Overall Discussion - Epic Four

[Zeviz]

After playing around with my saves in Worldbuilder, it looks like Barbs have to have access to resources to build corresponding units. In my game, Horses were right next to one of early Barb cities (the one on the island east of my continent), so they started building HAs as soon as they got HBR. However, none of their cities had metal until mid-fourteenth century. So I saw a lot of Horse Archers and Longbowmen, but no axemen, because by the time barbs got Iron (after a 3rd border expansion in one of their smaller cities) they were able to train maces.

It also seems that Barbs stop spawning when they have enough cities in the area. After the initial wave was over, I felt much less preassure than in any of my trial starts, and when attacks picked up later, all the units had 1 promotion, so they were trained, rather than spawned.

So the randomness of whether barb cities poped up near metals, and/or horses, combined with the fact that most barb units were trained rather than spawned, probably was the reason for difference of unit types Barbs used in different games.

PS Barbs got Sailing early and had several ports, so just pillaging their roads wouldn't be enough to disconnect resources unless you were luckly enough to fully disconnect an inland source.

[ThERat]

But I think there are two important things that hugely affect the game.
1.) The founding of a religion in the free world
and somehow loosely connected to that
2.) Was there an India-China war at some point, hence leaving one AI weak and one AI to grow into a modern time monster.

I agree with that.
For me Spain got all religions and more importantly India grew into a huge monster gobbling up their whole continent.
I also never got a tech from a hut, all i got was hostile fellows

As we can see the outcome of the game is really so different.

[Iustus]

Gandhi formed Taoism in my game, so, Qin, Gandhi and I were all Taoist.

France got some ships outside the wall (see the other thread on how to get inside the wall), and proceeded to make himself Gandhi's worst enemy.

So, while Qin and Gandhi did not like each other, in my game they never fought until I paid Qin about 25k to attack Gandhi (who was way above him in the power graph), and then I turned around and eliminated Qin.

I actually thought that it was almost impossible to lose (assuming you were willing to play a long game) if you made it to the point that there were no more barbarians, because Gandhi and Qin seemed so pacifist. I guess the shared religion was doing more help than I thought.

In retrospect, I think I just did not realize how much I was helped by France getting some ships out, and becoming the worst enemy of Gandhi.

-I

[uberfish]

Wow, a lot of really unpredictable things happened this game and it's made for some entertaining reading. Despite Sirian's misgivings I think it was a well designed scenario and worked for most of us, although I feel bad for the few players who got screwed by AI-related bad luck. Notably TheRat and rho21 having Qin or Gandhi totally eliminate the other before 1900, or Gandhi declaring war early on Sirian (ouch.)

I'd like to comment on the AI issues. AI-AI wars caused by border/religious tensions or declared to grab land are a regular feature and actually increase in frequency as the difficulty level goes up. Often this allows the human player to slip under the radar with zero defence for large parts of the game due to the entrenched hate that results. The flip side is that on Deity the AI attacking forces are large enough to overwhelm opponent AI static defence. Whereas at monarch AI-AI wars tend to be pointless and inconclusive, at deity they very often end with the loser being wiped out. I have lost games to runaway conqueror AIs and some people did in epic4. This problem could be addressed by putting the AIs on separate islands, maybe connected by an Archipelago style string of galley-accessible islands to allow positive and negative diplomatic relationships to be established. I think the Gandhi sneak attack problem was really a result of having a no-religion world and too few players which the civ4 diplomatic model really seems to rely on, so that could be avoided too.

The barbarians were also pretty erratic this game it seems, and I think it's probably to do with what resources they had available on their trade network too. Maybe in future scenarios, giving them horses and iron on a distant island connected by coast to the mainland would make it more likely that everyone faced a similar barbarian unit progression?

[Blake]

I didn't mention my hut results because they were universally negative or at best neutral. I popped 2 scouts (both died without doing anything) and I believe EVERY other hut was useless (map of nowhere) or barb warriors... I thought of huts as "+3 exp"... but some people seemed to have lucked out and popped half a dozen techs.

I found the Qin/Gandhi interactions over the various games really interesting... most amazing the games where Gandhi killed Qin - because in my game Gandhi had nearly no strategic resources! So in other games either Gandhi's expansion gave him more strategics, or he killed Qin despite the lack of strategics...


I don't think that a super-AI would be particularly harmful to a player unless it actually achieves domination...

1) With warring and a big/small configuration tech pace would probably be slower, as long as they're buddy-buddy they'd trade techs. The slower tech would slow down the barbarians (I believe).
2) Between being busy fighting and being slower to astronomy they'd take longer to colonize the Roman continent, giving more time to secure it peacefully.
3) By befriending the big guy you'd be extremely secure... the weaker AI could be quite easily killed off. Killing the big one would only be a matter of time - configuring roman territory for maximum hammers and building up a giant force then letting the AI embarass itself.


There were some real wildcards too... religion, from being founded by Qin/Gandhi, to even being founded by the player (And with Deity Isabella and Gandhi in the game?!!!!).
And the possibility of an inmate AI escaping containment and opening up tech trading much earlier.
And just generally time of contact and scope for trade - I was astounded to see a Gandhi contact with Gandhi still not having Code of Laws.


