Max level - no AI cheat

[headcase]

Players have been whining about "I want a smarter AI and one that uses less cheats" for a decade, and when they actually get that (at least as far as diplo AI is concerned) they throw a complete fit because now they actually have opponents who aren't just paper tigers there to be killed at your leisure. It is hilarious.

I like the new diplo but it's annoying when a faraway civ uses their cheats\boosts to become a powerhouse early on, then decides to go to war with you while you're still fighting someone else. Sure on Prince (even Warlord) the boost are mostly just happiness, but happiness is a huge factor that directly results in allowing a larger population.

If, on Prince, they had a 5% boost on apples, hammers, gold, beakers, culture, great people, and happiness, I think that would be a lot more enjoyable to play against than the current setup of what is effectively a near +100% boost on happiness resulting in near-permanent golden ages while owning half the world in the Medieval age.

This is the logic I used when saying the current boost on Prince is +100% happiness, it's off the top of my head so feel free to point out any improper logical leaps:
- They get only 60% of the unhappiness- and +1 happiness per luxury, and I think +3 happiness off the bat compared to Prince. Let's say that's around 50% of the total unhappiness roughly.
- Getting half the unhappiness means each goes twice as far.
- so it's sort of like double happiness(?) Which is +100%.

 

[Peng Qi]

Actually, it is cheating, difficulty in a decent strategy games is almost always determined by the choices the AI makes, not by the absurd bonusses it gets.

Even if the AI is'nt technically cheating, the dev's did by, instead of making a real difficulty they just made a joke version that takes about 5 minutes to program and shows very well how much they care about their players.

Can you name one of these "decent strategy games?"

 

[Derpy Hooves]

To the one dimensional question above me, SC2, M.A.X., HoMM, the older incarnations at least.

And SC2 should not be used as a direct example, it is a completely different beast, altough I still maintain a more complex one.

And on IV vs V diplomacy:
They are both at about the same level of complexity, V uses a bit more RGN values but that's about it.
Only one of them is terrible at what it is supposed to do though.
The main difference is modifiers and "realism", where CIV had a large amount of positive modifiers you could employ, most of these have been removed or nerfed (religion, open borders, trading, etc) and a good amount of negative modifiers, most of them very hard to control have taken their place.
"Realism", appearantly CIV AI was too easy to game so they responded by makin everyone likely to backstab, the AI ganging up against the leader and asking discounts, to simulate real players, not leaders.

So CiV ends up with an incredibly hostile AI that is almost impossible to influence because they removed the positive modifiers and its impossible to control absurd negs like "coveting lands" or "coveting wonders", so you end up with a AI that makes it almost impossible to engage in actual diplomacy.

CIV diplo AI might have had it's problems, being a bit too simple and easy to exploit, but CiVs version can still be easily exploitable and CIV gave you the feeling you where engaging in actual diplomacy because you had a lot of options to work with and allies would not backstab you over a single rgn roll.
While CiV feels like trying to engage in diplomacy with a pack of rabid, feral toddlers.

 

[headcase]

Oh the older HoMMs had plenty of "tricks" to get the AI to perform better. The difference is that to the best of my knowledge, it didn't have a big banner saying something along the lines of "The AI is treated just like a player so if you lose it's because our AI is better than you lol", like the tooltip for Prince in Civ 5.

Just read GameFAQs and you won't find many strategy games without at least one seemingly small detail that works in the AIs favour. That's fine, but the problem with Civ 5 is the massive happiness bonus instead of small bonuses everywhere. Makes it feel too asymmetrical.

 

[Camikaze]

Not sure if it's been mentioned yet, but this page might help answer the OP's question.

 

[BuilderFTW]

Regarding the ai "cheating" I do not think of it as cheating. Think of it from a role playing perspective, did Alexander the Great sit down and cry, saying his enemies where cheating because the armies he faced were 10x the size of his? No. The point of being a great leader is that you take your people(a small upstart nation) and drive them to greatness above other larger nations. Think of the ai bonuses not as cheating, but rather as the other nations being already established countries which you must either ally with or conquer to make your way to the top.

 

[Trias]

And on IV vs V diplomacy:
They are both at about the same level of complexity, V uses a bit more RGN values but that's about it.
Only one of them is terrible at what it is supposed to do though.
The main difference is modifiers and "realism", where CIV had a large amount of positive modifiers you could employ, most of these have been removed or nerfed (religion, open borders, trading, etc) and a good amount of negative modifiers, most of them very hard to control have taken their place.
"Realism", appearantly CIV AI was too easy to game so they responded by makin everyone likely to backstab, the AI ganging up against the leader and asking discounts, to simulate real players, not leaders.

