WHEOOHRN Hiding

[scu98rkr]

"I suggest you to look to the SG I'm hosting, where, due to the fact that it wasn't clear who was the target of one of the AI, the team was forced to make a military buildup that proved to be completely unnecessary and that it came in a very bad time ( lib race )."

No this is wrong. If the AI had not of given the warning you would either had to build up your military in case of an unexpected attack. Not having a defense force is always a risky business and your playing with fire. It would be better is the AI actually attacked you when weak than just forcing you to build units.

"Bottom line: ok, I admit that in certain situations the AI will be stupid to say to the world that is preparing a war, especially when it is dead in the water who is the target. But those cases are a minority, so making a case in top of that is a little bit shaky."

I totally disagree with this I would say the opposite is true. In most cases its best for the AI to keep you on your toes. If you know that none of the AI's are planning attack you can neglect military to concentrate on infrastructure and then build more advanced military unit when you ahead of the AI tech wise.

In a minority of cases it would be best for the AI to let you know its going to attack but some of thoose are covered by demanding tribute as Yakk says.

I mean if its better for the AI to warn you of an attack it would be better to have the power bars on show by default rather than related to espionage.

What needs to be improved is that the AI needs to be able to quickly spot when it has a military advantage and to attack. That would force the human to always have a good defense force.

 

[r_rolo1]

@TMIT

I respond to you the same way that I responded to Yakk and others that followed the same line of reasoning : first , "hands full" is not a war declaration sentence even when the AI is left undisturbed ( people seem to forget that ), second you can't assure in 95% or more of the cases of who is the target ( even with all the tile count, worst enemies and power ratings mumbo jumbo ) and third and most important ( it looks I need to bold this ) it does not make the AI better, it simply ties the hands of a player that uses it , and in that sense it is the functional equivalent of other less than pleasant things, like ( one you love ) a quirk that makes the interface assume that you clicked the DOW button , or of playing without a screen or with the hands tied behind your back ... or of other things that were already introduced in game , like the new diplo win rule or the new culture win numbers for No Espionage, that have 0 to do with making the AI better, but that make the life of the player ( human or not ) harder.

In fact I think will use the sentry net example you brought from MP to ( try to ) explain this more clearly ( someone gave a peak on fastmoves site, it looks ). As you know the AI does not make sentry nets, but the human has the ability to do them and to exploit the lack of knowledge of less savy players ( humans or not ) to surprise them via the diagonals. You have two solutions for that: one, make the AI to do their own nets and code them to exploit any human that does not use them in case they decide to go to war, two, forbid or make harder ( via increase of maintenace , for example ) to maintain a safety net and to attack via the diagonals ( say, make the diagonal movements to cost more movement points, as IMHO they should had done in the first place.... but that is besides the point ). Both solve the assymetry between the human and the AI ... but only one makes the AI smarter

On the dogpile comment: that would be a factor IF you could say with 100% certain that you are not a target AND ( not screaming, just highlighting the boolean operators ) you have conditions on backstabbing them effectively in time without putting yourself at risk. That is not exactly the majority of the cases... and in those cases, making the AI that can do that do that would actually be a improvement If you want a RL example, the US is with "hands full" against a certain country in the Middle east for a couple of years until now ... and yet you don't see Canada or Mexico preparing to attack the US Now if ( lets say ) Quenia decided to say out loud that it wanted to invade Burundi for a couple of years, maybe Tanzania would feel more eager to readjust their northern border

@ scu98rkr

First, and try to not take this as a critic , your comments reveal that you normally play games with little AI ( probably standart with 6 of them ), most likely hemispheres or continents and more importantly, in a level you can win with confort in maybe 80-90% of the times. That is most certainly the base of your comment that there is no point in making the AI know the "hands full" of the other AI because the human is the biggest threat of them. The human is rarely a bigger threat in any civ game to a AI than the rest of AI controled civs, atleast if the human is playing their actual level ( say, a level where they have roughly the same chances of winning a game as any other civ controled in the map ) with even a rather conservative number of 17 AI , you will not be a bigger threat in the game than the other AI for any specific AI 95% of the times or more. So , yes, making the AI aware of the possible DOW of the part of the other AI would be a big improvement.

