AI doesn't try hard enough to win

[Gurgeh]

I've always found that a frustrating aspect of the AI's make up - about 4 months ago I had a similar kind of game where I was going to win by space race - and did so - even though it would have taken Pericles no effort to wipe me out - he had about 15x my power and 45% on domination stats (+5 Vassals) and it kinda seemed a shame that the AI didn't bother to declare war on me to stop me - instead with 4 turns till my ship arrived he DOW'ed Mansa Musa - who was neither big enough to matter and a far harder initial fight.

HOWEVER.....

I would be a bit concerned if the AI was overly programmed to stop you (or any player/AI winning) as it would make cultural victory virtually impossible - i.e 3 AI's DOW you and all focus on 1 big city!! Also generally I think the game would feel frustrating - so not quite sure how this could be done.

Generally would feel nice to see the AI make more of an effort though or prioritise correctly - I play knowing if Mansa is in the game - even though he may be way ahead in tech for Space Race - I've never seen him pursue it - he always seems to stop and try culturual victory (which he can be dangerous at - but thats not the point!)

I alwasy just turn off all other victory conditions other than Domination.

 

[TheLazyHase]

I would tend to think that the preference for how to make the AI harder would be to improve the AI's gameplay (as this does), and, only upon failing that do we tolerate a simple brute-force handicap for the AI to simulate such a challenge.

Note that thoses modifications does not create a smarter AI. It's more of a stupider AI, more easily trecked out. But it exploit :
a) a strategy that is badly countered by human since they often build way too much instead of fielding troops.
b) the AI advantage over unit building at the max.
So, no it does not improve AI gameplay. It just force a brutal strategy.

 

[Gurgeh]

If the AI played to win, it would be far less dynamic to counter it and the game far less practical. Let me give you an example of an almost-impossible-to-beat-on-high-levels AI I did in the past, using ONLY XML value changes:

1. It is creative and imperialistic.
2. It has a unitprob of 80, so it builds units twice as often as shaka
3. It is not willing to declare on a target unless it has over double its power
4. It will trade tech at any disposition
5. It will declare war at any disposition. In fact, the moment it has enough power it plans a war on that hapless target.
6. It is willing to trade resources at any disposition
7. Its favorite tile improvement is the cottage

A deity player tried a game with this AI at immortal and promptly got spanked (essentially, the way it was set up guaranteed the human was the target if on the same continent, and he ate a TON of units on that declaration...!). However, such an AI is a threat anywhere:

- It will tend to vassal every civ if on another continent, and actually has enough units in its naval stack to be a threat
- Since it trades MORE often than mansa musa, it isn't as far behind in tech as expected.
- If it's on your continent and the AIs are getting a lot of bonuses, you're probably toast. For one, it will never, ever leave you alone and the other AIs tend to like it.

There is a good reason fireaxis did not make an AI like that.

Hahah! Sounds like the frikkin Terminator.

 

[r_rolo1]

I always had a hard time beliving in the notion that making the AI play like a human will make necessarily make the AI play better. Sure, a Ai like that will smash the current AI, that are coded to have some roleplay on it, but it seems nobody thinks what will happen if you put 7 or 17 of those in a map with a human. Will a AI like TMIT made perform better? Or they will pass the rest of the game killing eachother, manouvered in the shadows by the human? To say the truth I think that the current AI ( this includes the better AI ) does not gain much in average and in fact can lose some performance if you go that way, just because the AI has no central planning, no sense of oportunism besides a very crude backstab mechanism and no memory, a thing that is in the core of how humans play ( humans backstab, but the normally prepare the backstab for ages .... humans engulf, but the first isolate the victims and/or try to make it lose their army elsewhere, humans will build cottages but only in cities that can sustain them in a sensible way ( atleast most of the times ), humans build units, but they don't build them all the time, etc )