Regarding AI warfare and choice of targets...
I tend to use the rule of "Pacifist AI's are not doormats" - you only need slight positives to be safe from attack - but if you totally neglect your relationships and try to RELY on their pacifism, they'll beat the snot out of you (one of the meanest early Axe stacks I've ever faced was from Asoka - I thought I could culturally oppress him and he'd take it lying down, I was DEAD WRONG). It seems that only Sirian, with his preconceptions of how it was meant to play out, got bitten by this - while other players built up somewhat friendly relationships with both the AI's.
I would guess that in some games for some reason one of Gandhi/Qin quite early on decided he didn't like the other and thus went into military buildup while the other remained in blissful pacifist mode, this would be how one of them could achieve a landslide military victory (in short, dice rolls determining if one of them enters military buildup invasion mode). I think it'd be unlikely for a paranoid warmonger AI to be able to kill another paranoid warmonger AI... because they'd both build up a giant military just as a matter of principle.

I think I probably got the challenge as it was meant to be... stiff resistance from the barbs, religion completely contained, two roughly equal AI's who have chilly relations with each other but only fight later in the game, but still I found it fascinating how much the "alternate realities" diverged.

[Sullla]

Since it looks like this is the main discussion thread, let me paste some of the comments I had after reading Sirian's conclusion:

Replying to some of the comments Sirian made at the end of his game... I agree that some of the things we worked on didn't come out the way we intended with Civ4 (and of course I can't talk about a number of those issues here either). I share your feeling that tying maintenance costs to difficulty level was a big mistake; it's bad enough that the AIs expand faster, techs cost more to research, and the health/happy limits are so much lower. Crushing the human with ridiculous economic penalties as well is just overkill, and it forces games to play out the same way too often on Emperor+ difficulties (as Kylearan and others have testified). It's also a big mistake to tie SO much to the human in terms of diplomacy; the AIs just do NOT fight amongst themselves enough, and virtually everything is driven by their relationships with the human. Something like 75% or more of all wars in Civ4 involve the human, and that's just poor design. If I were to continue to work on the Civ4 AI, I'd probably start somewhere in this area.

But the strategic AI in Civ4 is not a total failure, as Sirian proclaimed in his conclusion. The AIs are NOT all the same - there are very real differences there! Isabella plays differently from Gandhi who plays differently from Montezuma who plays differently from Mansa Musa. Even diplomatically, "the things that count", they are not all the same. Certain civs love to go to war (Cathy), others are almost impossible to bribe into a war (Gandhi). Mansa Musa loves to trade techs, Tokugawa will never give you anything. Yes, there are some repetitive and irritating features about the AI civs, no doubt about it, but I do not feel that the AI has worn itself threadbare. If nothing else, try turning on the "Aggressive AI" feature. That tends to shake things up and make the AI less inclined to sit back and do nothing but tech its way into space.

If there was a flaw in this scenario design (and it seemed to work for just about everyone except Sirian - he has the worst luck in these kind of games!), it was picking Qin as the other AI versus Gandhi. Qin is one of the more peaceful AIs in the game! If you wanted the two of them to be at each other's throats, why not pick Alexander, or Temujin, or even someone like Cathy or Louis? I think that a wildly improbable war declaration from Gandhi in Sirian's game is not enough to write off the AI for good.

[Iustus]

Did anyone else build the internet? I noticed that no AI had built it in 1964, and managed to tech straight to it and finish it in 2001, gaining the 9 or so missing techs.

-Iustus

[mostly_harmless]

Did anyone else build the internet? I noticed that no AI had built it in 1964, and managed to tech straight to it and finish it in 2001, gaining the 9 or so missing techs.

-Iustus

Roughly the same for me although I finished it in 1943 getting 10 techs.
I never built it before I think and it occured to me too late, otherwise I would have beelined more towards it.

[uberfish]

I built the internet in 1902. I thought I would be in a race with Isabella and Louis, they had advanced technology in my game and I was never able to trade with them. Maybe the AI isn't programmed to build it.

In most Immortal and Deity games, I find that the AIs do frequently start wars between themselves without any prompting (probably due to quicker expansion and religious founding bringing them into conflict) and effectively capture cities (well, 20+ attackers will just get through by brute force.) I agree it would be nice to see more, and more effective AI-AI wars on the lower difficulty levels, but as it stands currently more wars would just result in the AI crippling itself through WW which isn't really what anyone wants.

I'm glad Alex wasn't one of the AIs on the other continent. I suspect most of us would have been killed off by his Infantry in 1700 if he had been. Not all AI leaders habitually prey on the weak contrary to what Sirian wrote - but Alex is #1 of those who do.

[Sirian]

But the strategic AI in Civ4 is not a total failure, as Sirian proclaimed in his conclusion. The AIs are NOT all the same - there are very real differences there!

Real differences, yes, but tactical in nature.

Isabella craves religion, but what does she do with it? Does she ever try to translate her strength in to a cultural or diplomatic victory? Does she try to gather allies who have been converted to her chosen state religion?