The diplomacy modifiers in civ4 had nothing to do with AI. Instead, they were an asymmetrical gameplay mechanism, that applied only to the computer controlled players. Meanwhile, the humans players were free from the modifiers and could do anything they wanted in the diplomacy. In terms you are using, this was a cheat for the human players, which persisted to diety difficulty.

The civ4 AI, made only a very limited attempt to make diplomatic decisions that were in its own interest. The civ5 AI, at least makes an attempt to determine a long term diplomatic strategy. Yes, this means that if the AI decides that player X's existence is detrimental to its goals (this is effectively what "covet lands" and "trying to win" modifiers mean), there is very little that player X can do about this. This is a problem with the game design of the civ series as a whole: in the end, it is almost always better to be rid of an opponent.

This is one area, where I would have made different choices then the devs did for civ5. Instead of removing all of the diplo modifiers, I would have changed the game mechanics such that the modifiers made sense for both human and AI players. (One way this could have worked is that the modifiers were made to refer to the attitude of the population rather than the leader. Making decisions against or with the will of the people (going to war with civ like by your population) could lead to happiness bonuses or penalties. This would have worked rather well with the global happiness mechanism.)

 

[Peng Qi]

To the one dimensional question above me, SC2, M.A.X., HoMM, the older incarnations at least.

The statistics I see show SC2's AI on Insane getting a 40% income boost, so it does, in fact, cheat. Not familiar with the other two games you mention.

 

[Derpy Hooves]

I already answered questions concerning SC2 earlier in the thread, read the entire thing if you really want to continue talking about it.
SC2 AI only cheats on the highest difficulty, the rest is all specific behavioral programming, CiV in contrast uses cheating (and the ability to declare war) as it's ONLY means of differentiating between difficulties, a very, very lazy solution.


And Trias, what you call assymetrical modifiers IS Civ diplomacy, whether you like it or not, I believe that most people prefer the old system though (I think this is one of the main reasons why people are constantly reffering to CIV when it comes to improving this game), the old diplomatic AI might have felt a bit static, but for it's purposes it was very sufficient and worked well as far as immersion goes, I think we need to dive into that subject a bit before we can fully understand WHY the CiV diplomacy fails so hard.

Civization has always been a simulation game, create a fun, workable rough equivelant of human development. You make the game workable by creating systems we can exploit, tech trees, buildings, diplomacy, warfare, good decisions get rewarded and bad ones punished.
The fun consists of 2 factors, rewards and immersion, rewards are gained from playing the game well and immersion is created by making scenario's conform to our expectations.
For instance, the immortal leaderheads are obviously highly unrealistic but work wonders for immersion, it might be normal for a country to get into 3 wars in 100 years, this however is not fun to experience with a system where 100 years can last a single turn.
Our human minds need a stable known system, human history might be chaotic on average, we can only think of single situations, we want trustworthy allies that simulate real world bonds, we want stable alliance blocks, we want a AI that acts like a real leader, not a player.
Noone but the most hardcore obsessed players that play Deity purely plays this game for the challenge, and most people will prefer a worse playing AI (just diplo, combat is still a disaster) if it means that they will have a more enjoyable experience.

Babies and bathwater, they felt the AI was too easy to game so instead of improving it's decisions they made a new one only focussed on this aspect, dropping all the improvements made in previous Civ games, I personally blame Schafer and his new <snip> philosophy, it didn't need one, just improvements on the old system.

Moderator Action: Inappropriate language removed.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889

 

[headcase]

Think of the ai bonuses not as cheating, but rather as the other nations being already established countries which you must either ally with or conquer to make your way to the top.

I wish they did make it that way; first thing I'd do in a mod is make it so, in higher difficulties, the AI starts earlier than you do, but does not receive any extra units or techs to start with (but still receives boosts to everything on King+).

Instead, we are left with the "AI receives no particular bonuses on this setting" lie that makes you feel like you're playing the game wrong. when AI is getting advantages when you are playing on "Easy" (Warlord), that's somewhat insulting to the player's skill.

Thanks to Camikaze's link, I now understand that the situation is worse than I thought; it appears that on Prince, they get policies at 67% the cost, techs at 95% the cost, extra catch-up on techs completed by other civs(?), lots of barb bonuses, cheaper upgrades... on top of the 60% unhappiness. Given how their civs do even with all those bonuses, combat might actually the AI's strong suit.

