Max level - no AI cheat

jeez. these AI threads are so simple to solve.

think you can write a better AI? go and do it then.


away you (you as in general to everyone reading this post), away you and get paid for writting said AI. go chase those dollars if you think you are up to it.

i bet there are thousands of companies that will pay you a fortune for good AI.



pretty well every computer game ive played, and thats been since the birth of computer games, has upped the difficulity by giving the computer more bonuses.
proof? :mischief:

:D
 
Thanks for the invite, but I will not chime in any more. I do not consider you a reasonable sparring partner. What you have going for you is that you can voice your opinions well. Otherwise, I think you are rude, unreasonable, and pompous. Also, your way of reasoning does not conform to the conventions of reasoning. Also you seem to have the notion that you can decide, on a public forum, who should and who should not chime in on the discussion. If that is not the height of arrogance I do not know what is.

Anyway, consider this discussion over from my part. I am very much willing to reason with reasonable people. You have a tendency of showing that you do not belong in that subset of people.

Moderator Action: Discuss the points, not other posters.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
 
Thanks for the invite, but I will not chime in any more. I do not consider you a reasonable sparring partner. What you have going for you is that you can voice your opinions well. Otherwise, I think you are rude, unreasonable, and pompous. Also, your way of reasoning does not conform to the conventions of reasoning. Also you seem to have the notion that you can decide, on a public forum, who should and who should not chime in on the discussion. If that is not the height of arrogance I do not know what is.

Anyway, consider this discussion over from my part. I am very much willing to reason with reasonable people. You have a tendency of showing that you do not belong in that subset of people.

If its too hot get out of the kitchen.
Consider that if you wherent so aggressive, with 3 people no less, the discussion might not have taken that turn.
Read the latter parts of the no iron thread I posted if you really need an example of what a good discussion looks like, its far more in depth then anything i have ever seen you post, im not sure what your defenition of the conventions of reasoning is but it most certainly does not feature logic or deduction.

My comment against wannabe was completely justified as he initially did not understand the discussion and only followed it up by attempts at insults, he had not brought anything of value to the table in about five posts, i fail to see what allowing him to continue harassing would accomplish.

Its ok though, from your initial post you did not seem knowledgeable on the subject and where just spouting random tangental stuff, i only included the invitation because it seemed like the nice thing to do, but your clearly still too emotional for reconciliation.

Good luck having reasonable and intelligent discussions with wannabe.

Moderator Action: And don't hit back.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
 
I challenge the first statement. In both starcraft and warcraft 3, some of the finest games rts has to offer, the AI would receive bonusses in skirmish matches on the highest difficulty levels. This completely contradicts what you state.

I would assume that difficulty almost always comes from the bonus the AI gets, rather than the choices it makes. The AI cannot compete with ever changing strategy and human adaptability, and it needs different tools than decision making in order to pose a challenge.

Let's not forget that RTS does not require the AI to make as many decisions from as many possible options.
 
Battle royal!!

Just kidding.

It’s amazing that this AI debate still evokes such a passionate discussion. Even among the haters, can you really say there is a huge difference between the AI in 4 vs 5? Isn’t this difference more than made up for by the new hex grid and single stack play which caused the AI difference in the first place?

Moderator Action: Please avoid calling people 'haters'.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
 
No.
Diplomatic ai is far worse with the new philosophy of playing like real players, which comes down to them focussing on making you lose instead of winning themselves somehow.
Combat ai is hard to compare due to differences in system, but what we can say is that CIV combat ai was actually good at what it was supposed to do.
I would compare it to an artist making a nice drawing with crayons in an hour to then conclude he needs to draw the Mona Lia in crayons and the same amount of time.
 
Well they tried something new. If it was just more of Civ4, then it would be an expansion.
 
Inovation is great, but in this case they threw the baby out with the bathwater.

If your studio is too small/incompetent to make a hexagonal war AI find an alternative that you CAN work with.
Or get a bigger/better studio to help you out, as far as the war AI goes its a great idea really badly implemented.
The diplo ai was just a massive mistake that should never have been touched by people this incompetent.
If your playing a single player fps do you want your enemies to bunnyhop, camp and call you names because that would simulate playing against humans?
They have strange ideas about games, the great power of a single player game is immersion, a feature which they thought was bad appearantly.
 
Well, there are a lot of people who disagree, that the game is still fun and there are a lot of hours of gaming to be had here. I'm one of them.

I do agree they lost ground in some spots, but gained in others that its almost evened out. Its a worthy building point to improve upon for the series. Versus all previous iterations where the new versions were mostly cosmetic.

Regarding the taunts, they always did that kind of comic relief type stuff. Remember looking better with your head on a pole? Classic...
 
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.

Which rules out Civs 1-5 as good strategy games (not necessarily an assessment I'd wholly disagree with, but for different reasons). This has always been the way Civ levels have been distinguished.