And about the OP complaint: yes, you are right, the AI should consider making a burning if another player is gunning for space or culture ( or even diplo ). And i partly agree with what you say in post #5 ( in spite of those "should I buddy the big boss or ally with #2 to take him down" calcs being extremely complex even for humans ). The big issue is what happens when we start putting more than a AI in game vs the human.... a perfectly sensible strategy in a 1:1 basis ( grabbing a nice SoD and going to burn a city the enemy needs for winning ) may not look so sensible when you start putting 3 or more players ( ok, torching the cap of my enemy to stop his win is a good idea, but he can bribe civ A to attack me in the moment I do it... then my army would be elsewhere and i would definitely lose the game...) ,that OTOH could bring a whole diferent set of strategical considerations to have (.... But i could bribe Civ B to attack civ A before I go torching the city or bribe civ C to attack my enemy instead, meaning that i would need far less troops to attack them, or... ). Managing this, even for a gifted human, is a hard job ... And things would obviously get worse when multiple fires start burning.... what would the poor AI do with 3 or 4 oponents gunning for space and culture? Should the AI drop it's wars with meaningless ( in the sense that aren't direct contenders for the win in this moment ) oposers to move it's troops against a possible runaway or should press for the victory first and attack later? Should the AI stop or slow down it's SS or culture plans for the win to stop others or forget about it and put more steam in their own plan? This are questions that even humans, that are normally blessed with a decent long term planning abilities and memory, find hard to manage unless the oponents are dumb as bricks or completely unimaginative.....

 

[Badtz Maru]

I used to love modding the AI XML files. I was disappointed in how Montezuma almost always was last in score, so I made him less likely to declare war on people more powerful than him, made his favored civic Theocracy, and tweaked his tech priorities a bit (I can't recall what exactly) and after that he was almost always one of the top 3 civs.

 

[TheMeInTeam]

Will a AI like TMIT made perform better? Or they will pass the rest of the game killing eachother, manouvered in the shadows by the human? To say the truth I think that the current AI ( this includes the better AI ) does not gain much in average and in fact can lose some performance if you go that way, just because the AI has no central planning, no sense of oportunism besides a very crude backstab mechanism and no memory, a thing that is in the core of how humans play ( humans backstab, but the normally prepare the backstab for ages .... humans engulf, but the first isolate the victims and/or try to make it lose their army elsewhere, humans will build cottages but only in cities that can sustain them in a sensible way ( atleast most of the times ), humans build units, but they don't build them all the time, etc )

Don't underestimate "runaway ai". There are some aspects of it that make it COMPLETELY IMPOSSIBLE on high levels, if all the AIs do it. For one, the human would be dogpiled instantly if every AI were like it, because this AI cares naught for diplo and will only declare on someone if it has a massive power lead.

Initially I thought this would deadlock it, but after DanF's AI DoW thread I learned that isn't the case. Because of how power gets added to one's own for war considerations, this AI is exceedingly likely to backstab civs at war. Although I've not playtested a game with only these AIs (it would be hell), my guess is that one with a favorable start would finally get a massive power lead on someone, declare, and then a chain reaction of dogpiling and backstabbing would be set off. Someone would emerge victorious from that and probably be in control of the game from there. Does this scenario sound familiar though? I see a LOT of MP games do just this as well...provided someone doesn't get choked/rushed and quit super early.

Also, you can mod an AI so it can't be bribed to war. I forget whether I did so but I think I did (I ONLY wanted it at war if it had a substantial power lead and could succeed).

Such an AI would probably seem very easy to a good player on noble, moderately challenging (but only early on, then the human would take off) on monarch, and completely impossible on immortal+ with a lot of starts if all of them played this way. Not only that, but this kind of AI would mop the floor with other normal AIs as you pointed out.

I argue that it is tough to classify it as anything BUT more effective, but is that better? Fireaxis wanted a roleplay element to the AI and unlike some of their other decisions (events, patching certain things, the controls, etc), I agree with them. An AI of the nature I describe would not lend itself to dynamic gameplay, it would feel like a pangaea FFA in MP every single game with minimal differences between which civ you face. That would lose its SP fun very quickly and every victory condition except a military-based one would be either impossible or impractical (the only way to win space or culture would be to subjugate everyone militarily first, but then why bother...and diplo is literally impossible because I make it so the AI can't like you enough!).

 

[r_rolo1]

I'm pretty sure that I could beat easily your AI*7 even on emperor ( maybe, with some luck on immortal ) by manipulation of the diplo, even if the civs were unbribable. In fact I would bet you could do the same I would only need to find a way of exploiting some initial border hates and play defense Greed is very exploitable, you know .....

 

[Badtz Maru]

I'd be interested in seeing how challenging the AI could be made on levels where it does not get bonuses over the players - how tough could you make a Noble AI? I know a good player could always dominate any computer player near it, regardless of how it's behavior would be changed, but I'm wondering if you could make an AI that would consolidate it's neighbors into a superpower powerful enough to be a challenge on Noble by the time the player came into contact with it.