If I understand what that link is saying, Chieftain is the max difficulty where you and the AI have equal boosts to your civ, to answer the OP.

The AI plays at chieftain difficulty level and get those bonuses as well; eg, if the player chooses deity, in addition to the 60% unhappiness listed below, the AI has 60% from playing at chieftain difficulty and those get multiplied together, resulting in 36% unhappiness from cities and population for the AI.

 

[Trias]

I already answered questions concerning SC2 earlier in the thread, read the entire thing if you really want to continue talking about it.
SC2 AI only cheats on the highest difficulty, the rest is all specific behavioral programming, CiV in contrast uses cheating (and the ability to declare war) as it's ONLY means of differentiating between difficulties, a very, very lazy solution.

I would suggest learning a thing or two about programming, and actually thinking a bit about the challenges of programming the AI for a 4X game. If you do, you will most likely reconsider your hasty judgement that this is a "very, very lazy solution".

(For example, you seem to completely fail to grasp the difference in complexity between programming a simple script type AI, such as in SC2, and the more adaptive type AI that is need for mod compatibility in Civ.)

And Trias, what you call assymetrical modifiers IS Civ diplomacy, whether you like it or not,

Whether YOU like it or not, this clearly is no longer the case. Which in principle is a good thing. Since whether you like it or not it was a form of cheating for the human player. (Since you insist on calling any sort of asymmetrical advantage for a player a "cheat".

I believe that most people prefer the old system though (I think this is one of the main reasons why people are constantly reffering to CIV when it comes to improving this game), the old diplomatic AI might have felt a bit static, but for it's purposes it was very sufficient and worked well as far as immersion goes, I think we need to dive into that subject a bit before we can fully understand WHY the CiV diplomacy fails so hard.

See the last part of my previous post.

Civization has always been a simulation game,

No, civ is not and never has been a simulation game. Civilization is a 4X game set to the back drop of human development.

Our human minds need a stable known system, human history might be chaotic on average, we can only think of single situations, we want trustworthy allies that simulate real world bonds, we want stable alliance blocks, we want a AI that acts like a real leader, not a player.

Most players also want an AI that poses a challenge. A big frustration with most civ players, is that the AI inevitably falls to far behind to be a challenge. (It also break the immersion by emphasizing that apparently the human player is not playing a real leader.)

Noone but the most hardcore obsessed players that play Deity purely plays this game for the challenge, and most people will prefer a worse playing AI (just diplo, combat is still a disaster) if it means that they will have a more enjoyable experience.

As I have said, the problem here is more fundamental. Namely, civ rewards being a backstabbing SoaB.

Babies and bathwater, they felt the AI was too easy to game so instead of improving it's decisions they made a new one only focussed on this aspect, dropping all the improvements made in previous Civ games, I personally blame Schafer and his new <snip> philosophy, it didn't need one, just improvements on the old system.

1) It had nothing to do with the AI being to easy. It had everything to do with the game being to assymmetrical between the human and the AI.

2) What makes you think that the AI, is completely new? (It is much more likely that they started out by modifying the code that was already there.)

Moderator Action: Please make sure you don't quote inappropriate language.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889

 

[sulx]

I just want to chip in here and clarify a few things about the AIs of Warcraft III and Starcraft II.
As someone that has played both those games pretty competitively (Was often around the top1000 ladder players in War3, masters in SC2) I have to say, the AI in both, even on the highest difficulty, is completely incompetent.
In Warcraft, as it is the older game, the insane AI (which revieved double the ressources from mining that you did) was no challenge for anyone that played the game for more than 60 hours in multiplayer, it was in fact so bad anyone decent at the game could beat it with only a single hero (given some handspeed).

The StarCraft II AI is actually half-decent at a lot of things (as is CiV's AI). It actually counters units when it sees them (instead of building the counters no matter what), but of course it is no where near smart enough to make guesses about your tech path based on the structures you make and the amount of gas geysirs you are mining, as a human would.
It actually moves its army together (I guess thats due to the SC2 unit pathing which makes that very easy) and it apparently has some routines to identify when you are weak and will attack you at those times. It also expands reasonably quick and does have some semblance of unit compositions (frontline blockers + backline damage dealers).
But it still sucks hugely. It does not retreat when it is losing a battle, waiting for when almost the entire army is gone to go back. It makes way too many production buildings for the income it has (on non-cheating difficulties, which is everything up to the 3rd hardest). It doesn't follow any stringent tech path. The "counters" it makes never work because it doesnt have the slighest clue about intelligent unit micromanagement. And very importantly it never uses spellcasters - how well you can use those often makes or breaks games in multplayer.
On the highest difficulty it is very much like the CiV AI - it is very challenging in the early stages of the game, but the more decisions have to be made, the worse it gets. In short: Once you hold the initial rushes of overwhelming numbers of units it is a cakewalk. That's assuming you play a somewhat "standard" game, because otherwise you can beat it in around 6-7 minutes with some AI-abusing rushes.