However, as I've noted in a number of past threads, there are fundamental difficulties in programming an AI that gets better at the game, rather than one that has particular bonuses or offsets. Even if you consider Civ V simplified compared with other Civ games, it remains much too complex for a Civ-style AI to play in the way a human would - the sheer length of the game involves too many variables that can influence the outcome at different game stages for an AI to 'plan ahead' or play well to a predetermined strategy, while a chess-style AI that calculates the consequences of its moves four or five turns ahead would not only be computationally extremely intensive with the number of options in any given turn in the game (and the ability to make more than one move) and extremely time-consuming (chess AIs often 'think' about their moves for a noticeable length of time with fairly basic calculations to master), but it would also make the AI incapable of any long-term planning.

Now imagine somehow programming an AI that could play the game at a reasonable level of skill without 'cheats', and then adjusting the AI so that it can perform strategies at 8 different difficulty levels without any bonuses.

The Civ V AI isn't as good as the Civ IV one in any event, but even if it were (and however simple the game may seem to a human), the AI would perform more poorly because the Civ V engine demands more complex decision-making from the AI than the Civ IV one did - most obviously in the combat system, but also with several new variables to consider in diplomacy.

All told, is it any wonder that Civ games have stuck with the bonus system that was introduced 20 years ago with much more primitive AI technology? The only AI likely to be able to play Civ games is one that can learn through experience, as has been demonstrated by an MIT team - however that's apparently also very time-intensive and its impressive win rate was against dumb Civ V AIs. It's not clear if that would perform equally well against a human.

Regarding the taunts, they always did that kind of comic relief type stuff. Remember looking better with your head on a pole? Classic...

But wasn't that the only one? I think Civ III did some fun stuff with diplomatic exchanges, but Civ V's are definitely characterful, with Civ-specific greetings and sometimes taunts. The one sad thing is that the way the player can respond is rather characterless - "Very well" or "You'll pay for this in time" (and is there any difference in outcome between these selections? I haven't noticed civs becoming less hostile if you refrain from 'taunting' them).

Speaking for the current state of the game I still haven't seen anything that makes me think the AI is particularly bad.

I'm enjoying Civ V and my experiences suggest that the AI isn't *as* bad as is often claimed - I have seen sensible city attacks by archers supporting melee units (I haven't even seen the infamous 'leading with siege' very often), and in some contexts sensible and consistent diplomacy ...

*but*

...I wouldn't go so far as to say that the AI is on a par with that in equivalent games. Fighting a war in the modern era, I saw the AI completely fail to understand that helicopter gunships can't take cities, bombarding Brussels down to 0 HP turn after turn and just moving endless helicopters next to the city, where they got shot down. I had a mental image of Genghis Khan looking puzzled as he tried to work out why the helicopters can't take cities.

Using fighters appropriately (as interceptors) and bombers to bomb things is an issue that was solved by past Civ AI, and can't be blamed on any change in the combat system. Aircraft worked the way they do now in Civ IV and, if memory serves, in Civ III, and the changes to stacking rules don't affect them at all. Yet Civ V AIs are consistently unable to use them with the same effectiveness, and pretty much invariably use fighters to bomb things.

Upgrades, promotions and Great People work much as they used to in older versions of the game. From recollection, the AI used all of them more appropriately than it does now. The AI can only make effective use of Great Generals, and even then seems unable to use the improvement or Golden Age abilities (as it can't with other GPs).

Old AI workers would sometimes make forts as defensive structures (although was often not bright enough to keep them occupied by units). I have yet to see the AI construct a fort in Civ V.

AI cities appear to be less specialised than in previous versions of the game.

And as far as your assertion that the AI just changes it's reaction times, it is just plainly wrong, a easy level opponent will always come at you at a pretty consistent time with a balanced group of units, for zerg roaches, lings and hydra's, which obviously is a very suboptimal build as it forces lair tech.

SC2 AI has a limited ability to learn counters, and to throw better mixes of units at you as one approach fails. However, it still plays in the same way - you can rely on being rushed at 7-8 minutes, and if you've planned to beat it once you can do so again with equal ease; the only change the AI will make is to attack with more units or a different composition. Even on the highest difficulty, the AI will not wall off, a simple strategy learned early by most Bronze players and considered essential against Zerg and Protoss. Starcraft is heavily based around responding based on the race your opponent is playing, since builds that beat one race's rush or other strategy won't work so well against the others. The AI will always adopt the same approach whichever race you're playing. etc. etc.

considering it is a game with a lot more variables to calculate as far as the warring goes (infinite amounts of positions, loads of different unit types and abilities, whereas civ just uses copy's of the same concept though the ages) it still holds up very well,

Poor example. Starcraft's "infinite amounts of positions" don't do anything - you gain no combat advantage for attacking in the flank that you don't attacking elsewhere. It's not a tactical game. There are no terrain or cover effects on combat effectiveness. Unit combinations have little inherent synergy - you've got Roaches and Zerglings vs. Marines and Marauders, all you need to program is that Zerglings counter Marauders and Roaches counter Marines. Zerglings don't counter Marines any better with Roaches backing them up than they do on their own, and vice versa - each unit does its own thing. This is a whole level of complexity below factoring in bonuses for supporting units, or determining whether to concentrate your ranged and assault forces against one unit vs. splitting to attack several. Civ units have lots of special abilities that dictate what they do (such as horses not being capable of attacking cities effectively) - granted the AI ignores them, but when was the last time you saw an SC2 AI use Forcefield or Fungal Growth (does it even build Infestors?)?