 

[Mittens]

As many others have stated, I think the game is far more interesting in that your opponents act realistically as opposed to being goal-oriented, it's one of the things that makes CIV such great game and so different from other games (and might I mention, probably the only non-broken game where the single player mode has more depth of strategy than the multiplayer mode). Keep moving up levels if you find yourself cruising to easy space wins.

 

[Badtz Maru]

As many others have stated, I think the game is far more interesting in that your opponents act realistically as opposed to being goal-oriented, it's one of the things that makes CIV such great game and so different from other games (and might I mention, probably the only non-broken game where the single player mode has more depth of strategy than the multiplayer mode). Keep moving up levels if you find yourself cruising to easy space wins.

It's not that I'm always cruising to an easy space win. I have to survive to that era and have good tech and decent production to do it, which doesn't always happen when I play on levels that are challenging to me. The problem is that, if I survive to the modern age and have decent tech (I don't even need a tech lead as the AI is very bad at building the spaceship), I can get a spaceship win with no challenge if I play diplomatically. I think that, if you are not the dominant force on the planet and there are other civilizations that are much more powerful than you and not your friends, they shouldn't make it so easy to win by spaceship (or culture). The AI should not all band together against you, or make huge sacrifices to keep you from winning, but if all it takes is a war declaration and a moderate-sized invasion force to keep from losing, the AI should try that instead of wasting time building the first spaceship parts when you have already finished.

I think it is realistic to try and stop a competing civilization from beating you to the stars. If there was still a Cold War situation between the USA and Russia, you can bet neither side would want the other getting a head start on colonizing the galaxy and would take steps to stop it if that was about to happen.

I've played Galactic Civilizations 2, and though there are things I don't like about it (mostly aesthetic), the AI was very good at trying to win, but it wasn't unrealistic in doing so. The AI seemed to put a priority on survival at any cost, and if one race was about to be wiped out by an evil civilization and had no chance of a comeback, they would almost always surrender to one of the good civilizations to ensure their survival. The AI was also aware when another player was getting close to one of the victory conditions and would modify their goals to keep you from doing that.

 

[Lay_Lay]

If it is a Noble level AI, like many players at that level, I might not expect that he even noticed that you launched your spaceship.

 

[TheMeInTeam]

I'm pretty sure that I could beat easily your AI*7 even on emperor ( maybe, with some luck on immortal ) by manipulation of the diplo, even if the civs were unbribable. In fact I would bet you could do the same I would only need to find a way of exploiting some initial border hates and play defense Greed is very exploitable, you know .....

It wouldn't work. They'd plan a war on you by about 1500 BC if not sooner, and they'd DOGPILE. You'd need a continents look or isolation, and probably a choke job or two. I don't see how you'd keep up enough power on them, and don't forget that they'd tech trade everything, even monopoly techs, like crazy.

Edit: or would you like me to prove it? I'm good enough with word and find/replace that I could make every AI behave identically without too much time investment...

 

[Isikien]

This might actually be really fun to play against...on noble or prince. I see nothing wrong with making the AI stronger in this way, if it is done in lieu of the random brute-force bonuses that the programmers resort to as a substitute. I would tend to think that the preference for how to make the AI harder would be to improve the AI's gameplay (as this does), and, only upon failing that do we tolerate a simple brute-force handicap for the AI to simulate such a challenge. But the better-gameplay AI is always preferable, I would think.***

Also, if I were going to play against an AI like this, I would insist on having all of the AI's like this. It would ruin the game's intrigue and surprise to know that one AI is going to dominate all of the others from the outset.

In fact, to partially simulate this, I always now play with randomized leader personalities so that I can never count on any particular AI not declaring at pleased, or not being a military or tech threat, or whatnot.

***Maybe this would be asking too much of the programmers, but if there is ever going to be a Civ5, maybe the programmers could include two options for AI personality: competitive AI that is meant to play as much like a human as possible (even with all of the frustrations that that might entail for newbies), and "sandbox AI" as the default toggled option (for either newbies, or those who want a more historical, sandbox feel to the game), which would place restrictions on AI personalities and on the AI's willingness to resort to dirty human-like tricks to win at all costs.