About the human not having denials avaliable... true, but it also has no diplo status The coders decided to be salomonic on that and say to the AI to assume that the human is always cautious towards them ( well, there is no better solution, because not all the players will be either friendly or unfriendly with the AI players and assuming otherwise would be forcing the players to go that way ... ). A similar thing could be done for this ( say, assume that the human has a 50 in the XML for cautious and then factor power and all the other jazz in )... or also to bring back the SMAC reputation index ( that allowed the AI to discern between berserk humans from peacenik ones ... not that the SMAC AI made much use of it, but again that is besides the point )

Oh and the quote you made of me was originally to point that Yakk was saying that it would A in one situation and the negation of A in a similar one....

On the comments on the second post: first I assume that you haven't read the thread I'm refering to , so I should had been clearer. In that game at that time we bordered 2 AI, one of them very pissed with us because of high border strain ( we, due to variant issues, constrained them a lot ) but with other civ as worst enemy by a inch ( war with other AI ongoing ) and the other AI cautious with us. In a matter of some turns both the cautious AI went hands full ( with 2 possible targets with roughly the same chances , one of them being us ) and the pissed AI got a peace deal without having used their army ( they couldn't attack the other AI because of closed borders in between ). So it looked that we were in a serious risk of facing a two front war and we started building army accordingly.... In the end, the pissed AI forgot a demand refusal and got cautious with us and the cautious AI target was the other civ. Note that we had 80-90% of the forces of any of the two AI we are talking about, a more than adequate force to defend yourself of one of them, but obviously not from both. So , yes, that "hands full" made us do unnecessary troop buildup ...

And I disagree on your disagreement of me And I already explained why: "hands full" is not a war declaration and it does not bring a tag of who is the target in the majority of the cases ( in fact, as more AI there is the game , the harder it gets to have certainties of who is the target of a warprep ... I would probably say that 7 AI is probably the upper limit to have certainties on warpreps targets in a usable number of times... if a AI can start war plans at cautious and it is cautious with 10 civs and annoyed with other 3, it is hard to pinpoint with certain who is the unlucky duck )... and this has 0 to do with "keeping the AI on your toes" unless your idea of defense implies having 3 times the power of the strongest AI in game ( to be able to survive a ( highly unlikely ) global dogpile on you ). As in most cases, and unlike you seem to believe, it is not possble to say with certains who is the target of a warprep and if the warprep will actually generate a DOW ( unless , OFC, you play the game with settings and in a way that most ( or the only ) warpreps against you come from refusal of demands with you being their worst enemy of becoming it because of that. That implies little number of AI, you having consistently atleast 1.2 of the power of most AI and trashy relations with most of them with all of them being relatively pleased with eachother.... believe it or not, those aren't the more common situations you can have in a game In fact, they aren't so common at all ). And to add, no, no "hands full" is not the same as no risk of a DOW, given that the AI does not need to go hand full to start a war on their own will ( not mentioning when they enter a war because of the AP or events ) and, even if it technically enters in hands full, it might decide it has already enough attack units and move to attack ASAP ( already seen that happening ) and there you don't see any warning if they can strike you in 1 turn.

I will not discuss turning the power graph visible as it was in warlords Will just comment that power spikes don't mean warprep and warprep does not mean power spikes ...

I couldn't agree more with you last point, but not for the reasons you stated. The AI should leverage any military advantage it has as soon as possible, not to harrass the human, but because it is good for them to do that . If you want a AI devoted to give troubles to the human, there is a nice and simple option called Always War ( or you can ressurect the Vanilla/Warlords Agg AI option if you desire something more soft than always war ( the code is there, just commented out ) )

 

[scu98rkr]

actually I normally play games with about 10 AI, me and one other human player.

I get the impression that most people never play multiplayer or with other humans and dont consider the options from this type of play.

If I play on my own I normally use 10-12 AI I like the game to be changeling, but I dont like the games where the AI sprints ahead then you spend the entire game trying to come back. Hence why I prefer to play with another human player.