Grandadmiral, considering they did quite well in CIV I suspect the diplomatic AI might just be another victim of the rather shoddy work done on this game, especially if you take into account that they made a lot of weird choices post release, almost like the did'nt really conduct a beta.

In its way, Civ V diplomacy is as different from previous games as combat, and yes it does require more complex programming, so yes it can be considered partially the result of technical limitations. Nearly every diplomatic option carried over from past games was either changed in some way or relies on mechanics that have changed in some way:

- Gold and GPT: Diplomatic option identical, value of gold very different. The AI seems to behave much as the Civ IV AI would in its valuing of the gold resource, offering and taking deals that would perhaps be reasonable in Civ IV. But no one's going to accept having to pay for peace unless they're in dire straits in Civ V.

- Research agreements of course work differently from tech trading. Nonetheless, this is generally the area the AI is most capable of exploiting, however it will still offer anyone with the cash an RA regardless of diplomatic relationships (then again, human players will often do the same unless there's a clear risk of war before the agreement completes).

- Declarations of friendship and denunciations are new, and are more complex mechanically than counterparts in past Civ games. The main function of both is to influence relations with third-party civs ("I denounce you, my friend now dislikes you more"). A human would use these to take advantage of changed relations to press a friend for war, say, and would target these actions at common rivals. This is outside the AI's ability to calculate; older games just used a system of "We're friends, you get an extra positive in relations with me". The result is what we see: denunciation seems to be largely random, without any obvious relation to friendly status, or is used against civs the AI has previously been at war with. It is not followed by any action that offers strategic advantage, such as denunciation followed by friendship with that civ's rivals, or joint war declarations.

- Trade. Both types of luxury now work differently. Again this doesn't generally appear to factor into AI decision-making - for instance it seems to treat any quantity of a strategic resource as being more or less equally valuable, and it doesn't discriminate between a resource it needs and one it doesn't. This latter is a step down from Civ IV AI, but the former is defensible as a product of added complexity. A human will trade for specific resources (and may even be willing to part with gold for them) if obtaining one will start a We Love the King Day or if unhappiness is hard to manage otherwise. The AI will trade regardless of this context - it will grab luxuries if it already has plenty of happiness (which, in fairness, is quite reasonable if aiming for a Golden Age, but not usually at the cost of gold), and it will always take the same amount of gold in exchange however valuable it finds the resource. I don't recall past AIs being any brighter in this regard - likewise they'd trade any resource regardless of whether they needed it for happiness or health control.

- Defensive pacts. No change, but an observation: Has anyone actually had a civ they have a defensive pact with go to war if they're attacked in Civ V? I can't recall it happening to me.

- Open borders. If I recall correctly, in the past it wasn't possible to have one-sided open border agreements at all (i.e. I give you open borders but you don't give any to me).

- The overall modifier system rewards and penalises a more complex set of behaviours. "We have the same religion", "We have different civics" etc. is a simple logical check - yes or no. Easy for a computer to calculate, and independent of context. "We want the same land, Wonder, victory, city-state" is situational and so relies on much more complex decision-making. Even apparently similar "Our close borders spark tensions" (Civ IV, test to see if your border square shares culture with my border square) and "You settled near us" (Civ V, doesn't require shared borders, and there is no 'cultural control' system over hexes that allows the kind of shared control of Civ IV) rely on more complex sets of variables in Civ V.

- City-states. The diplomacy system is designed to exploit city-states - several diplomatic modifiers revolve around them, their influence system is designed to allow civs to compete for control over them, the exclusive nature of city-state alliance (if I'm allied with Helsinki, you can't be) promotes inter-civ conflict and the inability to influence civs you're at war with adds an incentive to declare peace every so often in order to rebuild influence. Add decision-making over which CSes to go for - which will be influenced by, among other things, what resources they provide, what your current influence level with them is, what their current mission is, what type of CS they are, how specific CS types benefit your strategy, where they are in relation to your empire, where they are in relation to rival empires, whether they are well-placed to make a good ally in war, whether they are placed where they would be a danger to you if allied with an enemy... - and you have a system which is extremely complicated for any AI to manage.