In my opinion CiV's AI is on par if not better than the AI for SC2/war3, especially considering the insane complexity that CiV has. Granted I only started playing Civ-games with this one, but it has taken me almost 300 hours of play time to beat Deity more than half the time (and I had the help of all these nice guides here) - the StarCraft II insane AI I could beat after playing the beta for a total of maybe 100hours.

 

[vexing]

to answer the op, i'd say player at warlord is pretty close to even bonuses, the mix of bonuses for player and AI penalties ends pretty close to even.

to get a perfectly "even" game, you'd want to change the AI_HANDICAP value in the defines.xml to 3 (from 1) and play at prince.

 

[Flaxton]

Who says you can't program a decent AI??

I find that offensive.

I am an AI, just ask Turing.

 

[Derpy Hooves]

Trias, it seems your not able to look past simple 'script' labels as you show no desire to find any programming compromise.
As I've discussed with Phil before, a perfect AI would take years to program, an adequate one however does not and can be done by a series of fairly simplistic scripts.
Yet they chose to try neither and just let the AI remain in it's broken state. IE lazy bastard solution.

Assymetrical modifiers are necessary to create a working diplomacy in single player, unless you want to program a AI for diplo that is actually good, something they clearly did not do here, again, you don't want an AI that can only consider its own short term gains because it creates bad gameplay in a game with a focus on this aspect.

You say it's a 4x game, I say simulation, people want a simile between the game and their experiences, if you don't your likely one of the obsessive Deity players and kind of irrelevant for this discussion as you have very particular criteria of interest.
Backstabbing is a problem but just a fragment of the overal problem where the AI can only think on very short term.
1)Based on the removal of a lot of positive and adding of negative modifiers I'd say you are wrong or our initial statements where not exclusive.
2)I have no doubt of this, but due to the result and the changes in the framework this was, once again, a lazy option if not a complete mistake.

 

[Derpy Hooves]

Sulx,
Like you state yourself, the SC2 AI makes use of counters and generally makes a good amount of decisions, the CiV AI does not.
It's selection of units is based on what 'flavor' the AI has, not on what it needs to counter you.
It still has no understanding of formation, not even for its stackable GG's.
There is not a single problem with the SC2 AI you list that the CiV AI does not also suffer from, and this is 2 completely different genres.

Your other points are misconceptions, CiV AI is far worse, brutal SC2 does not get even close to the bonusses a deity AI gets, not to mention that you don't fight 8+ brutal enemies, if Brutal AI just got free units, couldn't be supply blocked and just got lump sums of gold and vespene I'm sure it'd be more challengin to a, let me emphasize, MASTERS player.
Not to mention that as you where top 1000, you where already very proficient with RTS and Blizzard games in general. (Not to mention that W3 was a lot less a-move focussed)

To create the current SC2 AI a lot more work had to be done then for CiV, they could make a combat AI with scripting and did, for CiV, they did not even try.

 

[Becephalus]

Derpy- I don't think you actually understand anything about programming and AI or game design so I am confused why you continue to make points in this thread when everyone is telling you you are wrong.

Take it from people who have studied game design and symbolic logic that coding an AI is no trivial task. Just because you want to rage against Firaxis for their so-so game doesn't mean every point of rage you have is an accurate one.

The AI is CiV is REALLY REALLY BAD compared to a good human player. But so is the AI in EVERY strategy game. The only ones where it is not is those where the AI is covered up because their are either:

Fewer decisions being made so it has less chance to screw up.
Better hidden bonuses so it is still getting cheats, but you simply don't notice.
Some real time element where it can use its superior speed to some small advantage (usually it doesn't even help that much honestly).

They cannot make an AI that can play Go and go is about a billion times simpler than CiV and I am honestly understating it.

The simple mini game of "what should I build in each city each turn" (only a tiny portion of what the AI needs to consider each turn) is much much much more complicated than go, chess, et cetera.

