Must the computer cheat on NOBLE?!
- Thread starter Endureth
- Start date
BeefontheBone
Windbag of the sea
- Joined
- Nov 7, 2005
- Messages
- 2,018
Yup, though it's worth noting that the player doesn't pay 100% either at Noble difficulty, and I don't know what sort of AI opposition you'll get, but you may see a lot of obsolete units hanging around which should have been upgraded or replaced.
MrCynical
Deity
I would say you'll probably end up back at the Civ 3 level where the AI never upgraded anything (which was awful, and made worse by spearman beats tank issues). If you want to weaken the AI just go down a difficulty level.
Psyringe
Scout
AIs don't have to cheat necessarily, however, few players would be willing to pay some hundred dollars for the development of a truely non-cheating, competitive AI and would still have fun to play the game when the AI needs half a day to calculate its turn. 
However, cheating as in "the AI magically conjures units it couldn't possibly get at the moment" would have been a showstopper. I'm glad that turned out to be wrong.
The AI upgrade bonus is something I'll have to watch though.
However, cheating as in "the AI magically conjures units it couldn't possibly get at the moment" would have been a showstopper. I'm glad that turned out to be wrong.
The AI upgrade bonus is something I'll have to watch though.
King Flevance
Deity
Psyringe said:AIs don't have to cheat necessarily, however, few players would be willing to pay some hundred dollars for the development of a truely non-cheating, competitive AI and would still have fun to play the game when the AI needs half a day to calculate its turn.
However, cheating as in "the AI magically conjures units it couldn't possibly get at the moment" would have been a showstopper. I'm glad that turned out to be wrong.
The AI upgrade bonus is something I'll have to watch though.
I cant wait to see the AIs in another decade or so. We may be able to see some real challenging stuff. I am sure cheating will still be there, but at a smaller degree maybe?
If the AI in CIV didnt cheat it wouldnt be CIV. It used to annoy me wen the tirmire's & planes used to just sail/fly and never need to land near a land square or a city at the end of there turn. Somtimes it paid off for you but theres used to sail with complete safty were's yours would be lucky to survive more than one turn in the ocean.
I still play origonal civ at work its still amazing. When i was a kid i used to think the settler unit looked like a clowns face & now im older i realise there were loads of units like that. I'll look at them now & think how did i come up with that image that bears no relation to what it stood for. The settler clown is the one that sticks in my mind the most though
I still play origonal civ at work its still amazing. When i was a kid i used to think the settler unit looked like a clowns face & now im older i realise there were loads of units like that. I'll look at them now & think how did i come up with that image that bears no relation to what it stood for. The settler clown is the one that sticks in my mind the most though
I'm not into the AI being able to do things that it shouldn't be able to do in any game ... it smacks too much of a cop out from effective and proper AI programming. I mean, when designers of chess games write a better routine for thinking out moves (often emulating different play styles and even former grandmasters), they don't let the computer play cheat and illegal moves just to make for a harder game, do they?
I think all games like Civilization tend to rely on this method way too much as a crutch ... "Hey, let's make the computer harder by allowing them to build units they haven't researched yet, or give them cheaper upgrades!" ... this just doesn't wash with me.
I think all games like Civilization tend to rely on this method way too much as a crutch ... "Hey, let's make the computer harder by allowing them to build units they haven't researched yet, or give them cheaper upgrades!" ... this just doesn't wash with me.
Huxley Hobbes
Anarcho-capitalist
- Joined
- Jan 29, 2005
- Messages
- 162
Kolyana said:
I couldn't have put it better myself, nor could I have provided a better analogy. I don't object to there being AI bonuses on higher difficulties, so long as those bonuses are noted and visible to the players (As they are in, for example, Empire Earth II), but I'd rather hope that when it's meant to be an equal footing that we'd not see cheating like this.
Jedoc
Warlord
I think most people who are outraged that the AI still cheats are either wildly overoptimistic about the current state of computer science, or selling the human brain short by a tremendous long shot. A game like any of the Civs (assymetrical, multiple player, non-zero sum, dynamic, imperfect information, ect.) is degrees of multitude above a game of chess. Creating a minimax algorithm for Civ IV (the most common "AI" for chess-playing computers) that could see more than a turn or two into the future would be near-impossible...and it certainly wouldn't run on your home computer. And even that would be crippling to the AI, since the Civ series requires that you think dozens or even hundreds of turns ahead in formulating even your basic strategies, such as "what should I build next in Rome?"
Consider: the absolute, theoretical maximum number of possible moves in a single turn of chess is 1024. 784, if you take into account the fact that if you have sixteen pieces on the board, that means fifteen spaces which cannot be moved onto by any given piece. Much lower, considering the extremely restrictive movement rules for each piece, and the large number of blocked paths. But that figure fluctuates constantly throughout the game and a true maximum is dreadfully hard to come by, so we'll stick with 784.