FURTHER EDIT (having failed to indicate the last edit before subsequent posts...):

Diplomatic ai is far worse with the new philosophy of playing like real players, which comes down to them focussing on making you lose instead of winning themselves somehow.

This is the most succinct summary of one of the key problems with Civ V's AI I've seen from you, Derpy Hooves! Agreed, an AI with the mandate "Kill all humans" is not the same as an AI that's "programmed to win". An AI that grabs all the city states and then declares war on you isn't one that's trying to win a diplomatic victory, although it's a tactic a human trying to win might employ - it's one that's trying to stop you winning a diplomatic victory. You'll rarely see the AIs use the same city-state denial tactic against one another (for instance Arabia and India were both competing for diplomatic victory in one of my games, but never went to war and indeed signed several declarations of friendship - although unlike Arabia and me they were on different continents).
 
Not to sound cheesy, but sometimes it takes a step backwards to move forward. As a gamer I recall when multi-core processors first came out. They were a rip-off. They were expensive and the single cores were kicking their collective asses. It took a while for software to catch up and make them a worthy purchase. Now there is no going back.

Hopefully it’ll be the same with the 1UPT hex.. and Civ6 will impress.
 
Damn you Phil! You and you long winded well thought out posts that are torturous to reply to on a mobile phone!

Anyway, blasto, you misunderstand, the taunts are not my problem, its the fact that they chabged ai logic to actually make it more like your playing vs people (only diplomatically ofcourse), for instance, all civs will become hoatile and attack you when you get close to any VC (apart from maybe diplo), them being extremely stingy with trading, the constant backstabbing and a lot less factors to influence them with (like how religion worked great in CIV but if you're up against a human player it could be ignored).

And phil as far as ai in strategy games goes, id say CIV had very little tactical elements, and I'm guessing the same applies to its previous incarnations, as such CiV is the first one with truly tactical combat (ofcourse CIV had tactical building and civics but i wouldn't call this strategy in this sense).
And yes, this problem would not be here if they hadnt made a terrible ai, but that's also exactly the problem, if you're gonna reinvent the wheel you need to make sure you have an axle it fits on.
Right now we got beautiful sophisticated wheels that won't fit and because of it the cart just stand there mass denouncing everything.

Ill continue this shortly when I'm home because scrolling down in a field that you're typing in on an android is absolute hell
 
Right now we got beautiful sophisticated wheels that won't fit and because of it the cart just stand there mass denouncing everything.

Another nicely succinct summation. And for me at least it's not just bad because the AI's worse than Civ IV at diplomacy, it's doubly bad because (a) the AI's worse than Civ IV at diplomacy, and (b) the Civ V diplomacy *system* has (to my mind at least) more potential than the Civ IV one. The trouble is, ideal systems aren't as good in practice as less complex systems when only the less complex ones actually work.
 
Ah, the wonders of having a proper keyboard.
Now where was I...

Ah yes, replying to Phil replying to the entire thread's more meaningful messages in a single post, good job by the way.

With planning ahead I'm assuming you mean the combat, not the grand plan AI (the grand plan AI should actually be pretty doable to create as far as time consumption goes, I expect modders can entirely fix this when Firaxis stops dancing around the DLL release).
I definitely agree making a chess level combat AI would be an extremely time consuming job and probably even require need of a program that can learn, but that can not be the only way it can work if you lower your standards a bit.
Planning ahead would be fantastic, but at this point it can't even create an army formation that is not just sending everything to the field of battle based on when they can get there, and with cities being fixed points it could just run a script to wait when in x range from the intended target.
I would assume a good amount could be programmed by relatively simple "if then" commands, ofcourse you'd have to set up a decent enough framework in which to give the commands in but most of it would just be the 6 directions, immovable terrain and combat modifiers and strengths.

I read in a thread somewhere when I started lurking from someone who I then thought was knowledgeable on the subject (criteria where most likely post count and formulation) that he managed to open the file relevant to the AI and that it looked like they, for the most part, copy pasted CIV's AI, which honestly explains a lot.
The blob movement might simply be that it's trying to move the lot in a stack and keeps getting confused why apart from his GG's everything else keeps getting stuck, and the diplomatic AI might simply be set for a game that still has positive modifiers like religion that don't exist anymore.

Oh, and maybe I should have been a bit more clear, I think the taunts are great, I was only using taunts in the context of a first person shooter in which case it is bad but because real leaders can be evil minded bastards I find it rather fitting, in contrast to them telling you that they're going to go to war with you because your winning the game, which is rather illusion breaking, admittedly I do get tired of the taunts in CiV but that is mostly because I find it almost impossible to see them as actual opponents anymore, but in CIV I loved it.

And about the gunships, I've seen England and the Ottomans try to take city states that where already at 0 hp with ships while being on a completely different continent where she had no ground army. Honestly, if they wanted ships to work, instead of nerfing it, couldn't they just have either added a meelee ship or giving ships a secondary meelee attack that might be pretty weak but at least works in situations like this?