I honestly feel this should be a seperate difficulty level or a mod. To be able to distinguish between handicapped play (which is basically what the difficulties are) and more calculating and doombringing AI would be good.

However for future civ iterations this shouldn't be the case. The AI and their little personality quirks are what make Civ fun, this should be an additional challenge rather than a fundamental concept of gameplay. If the AI didn't have their flaws and would chaotically backstab each other left right and centre including you it wouldn't be all that fun anymore

 

[Zeiter]

The AI and their little personality quirks are what make Civ fun, this should be an additional challenge rather than a fundamental concept of gameplay. If the AI didn't have their flaws and would chaotically backstab each other left right and centre including you it wouldn't be all that fun anymore

Well, here's the way I feel about role-playing AI: it only really works if the human player is willing, to a certain extent, to role-play as well. By this, I don't necessarily mean handicapping one's self when there is obviously a better alternative course of action at hand (although sometimes this can be fun--such as limiting one's self to certain civics or strategies for historical accuracy, as MadScientist's role-play challenges often do). What I mainly mean is: as the human player, not pulling apart every piece of minutiae in the code and shamelessly taking advantage of the AI's role-playing quirks that were put there originally for your enjoyment (and not in order to help you "game" the system, no pun intended). I think Civ should be flexible enough (and it is, with enough modding, thankfully!) to give human players the AI that they deserve. And a human who plays Civ4 only for the competitiveness of it, and who has teased apart every little predictable quirk about AI arbitrariness, deserves an AI that doesn't allow humans to take advantage of itself like that. Humans cannot complain on one hand that the AI doesn't play to win, and on the other hand that it has no character or anything other than a robotic drive to win.

For example: instead of approaching my games in terms of win/loss, I like to approach my games in terms of "novel and interesting accomplishments." A game is good if I accomplish or experience a lot of novel and interesting things during the game. In this sense, having to scramble after having been unprepared for that AI backstab would make that a good game, even if I were to lose. It created a novel and interesting situation--much better than hitting "end turn" and watching either my score, or the AI's, exponentially creep upward, without event, turn after turn.

Alternatively, I would have effectively ruined that game if I had peered into the code and examined every possible factor that could have possibly led to an AI backstab, and if I had used those revealed pre-programmed AI quirks (originally meant to spawn novelty and interest in games) to effectively stamp out all chances of any novelty from happening that might have threatened my game score or my prospects of "winning" the game according to the game's own pre-fabricated judgment scale. I think, after a while of playing Civ, there's only so much fun you can get from doing well according to the set of normative standards that the game gives you. After a while, to keep the game interesting, you have to create your own normative standards to shoot for in a Civ game, and sometimes this means roleplaying a bit and suspending one's disbelief in temporarily imagining that you are playing the game as if your opponents were unpredictable humans. That for me is far more fun. But for those who want to game the system, they deserve (and it sounds like they would much more enjoy) the doombringing AI.

 

[r_rolo1]

It wouldn't work. They'd plan a war on you by about 1500 BC if not sooner, and they'd DOGPILE. You'd need a continents look or isolation, and probably a choke job or two. I don't see how you'd keep up enough power on them, and don't forget that they'd tech trade everything, even monopoly techs, like crazy.

Edit: or would you like me to prove it? I'm good enough with word and find/replace that I could make every AI behave identically without too much time investment...

It wouldn't be worse than AW if they dogpiled and my previous experience says that it is managable up to Emperor atleast ( the AI also gets cozy cozy in AW and tech trade like mad people, so it would not be that diferent ).

You seem to be forgetting that the power count is basically meaningless for the AI war decisions as long as you are below their war threshold ... because of that I AM expecting a dogpile in 1500 , but i ( and every player ) also have a set of tools to deal with it or even avoid it completely...

I would play the game, but I'm in the middle of a changing job process, so I am short on time to play ... and a game like that would be time consuming to say the least.

 

[Isikien]

Well, here's the way I feel about role-playing AI: it only really works if the human player is willing, to a certain extent, to role-play as well. By this, I don't necessarily mean handicapping one's self when there is obviously a better alternative course of action at hand (although sometimes this can be fun--such as limiting one's self to certain civics or strategies for historical accuracy, as MadScientist's role-play challenges often do). What I mainly mean is: as the human player, not pulling apart every piece of minutiae in the code and shamelessly taking advantage of the AI's role-playing quirks that were put there originally for your enjoyment (and not in order to help you "game" the system, no pun intended). I think Civ should be flexible enough (and it is, with enough modding, thankfully!) to give human players the AI that they deserve. And a human who plays Civ4 only for the competitiveness of it, and who has teased apart every little predictable quirk about AI arbitrariness, deserves an AI that doesn't allow humans to take advantage of itself like that. Humans cannot complain on one hand that the AI doesn't play to win, and on the other hand that it has no character or anything other than a robotic drive to win.