I would guess your comments come from playing games at very high difficulty levels probably deity where the humans needs every last trick in the book. Hence why you are so obsessed that you can see WHEHORN.

Personally I dont think we really should be considering this style of play too much as often it revolves round knowing the AI and exploiting trick and loops holes in said AI to gain an advantage. Or if we do consider it basically we should looks for exploits deity level players use and fix them ie WHEHORN !

As far as im aware as the objective of the better AI is too make a more human AI one than can challenge at all stages of the game and is enjoyable for all types of player and one thats seems to make intelligent decisions.

Therefore I think it would be more use considering the ideas of expert multi-player players rather than deity level AI players.

You obviously have you reasons for disagreeing with the majority of posters and it sounds like they wont change.

PS I am in no way an expert multi-player player, but I did have to change my style of play considerably when I started play against other humans.

 

[r_rolo1]

True, I'm more a high level SP player than a MP one ( and even on MP I definitely prefer things like Pitboss to "real time" games that have that annoying tendency of becoming a clickfest ( if OFC ,no one decides to pause the game ad aeternum before that )). That most surely influences my view of things ( how could it be otherwise? ), but, unlike you state, being obsessed with tricks or so is not the reason I am argumenting in here ( I already said atleast 3 times in this thread I do not like how things are now on this regard Oh this make atleast 4 ).

Let me state this again : my point is not the value of the measure, is the simple fact that IMHO this does not make the AI smarter, and , because of that, IMHO is does not belong to this mod. The only reason I am still debating this is because no one convinced me otherwise ( and unlike you seem to be suggesting, I think that previous discussions I had in this forum and others show that I can be convinced that my opinion is wrong ) and in fact, the majority of the posters you mention are argumenting like if I had defended that things should be as they are ( another proof that majority can concur in error, I guess ). The few that actually argumented something against my previously stated postion pulled the "must tie the human to protect the stupid AI" and the "plans of war should be kept in secret always except to our dear friends" arguments. I responded to the first by saying that making the human life more dificult is not making the AI smarter ( unless in comparison as Yakk pointed out, to which I responded that playing without a screen also makes the human life harder, but does not make a iota for the AI smartness ) and to the second by pointing that saying that you are planning a war is not exactly the same as declaring, that, because the curent hover does not permit unless in very specific cases to pinpoint with certain who is the target , a war prep annoucement often ( in not almost always ) can not be used reliably as a early warning, and that a war prep announcement can even be beneficial to the AI that makes it.

The rest of my responses in this thread had been in the line to defend that knowing that the AI is preparing a war is not a exploit, atleast no more than doing sentry nets and attack via the diagonals ... what makes it a exploit is that the human understands it's uses and the AI don't ( hence having two solutions: making the AI understand and use it as the human can do or make human unable ( or force him to pay a steeper price) to use it. For the record, in doubt, my position is always the second on any issue regarding the improvemnt of the AI ). But that is off topic ( and I regret to have written academic essays sized posts just for that in here ... because that is not the issue in here ) .... on the topic, I still have to see someone pull a coherent argument of why my line of reasoning fails ( and no, that does not include "OMG can't you see the exploit?" ) and that this change makes the AI better in any sense besides comparative ( that by it self is meaningless ). If that happens, i'll be happy to drop my hat and recognize I was wrong and thank everyone for the intelectually exciting and knowledge filled discussion we had until now.

On what consists the objective of this mod... oh , that would fill pages and pages. My position had been always of trying to make a AI as close as possible of passing a Civ IV Turing test ( note, this is not the same that making a gamer AI : as ( I'm going to regret quoting him most surely ) troytheface said in one of his lucidity moments, humans can rolepay, humans can be lazy, sleepy, unfocused, distracted, wonderhogging, peaceniks, .... I mean a AI that given the same resources and the same ammount of intel the human has acess ( that is, the UI response to the human as it is now ) gives answers ( plays ) that can't be distinguished of a human. I probably think this a minority response OFC, but I don't see any other position that can be as logically founded and unambiguous as this one ( the enjoyiment and chalenge to the human can lead to pretty nasty things if pulled to the last logical consequences ... )