And about the GP's, I have seen them make stuff like manufactories before, and as far as golden ages go, the computer can't actually have a golden age for some obscure reason, maybe because they wouldn't understand it? Then again, what is there to understand about more gold and hammers...

And the SC2 AI is not that simple actually, it's still not a work of art but its still more advanced then the CiV combat AI (or the other Civ games, as noted before, you don't need much AI for a SoD), the rushing is predetermined yeah, it varies with difficulty though and at least has different builds and a limited understanding of counters, whereas CiV's just goes with whatever it's flavor is, or at least I've only seen 'natural' counters, i.e. Bismarck making panzers against infantry or Alex countering horses with hoplites.
It might not know race specific strategies but as Blizzard has been a bit more open with the ability to make your own AI, there are already quite a lot available that might very well have this feature (not to mention that the game has 2 expo packs left to go, whereas CiV...) and I know for certain that they made map specific AI as well, so it even knows how to exploit high ground and positioning.
And attacking in the flank might not give a +x% bonus, but it does give the ability to engage units that you might be able to counter, with Starcraft 2's heavy counter system this can be highly effective, see: the entire zerg race.

Tactics =/= terrain combat modifiers.
Admittedly it's a good system, but would not add much to a game like SC2, just try imagining playing SC2 with lava patches that make you lose 30% armor and attacking power.
And making unit's attack counters is fairly easy to program but I feel it's still just another component the CiV AI lacks.
And when it gets in the problem where it's getting harassed by single units on it's way to a specific target, let it do what real armies did, use it's cavalry to clear the small fry.
And on special abilities, I know for a fact sentries can't forcefield if the AI's controlling them, but considering forcefield is so beyond CiV it's not really relevant, thank god the AI does'nt have to deal with mountains you create during the game. And I might not have seen fungals in a standard AI game but I have on a custom map called Desert Strike (pretty hard to miss), from observation I'm fairly sure it uses a "if x amount of units get into a x radius cast fungal on center" command, not sure it's standard or custom AI though.

And I might add to the list you followed with about the diplo AI that a lot of these problems can actually be explained if they just copied the CIV code, CIV had tons of positive modifiers you could use while CiV has rather limited that while adding a whole new slew negative modifiers, not to mention that research agreements is one of the few things the AI does very well, somewhat remniscent of how capable to AI traded technologies in CIV.
And as noted before the AI can't get golden ages, so excess happiness if even more useless then for a human player.

And the whole idea of wanting it to play like a human is actually kind of hilarious, I'm guessing that after CIV the dev's got feedback that the AI was too easy and predictable, so they thought that by learning 'human' tactics and shoving in a load of random things they could do people would be more challenged.
If you would put this diplo in CIV it would be borderline unplayable because in that game a massed attack from multiple enemies could not just be fended off by some rushbuying and range unit abuse.
So, in an attempt to make the game more difficult they essentially made it hostile, definitely making the diplomacy part harder but also taking away most of the fun.
 
Ah, the wonders of having a proper keyboard.
Now where was I...

Ah yes, replying to Phil replying to the entire thread's more meaningful messages in a single post, good job by the way.

With planning ahead I'm assuming you mean the combat, not the grand plan AI (the grand plan AI should actually be pretty doable to create as far as time consumption goes, I expect modders can entirely fix this when Firaxis stops dancing around the DLL release).

I was mainly thinking grand plan - hence pointing out that a chess-like AI couldn't think far enough ahead. In chess you can't plan your whole route to victory, the game's far too reactive and any competent opponent will have responded to your moves before you get close to completing your strategy. Chess AIs evaluate each move they make in the context of what will potentially happen X moves ahead. Then after settling on a move and their opponent does the same, they do exactly the same thing based on the new game positions. There are far too many variables to do that in Civ, there's fog of war limiting knowledge of what the other player could be doing (and incomplete knowledge of each other player's techs, what cities they already have, etc.) - and in any case you can't play a Civ game thinking what you're going to do a small number of turns ahead because of its scale, you need to think in terms of grand plan.

I don't think that's easy to program - you could do a better job than an AI that can't even work out it's meant to move a settler if someone sticks a city in its way, but you're still going to have all sorts of situations where the AI's subgoal is "get Sistine Chapel", and if it fails in that task it has to find another way of playing that will grant it sufficient culture to make up that loss. It's already been observed that even with Civ IV's AI you could always defeat a cultural victory by rushing the Sistine Chapel because the AI had no alternative route to that victory condition to fall back on.

I definitely agree making a chess level combat AI would be an extremely time consuming job and probably even require need of a program that can learn, but that can not be the only way it can work if you lower your standards a bit.
Planning ahead would be fantastic, but at this point it can't even create an army formation that is not just sending everything to the field of battle based on when they can get there, and with cities being fixed points it could just run a script to wait when in x range from the intended target.