In the first turn of Civ 4, you have two pieces, a settler and either a warrior or a scout. A warrior has twelve possible moves. He can move in one of eight directions, disband, fortify, skip his turn, or assume a sentry position. A scout, depending on how many forests and hills are scattered around, has between twelve and one hundred eight possible moves. The settler, depending on the same variables, has between thirteen and one hundred seventeen. A warrior and a settler together? The range is sixty-nine to one thousand four hundred and four possible moves from the AI to choose from. So on our very first turn, in 4000 BC, a true Nashian artificial player has a good chance of doing more work than a purely reactive chess simulator.
And even that's a hugely simplified analysis. Because unlike chess, every square is different from every other, independent of the pieces on the board. Each square may be one of thirty-five effective terrain types, might contain a resource (or might not) and is probably capable of supporting one of several improvements (but might not.) And every move the computer makes would have to be based, to one extent or another, on those variables.
Assuming we could put together this organic, perfectly calculated minimax algorithm, which would take several teraflops of calculating power to run past the Middle Ages, it would still make up only a significant challenge for a sufficiently skilled player. Because the human factor of unpredictibility is even stronger in Civ than it is in chess. The computer would inevitably end up compensating for advantages that the human player illogically neglected to build up, and leaving itself open to crazy, inefficient gambits that a perfect player would never even attempt.
In short: yes. Yes, the AI does have to "cheat" to present a meaningful challenge for a human opponent. Get used to it.
Consider: the absolute, theoretical maximum number of possible moves in a single turn of chess is 1024. 784, if you take into account the fact that if you have sixteen pieces on the board, that means fifteen spaces which cannot be moved onto by any given piece. Much lower, considering the extremely restrictive movement rules for each piece, and the large number of blocked paths. But that figure fluctuates constantly throughout the game and a true maximum is dreadfully hard to come by, so we'll stick with 784.
In the first turn of Civ 4, you have two pieces, a settler and either a warrior or a scout. A warrior has twelve possible moves. He can move in one of eight directions, disband, fortify, skip his turn, or assume a sentry position. A scout, depending on how many forests and hills are scattered around, has between twelve and one hundred eight possible moves. The settler, depending on the same variables, has between thirteen and one hundred seventeen. A warrior and a settler together? The range is sixty-nine to one thousand four hundred and four possible moves from the AI to choose from. So on our very first turn, in 4000 BC, a true Nashian artificial player has a good chance of doing more work than a purely reactive chess simulator.
And even that's a hugely simplified analysis. Because unlike chess, every square is different from every other, independent of the pieces on the board. Each square may be one of thirty-five effective terrain types, might contain a resource (or might not) and is probably capable of supporting one of several improvements (but might not.) And every move the computer makes would have to be based, to one extent or another, on those variables.
Assuming we could put together this organic, perfectly calculated minimax algorithm, which would take several teraflops of calculating power to run past the Middle Ages, it would still make up only a significant challenge for a sufficiently skilled player. Because the human factor of unpredictibility is even stronger in Civ than it is in chess. The computer would inevitably end up compensating for advantages that the human player illogically neglected to build up, and leaving itself open to crazy, inefficient gambits that a perfect player would never even attempt.
In short: yes. Yes, the AI does have to "cheat" to present a meaningful challenge for a human opponent. Get used to it.
Spatzimaus
Mad Scientist
Kolyana said:
There was an old console RPG (I forget which) that had something related to this. As always, the heroes defeated the various Bosses on their way to the Big Bad. Well, right before the end, one of the earlier bosses shows up again. The heroes say something like "Wait, we killed you already!" to which he responds "Ah, but I reloaded my savegame!"
The point is, even if you try to make a perfectly level playing field, by making an AI that knows how to make good trade deals (including going into a deficit just to pay for new techs), improve terrain correctly, manage stacks of units competently, keep units upgraded or scrap obsolete ones, and balance settler sprawl... the human will STILL be at an advantage, since if worst comes to worst he can reload the game. This doesn't even include the "look ahead" factor; the human knows what techs are coming, and plans accordingly. He waits to start a war until he has the key military tech; he plants mines early on the off-chance one of them will randomly spawn the late-game strategic resources he knows he'll need; he decides before he even founds them which cities will specialize in research and which in production.
Most of those AI improvements I mentioned are exponentially more complex than what currently exists. While I'd love to see some of these coded in, I just don't see it happening. As an example, consider mid-game workers. In Civ3, once a city hit size 12 and filled up its food box, you could make a quick 1-turn Worker, the city would drop to 11, then immediately jump back up to 12. But, by the time you could do this, you'd pretty much fully improved all of your core cities, so there wasn't a huge need for it. Even so, in the 10-20 turns before I unlocked Steam Power, I'd have every city churn out a Worker this way as often as possible, then two or three turns before I got the tech I'd assemble all of my workers into groups just large enough to railroad a tile each turn. Now, as much as I'd love to have seen the AI manage something similar, it just never happened, and I doubt it WILL happen until an expansion, if ever.