Sorry I did forget to say, I do agree with your post and I also neglected to say that I would like a setting seperate from difficulty level that offered a different flavour of AI

Something along the lines of "Historical" or "Roleplaying" AI or having an alternative choice with a "Ruthless" AI.

With TMIT's xml alterations, the runaway AI is good but if that was spread out on all the leaders i think it would make for an interesting game. If each iteration of this AI was given slightly different peaceweights to stop dogpiling of a human opponent (that's the only thing i could think was unfair, if the human entered and became the main target straight away for all AIs) and better yet, if some were given preference towards building less units and more emphasis on science, I think we potentially could have an interesting mish mash of a civ game without compromising too much on individual leader head dynamics

 

[TheMeInTeam]

It wouldn't be worse than AW if they dogpiled and my previous experience says that it is managable up to Emperor atleast ( the AI also gets cozy cozy in AW and tech trade like mad people, so it would not be that diferent ).

You seem to be forgetting that the power count is basically meaningless for the AI war decisions as long as you are below their war threshold ... because of that I AM expecting a dogpile in 1500 , but i ( and every player ) also have a set of tools to deal with it or even avoid it completely...

I would play the game, but I'm in the middle of a changing job process, so I am short on time to play ... and a game like that would be time consuming to say the least.

I won't exactly be able to generate such a thing INSTANTLY, either. I'll be wanting to test a few new things/ideas.

But it probably wouldn't be a total dogpile, and that might serve to make it harder. You are correct about the dogpile war mechanics. Every AI considers the power of either side at war as part of its own when deciding its "power" against the target. If maybe one and DEFINITELY 2 AIs declare on the human, both of those AIs are eligible targets to anybody else around, but having your enemies losing most of their cities might not be the best development.

Also it might be similar to AW but there is some extra stack buildup so it depends.

 

[mfie]

I think it would be fun if an AI attacks someone who is about to win (not just human), ofcourse taking in account how much they like, their personalities and your power in to account. That would meen you have to focus on military too for winning by space (or planning to complete most of the parts on the same turn). That would bring a new aspect to the game

 

[TheLazyHase]

I think it would be fun if an AI attacks someone who is about to win (not just human), ofcourse taking in account how much they like, their personalities and your power in to account. That would meen you have to focus on military too for winning by space (or planning to complete most of the parts on the same turn). That would bring a new aspect to the game

Yes, the "you must have an army able to conquest the entire world to win" aspect. Clearly, it is lacking for now.

 

[TheMeInTeam]

Alternatively, I would have effectively ruined that game if I had peered into the code and examined every possible factor that could have possibly led to an AI backstab, and if I had used those revealed pre-programmed AI quirks (originally meant to spawn novelty and interest in games) to effectively stamp out all chances of any novelty from happening that might have threatened my game score or my prospects of "winning" the game according to the game's own pre-fabricated judgment scale. I think, after a while of playing Civ, there's only so much fun you can get from doing well according to the set of normative standards that the game gives you. After a while, to keep the game interesting, you have to create your own normative standards to shoot for in a Civ game, and sometimes this means roleplaying a bit and suspending one's disbelief in temporarily imagining that you are playing the game as if your opponents were unpredictable humans. That for me is far more fun. But for those who want to game the system, they deserve (and it sounds like they would much more enjoy) the doombringing AI.

Part of the reason for the OVERWHELMING advantage granted the AI on deity is the fact that it roleplays. If human players are not willing to code dive to at least some extent, that level isn't reasonable. Knowing the AI is the human's handicap advantage, while the AI gets double production speed, an extra settler, practically no maintenance or war weariness, and 50% upgrade costs among other things to compensate.

Pull the AI nonsense, and you have to pull some handicaps too. Same thing with the code - many aspects of the code are essentially convoluted gameplay rules. I must admit I am somewhat disgusted by the number of people who are willing to accept or even desire hidden rules or a misleading (or flat out false) GUI.