I did actually get a wave attack from Persia in one game recently - a few Warriors and an Archer in the first wave; I beat those and started pushing out on the offensive, only to find another Warrior and a couple more Archers following up (who promptly killed the unit I had pushing out for a counterattack).

I read in a thread somewhere when I started lurking from someone who I then thought was knowledgeable on the subject (criteria where most likely post count and formulation) that he managed to open the file relevant to the AI and that it looked like they, for the most part, copy pasted CIV's AI, which honestly explains a lot.
The blob movement might simply be that it's trying to move the lot in a stack and keeps getting confused why apart from his GG's everything else keeps getting stuck,

After a couple of games where the combat AI seemed moderately capable, I've now had a drawn-out combat in my third Immortal game as Songhai. I had Troyes (again the fixed settler thing, I think - I had a bunch of archers and warriors ready to attack next turn, and he just turned his settler into a city. It held out a few turns, actually), and hadn't yet got Iron Working, so I just had my couple of Warriors and a couple of Archers. I repelled his first attack, with a spearman and a Swordsman, I think. Then my GG turned into a Citadel at my border (incidentally clearing a forest that gave my garrisoned archer a clear line of fire into the bargain), with one of my Warriors inside. The AI did the usual thing of sitting round the Citadel taking damage without bothering to attack, and with the garrison as well he lost a lot of Swordsmen, Spearmen, and the occasional Archer (they were actually doing a better job of hanging back, and he actually used his chariot archer very well, having it placed where it was protected by Paris but could regularly snipe my non-Citadel fortified Warrior). The war lasted long enough for me to both reach Iron Working and develop a small iron source, at which point Napoleon got bored of having me kill all his units and promptly gave me his two remaining (non-capital) cities (somehow, despite having two new puppets, each of which has only one building in, I'm suffering negative finances due to building maintenance, as well as unhappiness because no one has duplicate luxuries to trade).

EDIT: I'll have to check when next in-game, but I think I've worked that puzzle out at least. Orleans and Tours are connected by a road, but this doesn't bring in any income because it's not connected to Gao.

Should probably have razed Tours anyway, if only for the Songhai gold bonus, though at least it has a library and it's well-placed for connection to Timbuktu (sorry, Tomboctou).

I would have carried on the war, only I'd promised Darius to help him beat up Washington in a few turns. San Francisco was protected by a single damaged Warrior last time I checked...

and the diplomatic AI might simply be set for a game that still has positive modifiers like religion that don't exist anymore.

I see a bit more design in it than that, but I agree that it seems to be designed in the context of a system that allows greater player control - the AIs aren't all going to war because it's a wargame or because they're programmed to be especially belligerent, they're doing it because it's the AI's way of prompting the player to pay more attention to his diplomatic relations to avoid that outcome. The only trouble is, especially early in the game when aggression is most pronounced, the player has very little control over said diplomatic relations, partly indeed because of the absence of many positive modifiers.

Oh, and maybe I should have been a bit more clear, I think the taunts are great, I was only using taunts in the context of a first person shooter in which case it is bad but because real leaders can be evil minded bastards I find it rather fitting, in contrast to them telling you that they're going to go to war with you because your winning the game, which is rather illusion breaking, admittedly I do get tired of the taunts in CiV but that is mostly because I find it almost impossible to see them as actual opponents anymore, but in CIV I loved it.

I haven't come across "I'm goin to war with you because you're winning the game" - the closest I've come is "You're very powerful, but I'll find a way to defeat you" (from Gandhi), which means the same thing but I think is fine in terms of the character of the game.

And about the gunships, I've seen England and the Ottomans try to take city states that where already at 0 hp with ships while being on a completely different continent where she had no ground army.

Honestly, I saw this in Civ IV and earlier Civ games a lot as well. And ships have never been able to capture cities - Marines could, but Civ V at least in concept has one up on that in that any ground unit can launch amphibious assaults. It's not the ships that are at fault, which work much as they always did and perhaps for that reason the AI uses them reasonably capably, it's the AI's inability to use the embark/disembark rules, and its reluctance to cross open water with embarked ground units.

Honestly, if they wanted ships to work, instead of nerfing it, couldn't they just have either added a meelee ship or giving ships a secondary meelee attack that might be pretty weak but at least works in situations like this?

I'd rather they just taught the AI to use embarked units to attack across oceans.

And about the GP's, I have seen them make stuff like manufactories before, and as far as golden ages go, the computer can't actually have a golden age for some obscure reason, maybe because they wouldn't understand it? Then again, what is there to understand about more gold and hammers...

That's odd. I was thinking Persia may have been tailor-made for the AI, and the civ does have a reputation for making lots of gold - which would make sense if it were exploiting the AI's happiness advantage to get the most out of the Persian UA.