On MP changing the way that people think on the game: definitely . A lot of SP strategies are no more that ripoffs of the AI or atleast rely extensively on knowing the AI patterns ( a Better AI that passed the Turing test would not be ripped in this way ... a game AI would probably be ), A lot of units that look underpar in SP are good or extremely good in MP ( and the other way around ) ... but that has little to do with the coding of a Better AI by it self IMHO ( OFC lessons like using speed, turn order, numbers, technological gaps , dynamic defense instead of static one ,... would obviously be welcome ) because a lot of MP strategies are focused solely on ripping non-gamer humans off

 

[bestsss]

Make everything returns DENIAL_JOKING you'll be set. Telling that sometime an AI can plan a war depending if there is a weaker civ w/ bad enough attitude (doesn't matter who the civ is) and enough attitude to be bribed.
So giving a free tech and an AI may tell it prepares for war which is just as dumb.

Just make it return the same denial for everything and cut the roleplaying nonsense if you really want better AI.

However even that wont matter since power might be known and the attitude too and can be calculated. In that aspect if there a weaker civ preparing still might be revealed by just not willing to be bribed.

 

[Yakk]

An AI should, 'realism wise', have no control over the player's screen going black.

Realism wise, the AI gets to determine what the AI says in response to a query (like "do you want to go to war").

Saying to the country you are planning to go to war with, whom you wouldn't even consider the request because you hate him that much, against a target that both of you know is so much more powerful that you'd never consider going to war against them even if you loved the country, that your main reason why you aren't interested in the "join me in this war" request is that you are gearing up for war on a 3rd party, is a poor decision on the part of the AI.

Moving the "I'm busy with other wars", in (at least) the case where you aren't actually openly at war, to being after other excuses, mitigates this AI stupidity to a certain extent. It would be better for the AI to learn deception and even fancier diplomatic tricks, but that is hard: mitigating this stupidity is easy.

Although, making the "I am busy right now, you idiot" code run when the AI is openly at war is probably a good idea; when the AI isn't actively at war, the "I am busy with something else" excuse should be deferred until after other plausible excuses (in theory, it would be probably be best for the AI to never say this when prepping for war: but I don't want human players to be frustrated making offers, or building up resources to make an offer, that can never be met).

In anycase, this bikeshed painting discussion has gone on too long. When I next get the chance, I'll toss the patch off at jdog.

 

[r_rolo1]

Make everything returns DENIAL_JOKING you'll be set. Telling that sometime an AI can plan a war depending if there is a weaker civ w/ bad enough attitude (doesn't matter who the civ is) and enough attitude to be bribed.
So giving a free tech and an AI may tell it prepares for war which is just as dumb.

Just make it return the same denial for everything and cut the roleplaying nonsense if you really want better AI.

However even that wont matter since power might be known and the attitude too and can be calculated. In that aspect if there a weaker civ preparing still might be revealed by just not willing to be bribed.

Well, that is definitely the logical consequence of the line of thought of the majority of the posters here... if the AI telling that they are planning a war is stupid, it is always stupid So the AI should never say they are preparing a war. Same applies to the other denials , so the logical outcome is that the AI should awnser always in the same way, if at all, to why it doesn't want to be bribed to a war ( as we are in here, why not simply remove all the hover texts in the diplo window? ). As a hommage to the BtS main coder and proeminent contributor for the Vanilla/Warlords version of Better AI we could even call it DENIAL_ANGRY_KOREAN_MAN

The only thing that I can argue against this line of thought is that the original coders obviously thinked otherwise in terms of the existance of diferent denials And again, IMHO this does not make the AI a picogram smarter

 

[r_rolo1]

I agree that this discussion gone out of the rails and that it looks that will not bring anything constructive from now on in terms of the OP ,unfortunately.

But just a final note:

Moving the "I'm busy with other wars", in (at least) the case where you aren't actually openly at war, to being after other excuses, mitigates this AI stupidity to a certain extent. It would be better for the AI to learn deception and even fancier diplomatic tricks, but that is hard: mitigating this stupidity is easy.

