutes. This must be the case because chess AIs are different--even at the highest level, the MS chess AI, for example, miss values certain positions that the chessmaster AI values correctly. Even deep blue (which no player here is likely to have faced) uses pruning techniques.

That said, it's certainly true that there's more of a brute force element in chess than CiV.
 
Fischer random chess (chess960) is a lot more complicated than civ5, is random and by its nature cannot rely on databases. Yet it has AIs that are significantly better than a civ5 AI and they don't use bonuses. http://lichess.org is one you can play online for free (warning: you may come to the sad realisation that this is significantly more challenging and fun than ciV). So I think the fact that people can create such AIs does make at least a comparison between a chess960 AI and a civ AI credible. Frankly I don't hate ciV I think it is an average strategy gane but it is perfectly reasonable to criticise its AI when the "premier" tbs game has an AI worse than "small" projects that faced similar challenges. I mean how can a fischer960 game online have a better AI at its lowest level then ciV? If ciV had great combat AI and poor economic, diplomatic AI I could accept they had at least tried. They should of been able to at least get combat right imo.
 
Fischer random chess (chess960) is a lot more complicated than civ5, is random and by its nature cannot rely on databases. Yet it has AIs that are significantly better than a civ5 AI and they don't use bonuses. http://lichess.org is one you can play online for free (warning: you may come to the sad realisation that this is significantly more challenging and fun than ciV). So I think the fact that people can create such AIs does make at least a comparison between a chess960 AI and a civ AI credible. Frankly I don't hate ciV I think it is an average strategy gane but it is perfectly reasonable to criticise its AI when the "premier" tbs game has an AI worse than "small" projects that faced similar challenges. I mean how can a fischer960 game online have a better AI at its lowest level then ciV? If ciV had great combat AI and poor economic, diplomatic AI I could accept they had at least tried. They should of been able to at least get combat right imo.

So the computer can't reference an opening book. Not a big deal... Positions in chess can still be evaluated with pretty good accuracy by an evaluation function which is simple enough to allow the computer to accurately assess millions of positions per second. Minimax search with alpha-beta pruning, and you can brute force anything, even without an opening book.

The big limiting factors in creating a tactical AI for Civilization are:
1. The difficulty in creating an evaluation function to judge how 'good' a position is.
2. The difficulty in figuring out a 'scope' to judge. (Do I only consider these 2 dozen hexes to evaluate my moves? Do I consider the entire continent?)
3. Some randomness in combat. (Not as much as Civ4, but it is there.)
4. Multiple moves per turn per player which can cause ridiculous increases in branching factor.
5. Speed. If Civ5 takes more than 15 seconds to proccess every single move, tactical or strategic, for 7+ AIs, people start complaining.

If you can figure out a way to solve all those problems (ESPECIALLY the speed issue) I'm sure you could make a fortune.
 
So the computer can't reference an opening book. Not a big deal... Positions in chess can still be evaluated with pretty good accuracy by an evaluation function which is simple enough to allow the computer to accurately assess millions of positions per second. Minimax search with alpha-beta pruning, and you can brute force anything, even without an opening book.

The big limiting factors in creating a tactical AI for Civilization are:
1. The difficulty in creating an evaluation function to judge how 'good' a position is.
2. The difficulty in figuring out a 'scope' to judge. (Do I only consider these 2 dozen hexes to evaluate my moves? Do I consider the entire continent?)
3. Some randomness in combat. (Not as much as Civ4, but it is there.)
4. Multiple moves per turn per player which can cause ridiculous increases in branching factor.
5. Speed. If Civ5 takes more than 15 seconds to proccess every single move, tactical or strategic, for 7+ AIs, people start complaining.

If you can figure out a way to solve all those problems (ESPECIALLY the speed issue) I'm sure you could make a fortune.

These reasons are valid ones on why ciV cannot have a great AI they aren't really good reasons on why ciV has a terrible combat AI. Currently ranged units run in before melee units, none of those reasons are so limiting as to cause that sort of stupid AI behavior. The current AI is poor even taking into account mitigating factors such as you mentioned.

But I do agree you are correct, the very design of this game is faulty to the point that a good AI is probably beyond any reasonable expectation. That is essentially what you said above, that is why I think this game has no hope. The very design prevents good tactics on the part of AI. 1upt might be a good rule but technology needs to be able to support the rule before it is enabled otherwise we end up with this situation.
 
the very design of this game is faulty to the point that a good AI is probably beyond any reasonable expectation. That is essentially what you said above, that is why I think this game has no hope. The very design prevents good tactics on the part of AI.
There was some discussion in another thread to use Monte Carlo style methods to do the AI, which eliminates some of the problems with standard AI methods. I'm not sure the processing power is good enough to throw it into Civ yet, but if not, I bet it's not too far off. (This is actually the method used by the top Go playing computers, since that's another game which has issues with evaluation functions and branching factor.)
 
Chess was so much better before 1UPT.
 
Top Bottom