Still, I don't think I've seen GP tile improvements, and certainly never a Citadel. Even in the Viking scenario (where there must be some slight improvement to the AI in some respects, since at least it can cross water), I've never seen AI Normans use Motte & Bailey, and that doesn't even need a GG to create it.

Do GPs count as units for score purposes? I remain sure that the AI makes a lot of its decisions on the basis of how it impacts its score (which at least is something an AI can readily check and respond to). If sacrificing a GP would reduce its score that would lead to situations in which it would rarely if ever do so.

And the SC2 AI is not that simple actually, it's still not a work of art but its still more advanced then the CiV combat AI (or the other Civ games, as noted before, you don't need much AI for a SoD), the rushing is predetermined yeah, it varies with difficulty though and at least has different builds and a limited understanding of counters, whereas CiV's just goes with whatever it's flavor is, or at least I've only seen 'natural' counters, i.e. Bismarck making panzers against infantry or Alex countering horses with hoplites.

That's not a bad way for it to start - the game does have natural counter divisions which make that sort of play workable by the AI. I usually find that cavalry preferentially attack ranged and siege units, which is good to see. Spearmen and pikes seem more random in whether they prioritise cavalry targets. Although Napoleon didn't much adjust his archer/swordsmen/spearmen mix when it didn't work against my city (perhaps, since I was defending with the same units throughout, he didn't know what to build as a counter - he did seem to try a larger number of archers at some point, but with Citadel fortification bonus and 1-2 ranged defence promotions per defender, that was a bit of a no go).

It might not know race specific strategies but as Blizzard has been a bit more open with the ability to make your own AI, there are already quite a lot available that might very well have this feature (not to mention that the game has 2 expo packs left to go, whereas CiV...) and I know for certain that they made map specific AI as well, so it even knows how to exploit high ground and positioning.

Hmm, I tend to use a single map for all my AI battles in SC so that I'm comparing new strategies across standardised conditions.

And attacking in the flank might not give a +x% bonus, but it does give the ability to engage units that you might be able to counter, with Starcraft 2's heavy counter system this can be highly effective, see: the entire zerg race.

Are you talking just about units surrounding opposing ones? That's a numbers issue rather than anything relevant to flank positioning.

Tactics =/= terrain combat modifiers.

Tactics are anything involving moving and positioning units for advantage - terrain combat modifiers are the way Civ does it, by giving you advantages/disadvantages depending on where your units are sitting. Games like Company of Heroes give cover effects, and benefits for being in buildings. In Total War games you gain advantages for being on a hill, and the ability to hide if you're in woodland (or some other terrain depending on unit rules). Starcraft gives basically nothing. You get an advantage for being at the top of a ramp or, if you're flying, in dead space where you're out of reach of ground units. That's about it.

And the whole idea of wanting it to play like a human is actually kind of hilarious, I'm guessing that after CIV the dev's got feedback that the AI was too easy and predictable, so they thought that by learning 'human' tactics and shoving in a load of random things they could do people would be more challenged.

I've mentioned in past threads that there is a fundamental tension in trying to treat Civ, which was always at least 50% sandbox, as a competitive game. It's not really the nature of Civ games - there's the issue I like to raise that it's hard to derail or interfere with opposing strategies, there's the fact that as you point out the AI has generally been along for the ride, to give the human an enjoyable rather than a challenging experience and without any particular intention to win the game in its own right, and there's the scoreboard with its various levels (and yes, Dan Quayle does look slightly dated now, but it's the traditional joke level). Civ was always about the human playing against himself to beat his previous best, not about playing a strategy game to outwit your opponents. In some ways I wish it was, and if it was the Civ V engine would be a good way to go about developing that kind of game, but the AI is never going to be capable of playing something as complex as the average strategy boardgame, and Civ V has more detail and complexity than that kind of boardgame can offer. If an AI can't be designed that can be a reasonable opponent in Tigris & Euphrates (and they haven't even tried to make a computer version of something as complex as, say, Britannia), trying to do so with Civ V is going to be a very hard sell.

If you would put this diplo in CIV it would be borderline unplayable because in that
game a massed attack from multiple enemies could not just be fended off by some rushbuying and range unit abuse.
So, in an attempt to make the game more difficult they essentially made it hostile, definitely making the diplomacy part harder but also taking away most of the fun.

It still can be fun when the diplomacy works - and it can be made to work. But perhaps not on Immortal; three games so far, three rush attacks (well, four, since I was attacked by two enemies in one game), all before I've had time to secure duplicate resources for trading or funds for research agreements and often not long after the development of Writing. I don't even know what prompted the French attack; Napoleon's war dec sounded as though he planned to wipe out everyone, which was shall we say somewhat optimistic of him. No doubt I shall find out if Darius remains loyal and if the joint attack on the US will earn lasting favour from him (we're into our second declaration of friendship), and whether declaring friendship with Sejong will be lasting (he's supposed to have high loyalty, but this is the first time I've encountered AI Korea).
 