Like I said before, I am not an adept of band-aiding unless there is really no other viable option and the problem is serious ( that is in terms of cosistency and/or performance ). IMHO neither applies here . You think otherwise. Fair enough... we will not agree on this as long as one of us change the thinking basis behind the arguments we put on the table and that probably will not happen so soon.

In anycase, this bikeshed painting discussion has gone on too long. When I next get the chance, I'll toss the patch off at jdog.

Fair enough. He had shown good judgement so far in the BBAI and UP changes.

 

[scu98rkr]

"on the topic, I still have to see someone pull a coherent argument of why my line of reasoning fails "

What do you mean by the above. My argument for getting rid of it is thus :-

But how do you suggest the AI checks if the human is WHEOOHRN. Should the human have to check a WHEOOHRN button 20 turns before it declares ?

If you dont agree with this argument fair enough but I dont see how its not coherent. Ok let me try better.

I think WHEOOHRN should be removed because the AI can not check if the human player is WHEOOHRN therefore it is an unfair advantage to the human player.

Is that more coherent ?

 

[r_rolo1]

It is coherent , but I already said why it "fails" : the original coder already implemented code that makes the AI assume a certain behaviour as the human one in some other areas. So there is no reason to not do that here as well... and if you think that making the AI assume a certain behaviour as the human is a bad idea, you also have to find out other solution for when the AI needs to get the diplo status of the human towards them, atleast if you want to keep coherence.

But I would definitely like to not discuss more off-topic in here, please, before some mod decides to close the thread. Yakk already said he would send the OP code to jdog and let him decide on that anyway....

 

[BobTheTerrible]

Nothing personal, but please, just please, can we get that example out of the discussion? It is pretty marginal even for the human as it is now.

I guess I'm ignorant, but I thought that was the whole point of this thread?

 

[r_rolo1]

No, it wasn't ... if you read the post from where you took that quotation, the awnser is already there.

 

[Fimbulvetr]

(Trying to summarize here)
It was not the whole point but it was a good example of that point. And rolo's answer that is already somewhere out there is that the whole Diplo Screen thing and any most likely unintended information leaks aren't even part of the entity that is the AI, and should therefore not be touched in this mod; and that it doesn't always mean they are preparing for a war against you, since in general the target can't be determined. Unlike rolo, most of us think that it is part of the AI, and that war preparations are not to be made public even if the target is not always clear.

So many posts with so many words, for such a tiny fix

 

[MartinHarper]

A war prep announcement can even be beneficial to the AI that makes it.

Can you give an example of a case where an AI at peace could return DENIAL_ATTITUDE or DENIAL_TOO_MANY_WARS, and it is beneficial for the AI to return DENIAL_TOO_MANY_WARS? I don't believe there are any such cases.

 

[r_rolo1]

@Martin

I already gave one ... twice A real game example to boot See post #42 of this thread, response to scu98rkr, second paragraph.

@Fuyu

The resume is close, but not quite accurate. Most of the posters in this thread simply asked to erase the DENIAL_TOO_MUCH_WARS hover. To say the truth the only ones that actually defended the OP proposal ( with maybe some other twist on top ) were martin ,Yakk and probably more one or two. This thread only got this big because it seems that most of the posters convinced themselfes that I was defending a position diferent of the one you pointed out, unfortunately...

 

[MartinHarper]

I already gave one ... twice A real game example to boot See post #42 of this thread, response to scu98rkr, second paragraph.

Is this is the game? I can't find the particular war plans that you're discussing.
http://forums.civfanatics.com/showthread.php?t=336966

You say that the "hands full" messages made the human player build more troops. How did that troop build up benefit either of the two AIs that gave the "hands full" message? I can see that it made life harder for the human civ. However, these AIs are now more at risk of invasion from the human, and must spend more of their hammers on defence. I don't see how this was beneficial to the AIs in question.

 

[r_rolo1]

The thread is that one ( the situation I am describing happened in this turnset and in the next one, in spite of charlie being preparing a war before that ), but I guess you misunderstood my description of the situation . The two AI were in hands full but one of them was in war with other ( so that one is basically meaningless for the argument... in fact it was the end of that war that putted in the danger of having a two front war, given that they were annoyed with us and had a unspent army )... the other AI was the one that was in hands full in peace and with 2 equally likely targets.