Using chess as a comparison is pretty bad. Chess' rules are very simple, the units don't change over time, the board is featureless AND doesn't change over time, and the number of possible moves for each unit each turn are rather limited. And even so, we STILL have a hard time writing chess games that can beat expert players. So, imagine how hard it'd be to write an AI that plays equal to the best players, without taking several hours per turn to run through all of the possible strategies. This isn't a twitch game like Quake, where raw reflex speed of bots can make up for weaker AI, and where success is measured solely by whether you're hit or not.
It's even worse than that. Computers can play good chess because they have perfect information. If you have a game without perfect information they struggle, because unlike humans they are unable to properly 'read' their oppositions moves.
To put it in perspective, in poker you generally have only 3 possible moves (fold, bet/raise and check/call), but nobody has designed a pokerbot capable of beating a professional poker player.
To put it in perspective, in poker you generally have only 3 possible moves (fold, bet/raise and check/call), but nobody has designed a pokerbot capable of beating a professional poker player.
Zhahz
PC Gamer
I couldn't find the thread but someone listed the actual setup for Noble and part of it was that AI units cost more up front but they have lower upgrade costs. It seems like a good tradeoff to me as part of the "cost" of upgrading for the AIs is built-in up front. You know they're going to build units regardless of the cost, but you want them to upgrade too or they'll be far too easy.
Generally speaking I don't think this should be a problem, but if you're going for an earlyish rush it could be very annoying to have it shut down by an AIs cheap upgrading (although the chances of them having significant cash to do this type of thing early are pretty low).
Generally speaking I don't think this should be a problem, but if you're going for an earlyish rush it could be very annoying to have it shut down by an AIs cheap upgrading (although the chances of them having significant cash to do this type of thing early are pretty low).
Psyringe
Scout
Regarding AI quality, I'm really curious what the modders can do with the SDK. We *do* have some very good programmers in the Civ community.
lysander said:
Actually if the AI didn't insist on spending every cent he had on trades he might be able to afford those upgrades. What I usually notice in my games is one AI has all the money, then he trades it to another AI for a tech who in turn trades it to another AI for a tech who in turn tries to trade it to the player for a tech.
That's why the computer never has the money for upgrades, because of the way he trades amongst himself. i.e. all India's money for Germany's one tech.
That make any sense? I just woke up.
-E
sadu
Warlord
I have no problem with the AI cheating, so long as there's no surprises.
Civ3 used to really annoy me how the AI would know exactly what units were in your cities, but you could only see the top unit in their stack. Am pleased this has now been fixed.
AI cheating makes up for all the advantages the player gets, such as knowing not to bother going for buddhism and shoot for hinduism instead. Or the suicide galley thing in civ3. Or the save-reload advantage that so many people use when a battle doesnt go their way.
Civ3 used to really annoy me how the AI would know exactly what units were in your cities, but you could only see the top unit in their stack. Am pleased this has now been fixed.
AI cheating makes up for all the advantages the player gets, such as knowing not to bother going for buddhism and shoot for hinduism instead. Or the suicide galley thing in civ3. Or the save-reload advantage that so many people use when a battle doesnt go their way.
Interesting discussion guys
I did not mean to infer in my earlier post that programming a complex AI engine would be as simple as programming a chess program, or even feasible, and I truly acknowledge the immense differences.
In saying that, I think certain crutches for the AI are a little too obvious, or perhaps just a little 'obvious' for my liking ... giving them a 70% discount on upgrades is such an example ... it just seems to lack thought and clever design.
I think a lot of the points Spatzimaus makes *could* be programmed into a game, and while a machine will not - at least this year - think like us or have the range of decision possibilities available to it, I *do* think that it could be programmed to think deeper and more cleverly.
I'm all about giving the HUMAN assistance on easier levels, rmeoving this as you go up, and then giving the AI a different type of help ... perhaps 2 warriors to start with, or two settlers, or the AI has already settled 50 years/100 years before and has a head start in everything ... to me it's more 'plausible' than an outright cheat or bending of the rules.
Just my 2 cents and not intending to flame, it's just that it's always bugged me regarding this type of game going back to CivI and Masters of Magic.
I did not mean to infer in my earlier post that programming a complex AI engine would be as simple as programming a chess program, or even feasible, and I truly acknowledge the immense differences.
In saying that, I think certain crutches for the AI are a little too obvious, or perhaps just a little 'obvious' for my liking ... giving them a 70% discount on upgrades is such an example ... it just seems to lack thought and clever design.
I think a lot of the points Spatzimaus makes *could* be programmed into a game, and while a machine will not - at least this year
- think like us or have the range of decision possibilities available to it, I *do* think that it could be programmed to think deeper and more cleverly.I'm all about giving the HUMAN assistance on easier levels, rmeoving this as you go up, and then giving the AI a different type of help ... perhaps 2 warriors to start with, or two settlers, or the AI has already settled 50 years/100 years before and has a head start in everything ... to me it's more 'plausible' than an outright cheat or bending of the rules.
Just my 2 cents and not intending to flame, it's just that it's always bugged me regarding this type of game going back to CivI and Masters of Magic.
Similar threads