Start trying to write rules for it that are not just build orders, particularly past turn 50.

Do you build a unit or a building or a wonder or wealth? How do you decide? What equations do you put in to weight them? What are those equations measuring?

Lets assume you somehow come up with a passable algorithm to answer that are going to build a unit? Which one? What about the next city, is it going to build the same unit? Shouldn't you need a more balanced force? What about maintenance, what about enemy proximity? What about your diplomatic status?

I could spend 2080 hours writing AI for CiV and maybe I could come up with some improvements in some areas. Heck it would probably take me 2080 hours just to understand the entirety of the current AI.

Just to be clear I fully support the following notions:
That Firxais should improve the AI as they are able, either through past sales monies or through AI focused DLC/Expansions.
That the AI in CiV is disappointing as an opponent.

 

[Derpy Hooves]

Becephalus, you mean those people who tell me I'm wrong and are all just quoting the same few guys in the couple of threads they read about this?
Whereas you just take their opinions for truth without doing any critical thinking yourself I will actually look at more possibilities then the perfect AI previous threads alluded to being impossible.
As I stated before, it's not perfection we're looking for, just something better then what we have now, all knowledgeable posts on the subject refered to chess level AI.
Even that post about GO's complexity your quoting from is talking about perfect AI.

I also find it amusing how you state I know nothing about AI programming and then you just theorize a load without any knwledge on the subject.
The rest of your post is just repeating an paraphrasing earlier posts.

I understand your (an other peoples) sentiments, you care about Firaxis because you have history so you want to defend it, this is not the company you used to know though.
Since the CiV release they have made: DLC (IE no effort), some patches, facebookCiv and X-Com, still no expo announced (which usually happened long before release, implying no expo) and the DLL should have been released months ago according to their own statements.

This is not the Firaxis you used to love and adore and who loved and respected their fans in return, just an empty corporate husk.

Moderator Action: Take the rant (the second half of this post) to the appropriate thread and don't pollute other threads with it.

 

[Trias]

Trias, it seems your not able to look past simple 'script' labels as you show no desire to find any programming compromise.
As I've discussed with Phil before, a perfect AI would take years to program, an adequate one however does not and can be done by a series of fairly simplistic scripts.
Yet they chose to try neither and just let the AI remain in it's broken state. IE lazy bastard solution.

A simplistic "script" type AI, would be completely unacceptable for civ, since it would not be able to deal with mods. You seem to completely missing that point.

you don't want an AI that can only consider its own short term gains because it creates bad gameplay in a game with a focus on this aspect.

I find this comment somewhat hilarious, since the AI in civ5 is the first in the series to actually make long term considerations. (The AI in civ4 basically made short term decisions only.)

You say it's a 4x game, I say simulation,

You may say it is a simulation, but you are pretty much alone there. The rest of the world knows civ as (the prototypical example of) a 4X game.

1)Based on the removal of a lot of positive and adding of negative modifiers I'd say you are wrong or our initial statements where not exclusive.

Actually, the only criteria used for adding/ removing the behaviour modifiers of the AI is whether they made sense in terms of the interests of the AI. That the result is that many "positive" modifiers were removed is indicative of a deeper problem with the civ series, namely that there is relatively little incentive for long term cooperation.

2)I have no doubt of this, but due to the result and the changes in the framework this was, once again, a lazy option if not a complete mistake.

I don't think you are in any position to make judgements about what is lazy or not.

 

[Derpy Hooves]

For gods sake at least try thinking outside of your usual confines, if a script AI would have problems it wouldn't be mods, the strange units from NIGHTS should give you your answer, either the AI treats them like its normal units, i.e. treats them like cannon fodder, or the mod maker has to write a specific script for it, giving exactly the same situation as thr scripted AI but without competent combat.
At least I think this is what you're referring to, it's a pretty vague blanket statement so it's a bit difficult to figure out.

I think in your part about the AI planning ahead you mean the long term AI that really only determines their VC? It still doesn't do any planning, and where the CIV AI would not make the best decisions for stopping your VC at least they where not suicidal.

Actually every post I've read classifies Civ as an empire building game, so as far as general concensus is concerned, you're wrong.

About the modifiers, what they did was rip away the band-aid that kept a massive festering wound under wraps, and instead of stitching the wound they just left it to rot and spread, messing up other parts of the game, lazy and a terrible choice.

And I don't think you're in any position to decide what I find lazy or not.