Ok, you have a point that the AI is ina bigger military threat, but you can't say that they haven't made damage to the player to whom they inconsciously bluffed to without moving a finger against them. Military units are worth 0 if they aren't used and they drain both the budget ( units pay maintenance, you know... ) and the hammer poolbase ( hammers spent in units are hammers not spent in city infrastructure ) and the argument line is the same for defensive buildings. Given that the player that feared the DOW did not had plans for a war, those extra units ( that were close of the obseletion time .... ) are a minus. As the player that built the extra units didn't had enough attack force to make a credible attack on the AI that pulled the stunt ( obviously most of the units and defensive improments were made in the expectative of defending , possibly against two foes , remember ... the human equivalent to the AI building a lot of counter and reserve units, but little city attack ones ), the AI had little to fear from a attack from us even if we wanted to attack them. So, compared with not announcing his intentions, where the only benefit they would have would be the one they could get from the war with the other guy, they also made the other civ ( us ) to burden themselfes with extra units we didn't needed, that weren't suited for a attacking spree even if we wanted to attack them and a ( temporary ) disruptions of the previous plans ( there wasn't a saying that stated that the first objective of a war is the disruption of the enemy plans ? )... and that for the price of saying to everyone that it was planning a war instead of only saying that to their BFF ( in that game, none )

Sure, this implied that there was a player that could read between the lines ( the current AI, BBAI included, is too stupid to be bluffed ) and a certain diplomatic conjution. But the fact is that, even if just by luck, that AI pulled a bluff on a human that hurted the human because it announced that it was planning a war to everyone. So, yes, there is atleast a situation where a AI can win something of putting the DENIAL_TOO_MANY_WARS in front of the DENIAL_ATTITUDE

 

[TheMeInTeam]

Haha! The reference to AKM even reaches the civ community. He was not well-regarded in high level play by some due to his insistence on playing variants, but his replays were fun as hell to watch.

Anyway I agree with Rolo that this is not a direct intelligence boost to the AI. It merely removes something that probably shouldn't have been there to begin with. If that is indeed outside the scope of this mod, so be it. However it is hard to make a case that it is ever helpful for the AI to reveal its own war plans, and humans certainly do value WHEOOHRN even if it isn't 100% complete information.

I don't buy the "bluff into wasting hammers" argument. It assumes high-level play and betterAI is trying to make the AI good enough to *not* have to rely on garbage bonuses, or at least fewer. If you're at the same unit cost, bluffing and not putting up troops yourself is going to get you killed. Even moreso for the AI, which still lacks anything resembling elite war tactics. And if that bluff turned out to be a DoW elsewhere, my earlier point of setting the human up for an easy dogpile holds. Sure, if you feed the AI a sick discount this bluff might actually be viable, but THAT is just the type of thing this mod is trying to check so that the AI is challenging without access to it. Just like the deity AI shouldn't archer rush the human on turn 8, high level AIs should not be using an advantage they only have because they suck to bluff the humans.

While the term sentry net was new to me, the concept hasn't been for years (since before civ IV) as scouting is so essential in so many games. Of course, the anti-AI sentry net is basically just parking a unit in a city or two nearest the player and checking for buildup, but that's besides the point...

 

[r_rolo1]

TMIT, that is more a argument againt the "sick discounts" than against my proposal And anyway a AI does not need those to get bigger than other players in the game ( human or not ), right?

 

[jdog5000]

To use a poker analogy, while this change won't affect the AI's strategy for playing its hand, it will improve its ability to win by keeping a better poker face. This change will make the game a little more challenging for the human player, while making the AI behave more realistically whether you like to think of them as historical figures or stand-ins for human players of the game.

In fact, the easy availability of this information is currently a bit of a cheat for the human player. The AIs do not use war plan information for their neighbors to decide to bulk up defenses even though there are many places where they could. To a certain extent it wouldn't make sense for them to do so, because the most dangerous player on the map (most of the time) is the human player for whom they get no such information. My sense is that the level to which this information is used and studied by human players is a bit of an exploit that was not intended by Firaxis.