Montezuma will declare war and try to take cities, but does he ever threaten to reach domination or conquest? What use are his efforts? In the end, he does the same thing all the other AIs do: he techs along. He's not very good at that, though. The fact that the AI (any AI) won't stop pillaging when on the attack means that any lands they do gain are stripped bare or close to it, and of what use is that? He just falls behind and then gets squashed by a tech leader, in most cases, unless he got the juiciest lands to start or the religion happens to break his way, bringing him allies. And still, to what end? Only to hurt other AIs, which at best results in him actually being the tech leader.

All roads lead to Rome in this game, with "Rome" being the name of their interstellar colonizer.

Even diplomatically, "the things that count", they are not all the same. Certain civs love to go to war (Cathy), others are almost impossible to bribe into a war (Gandhi). Mansa Musa loves to trade techs, Tokugawa will never give you anything.

Quirks. The quirks are good. They provide some variance. That variance is far more cosmetic than you credit, though. The AIs, as a body, do not threaten one another, offer tribute, or form any kind of coherent military bloc. There is no multilateral dimension to anything they do.

Diplomatic effects take time to accumulate and even more time to wear off. If you aim to befriend an AI and they backstab you before you've had enough time to bring relations with them in to the "safe" area, your diplomatic game is over. You've incurred lasting penalties with your betrayor's enemies, so you can't switch sides. You've lost most of your progress with this civ. You're lost, friend. Maybe you can prevail on the battlefield, but the Big Huge Dice Roll that is the diplomatic game has broken against you and you're toast at the discussion table. What good do any of these personality quirks do for you, then?

No, the diplo system is too cumbersome. There is too much mass to it, too much inertia. It must be lighter to perform better, even though in many ways that would open it to exploitation. The player has to be able to recover from making a mistake or it's no fun. That's probably THE signature element in Sid Meier's designs. From Covert Action to Pirates to Railroad Tycoon, the games were balanced to enable the player to recover from even significant mistakes. Civ4 has that in most areas, but not in the diplomacy.

Yes, there are some repetitive and irritating features about the AI civs, no doubt about it, but I do not feel that the AI has worn itself threadbare. If nothing else, try turning on the "Aggressive AI" feature.

One hears this suggestion often. It's become the stock rebuttal to any criticism of the core AI. Yet increasing the number of war declarations changes nothing. Even if we removed the bias against the player from the AI, and had AIs regularly fighting other AIs, it would be in vain. The game's core economics shortchange war as a strategy. All it's good for is slowing down rivals. The AIs, all having the same warmaking strategy, will merely pillage one another's lands, maybe capturing a stripped-down city or two, with those who manage to stay out of the warring sneaking past on the quest for space. A "runaway AI" is not possible in the early game, and by the time the tech opens the economic spigot, we're in to the Rifle/Infantry era. Even if an AI could manage it (and they can't, because they are inept at taking cities from technologically competitive rivals), there isn't time.

Worse, if we turn off space, the AI is completely lost. That's the only objective it pursues with intent. Even though it only suffers something like 10% of the War Weariness of humans, that's still enough to force it to stop.

Epic Four stacks the deck in favor of the AI gaining domination victory. The AI only has to overcome ONE other AI plus the hapless barbarians and the severely hogtied player. We did manage to see one or two such victories, but these are the first AI domination victories I've ever seen.

I think that a wildly improbable war declaration from Gandhi in Sirian's game is not enough to write off the AI for good.

That's NOT a wildly improbable war declaration, though. That's the point.

The AIs, all of the AIs, are set to attack weak targets with whom they do not have good relations. Gandhi is the easiest to befriend because his "never attack" threshold is set the lowest. Just get him to Pleased and you're permanently safe. BUT, fail to get him to Pleased, and you're a target. He does not actually PLAY with a peacenik strategy. It's the same bloodthirsty warmaking routine under the hood, the same strategic AI. That AI is constrained by not wanting to attack strong targets (other AIs), and not attacking civs it likes, but if you aren't on either of those lists, he WILL BE coming for you. Period. ... That's not a builder strategy or a leader who is peaceful at heart. It's a warmonger who only picks on those he doesn't like, and only if he can bully them around at ease -- no different than any of the other AIs, except for where his triggers are set.


I think Soren has made huge strides with many of his design elements. He has managed to contain the snowball effect. That's a critical development, one worthy of the highest praise. The problem is that it is not enough. The core design has shifted from one in which endless cities are the ticket (the more the merrier) to one in which all civs must tech up up up. One path dictates that war must be the most effective option, to acquire as much territory as possible as quickly as possible. The other dictates that obtaining more territory is HARMFUL to your cause until you've teched up first, and by then, warmaking reaches the tedious mop-up stage, since it's a foregone conclusion from a technologically superior position, but not enough time left on the clock (for the AI) when facing well-prepared rivals.

The design, in effect, has slain the giant. Conquest is dead. Long live the space race!

We need a game where both paths are roughly equally viable. We need some runaway AIs in terms of land, but for that extra land NOT to translate in to a runaway tech lead. We need a game whose core economy offers not one path or the other, but both, without allowing them to combine. Such a game is years away at best, I'm afraid.


- Sirian