To make it easy for you, OP. Why don't you start the game I've uploaded in the General Discussions area.

You will have the time and everything to do what you ever want to do.

Prince lvl only, it suits you well.
 
I disagree. I think the AI tactics are pretty incomplete. The game was release was rushed (I think) and there were quite a few bugs and such on launch day. Probably AI tactics were cut short. Also they didn't really have data on how people would be playing so they couldn't really develop the AI that easily.

That is great you disagree, but you are simply wrong. Coding an AI to move in formation is extremely hard. We have a bunch of people whining about a lack of a solution to a problem they don't understand.

I might as well make threads asking why Civ5 doesn't give me a massage while I am playing.
 
I read in a thread somewhere when I started lurking from someone who I then thought was knowledgeable on the subject (criteria where most likely post count and formulation) that he managed to open the file relevant to the AI and that it looked like they, for the most part, copy pasted CIV's AI, which honestly explains a lot.

If somebody said that, they were talking out of there arse, because nobody outside of Firaxis has seen what the current AI code looks like.
More likely they were talking about the XML code for the difficulty modifiers. This (as many parts of the XML) indeed seems only a slightly modified version of the modifiers XML code in civ4. This may be somewhat of a red herring, because many of the "leftover" xml values, do not do anything. (Not properly cleaning the XML before release can definitely be seen as laziness on part of the devs.)

(Note, that this also is somewhat a sign of incomplete development/play testing. I clear flaw in the AI modifiers, is that the AI is compensated for its lack of skill by giving it greater numbers. Due to the 1upt nature of civ5, this is a lot less effective than it was in civ4. (Simply, because civ5 combat was designed to favor quality over quantity.) The game would have been better off, if instead the AI was given stronger units as a compensation.)

About the DiploAI:
The diplo AI in civ4 was terrible from an AI point of view. Almost none of the considarations going into the AI diplomatic decisions had anything to with the interests of the AI. Rather they were obsessed with irrelevant issues like what state religion other players had, and what civics they were using. Yet, the result was a behaviour that, while self-destructive, was acceptable to most of the players.

From an AI point of view the civ5 diplo AI is many steps ahead. It at least makes an effort to make decisions based on the actual gameplay impact of those decisions. However, its decision making is still rather bad. (but less terrible). More, problematic, is that these decisions appear to rather erratic to some of the human players, with the AI often ending up disliking the human player. (However, be honest to yourself, wouldn't you dislike yourself, if you were playing against yourself :lol:)

About civ AI vs. RTS (sc) AI.
There are some very different beasts. Some things to keep in mind are the following:

-The AI's of games of like starcraft are almost always purely scripted. That is, they are based on a pre-determined build order/ order of actions. (Usually with branching options based on its opponents actions.) Constructing such an AI to give a decent challenge to new players is relatively simple. However, they are very inflexible and cannot deal with situations which where not fore seen when being programmed.
This makes such an AI completely unsuitable for a game of civ. This is because the AI be can be confronted by a completely different set of xml rules. (For example, you cannot script the AI build order, because some (or all) of the units in the script may not even be available, resulting in a zombie AI.) Instead, the civ AI must always make it decisions by weighing all it available options against each other. This obviously, is a lot more complicated to program, and requires a lot of tweaking of the different weighing factors to get right. (The civ5 AI appears to need quite a bit of work on both aspects.)

-A useful principle in RTS games for the AI is that it can do the following: Make decisions for all units based on the current game state, and then reevaluate those decisions a game cycle later. This (if done properly) allows the units to dynamically make organized decisions. In particular, since the cycles are very short, it is easy to compensate for bad individual decisions.
In a turn based game all the individual moves matter, and making a mistake immediately leads to disadvantage. (Leaving a GG or archery undefended vs melee for just a single turn can be disastrous.) So again this working principle does not work for turn-based games.

In this manner there are many more fundamental differences, which makes the AIs very incomparable.

tl;dr: It is insensible to compare the AI in civ5 and SC2.
 
If somebody said that, they were talking out of there arse, because nobody outside of Firaxis has seen what the current AI code looks like.
...
About the DiploAI:
...
About civ AI vs. RTS (sc) AI.
There are some very different beasts. Some things to keep in mind are the following:
...
tl;dr: It is insensible to compare the AI in civ5 and SC2.

Good post Trias, I think it is hilarious when people laud the Civ4 "Diplo AI" and attack the Civ 5 Diplo AI. CIv 4 didn't have any diplo AI! It was just a naked list of modifiers, that was pretty much it. No mystery, certainly no challenge, and almost no ability to perform on an adequate basis.

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.
 
The discussion about the SC2 AI took write a wrong turn somewhere along the way... Please note that the SC2 AI was brought up in a discussion about AI in general. I never did this with an idea of comparing the AI in the two games.
 
Top Bottom