Sirian's Shadow - Epic Four

[Sirian]

As always, my report can be found at Dawn of Civilization:

http://sirian.warpcore.org/civ4/index.html

As game sponsor, with a customized map, I had spoiler info, so my report for this one is only a shadow. I hope you find it worth reading anyway.


- Sirian

[VoiceOfUnreason]

Do I have your initial tech order right?

Hunting
Archery
Bronzeworking
Mysticism
Animal Husbandry
Wheel
Pottery

[darrelljs]

Sirian,

As always, great report.

I think you are being a bit hard on the AI, but then I don't know what is going on under the hood as well as you do. In my games, I see AIs declaring war on each other all the time (especially the agressive ones). In your game I think Qin didn't attack Gandhi because as the power graph shows he never had an opportunity to. He's ratty, but he's not stupid. Gandhi didn't attack Qin because he didn't really see any gain from it.

So the real question was, why did Gandhi attack you? I can only speculate, but maybe it was economic. That is, as his economy matures the benefits to expansion became the dominant factor in whatever equations dictate his behavior (even more dominant than the peace preference), and you were an easier target than Qin or the barbarians?

Darrell

[Sirian]

Do I have your initial tech order right?

Hunting
Archery
(Sailing from hut)
Bronzeworking
Mysticism
Animal Husbandry
Wheel
Pottery

Hunting
(Sailing from hut)
Archery
Bronzeworking
Masonry
Mysticism
Animal Husbandry
Wheel
Pottery
(Agriculture from hut)
Iron Working
Meditation
Priesthood
Monarchy

Probably not ideal, but not too bad. I ended up not building walls in Rome, so the Masonry could have waited, but so could any of the other improvement techs, since I didn't train a Worker until I already had three cities -- and I did start Walls as my first build in the third city, as well as build them fairly early at the second city (which never came under direct attack, actually, if I remember correctly).


- Sirian

[Sirian]

I think you are being a bit hard on the AI, but then I don't know what is going on under the hood as well as you do.

Being tough but fair on the Civ3 AI is in large part how I got the opportunity to contribute to the Civ4 AI. Most (and I do mean MOST) of the Civ3 AI's tactical weaknesses were improved upon in Civ4. I think to a large extent, the moddability of Civ4 works against it. There is no hardwiring of unit AI, for instance, since any assumptions made about units might be totally off base for any modified units. Thus there is no code to handle swords and axes, but only to handle the "best unit" to fit certain roles. With only a few roles to fill, the AI rolls out less-than-ideal unit mixes. This is a totally new problem for Civ4, introduced by the more complex combat model and the total commitment to moddability.

The harshness arises in large part because I've already said all the nice things about the AI before, in most cases directly to the AI designer-programmer, and I don't tend to repeat old news (old to me). Most of what is being discovered at this point is a better understanding of the various problems and shortfalls. I can't help but continue looking for root causes behind AI behavior, especially undesirable behavior. Often the problems emerge from overlapping functions, where systems intended to do X and Y and up doing Z when they intersect, and sometimes Z is very bad. No way to predict Z in advance, though. Nobody is clairvoyant.

In my games, I see AIs declaring war on each other all the time (especially the agressive ones).

Nothing like they did in Civ3. A couple of war declarations total for the whole game, with a couple of civs involved while the rest turtle and tech away.

Yes, they target weaker civs. That SOUNDS good, on first blush, but it is exactly what creates the imbalance and the predestiny. I would have preferred to see the AI pounce on a human who is weaker than he should be by gambling on running low defenses -- but "weaker than he should be" is not a standard we ended up with. Instead we have "weakest on the board" getting targetted, and on higher difficulties, that is inevitably the player, going by the measures the AI takes. Something intended to do one thing instead does something else, and it's a universal problem.


Just being "at war" is also not the issue. What are the AIs doing with these wars. Why are they entering a war? Why aren't they entering a war? The triggers and methods being employed here, while not without their merits, have not carried us to the promised land. They haven't even gone as far as I/we expected to see them go. This isn't anybody's fault. I'm simply looking toward future possibilities, potential solutions, and goals.

In your game I think Qin didn't attack Gandhi because as the power graph shows he never had an opportunity to. He's ratty, but he's not stupid.

That's a fair point, but again, the AIs are all designed to seek the same level of military, so it's only by accident that they ever end up with enough variance between them to justify a war, if odds of success in a one on one match are the measure of worthiness. That kind of thinking can spot the trees but miss the forest entirely, since it is not the logic of one on one that matters, but the big picture of how all the individual choices mesh in to the world reality. Getting locked in to bilateral-only thinking is a big part of the problem here. The whole (diplomatic behavior) needs to be greater than the sum of its parts (the various bilateral relations and evaluations) -- but it isn't.


As for why Gandhi attacked me... It's all about the attitudes. Get them high enough (the threshold varies per leader), they will leave you out of their targetting routines. Still, they are all willing to attack the player whenever the player is not "off limits" due to high relations, because the player is "too weak" -- an inevitable situation on high difficulty.

Blake's game saw him sign a Defensive Pact -- a great move, since it essentialy merged him with Gandhi in Qin's measurements, shielding him from attack. I didn't manage to last long enough to get the option to sign a DP with Qin, though. Still, in Blake's game, although Qin eventually attacked India, it was not until after the game would have ended by space if that condition had been enabled.

Rat's game saw early war between India and China. I wonder how that got started. That was what was supposed to happen with this scenario, that there would be some early war on the other continent. Could just be that the raging barbs got in a lucky strike, though, and left Qin weakened just enough for Gandhi to decide to attack -- Gandhi the peacemaker. (Heh.)


- Sirian

[ThERat]

wow....reading your conclusion left me somewhat desillusioned

I mean, we had all these discussions and of course, you tried to explain to us what was happening and why decisions were made the way they are. But, now reading from you such an analysis of the AI is making me feel sad.
Sad, as I know you really wanted the best AI to give us all a real great gaming experience. And sad, because I don't feel that optimistic about the game, that I would eventually find my love for it again.

I agree that the initial phase of this game was great, like AW in C3C (do you notice that part of the fun is the absence of WW when fighting barbs?). It started to get tedious when Gandhi appeared, while barbs slowly disappeared. I hope I did not delete my savegames so that I can check what happened between Gandhi and Qin early on. I was friendly with Gandhi from the start to gain luxes etc. That was my luck, the few cities Qin managed to plant on my continent were easily taken by me and Gandhi.

[Kylearan]

Hi,

Sirian, good report as always, and thanks for a really great scenario, even if it didn't work out as you had hoped! You wrote somewhere in your report that fighting the barbs felt like Civ 3 AW, and I had exactly the same feeling. It was great fun, and I really hope we'll see this kind of scenario again. Now that I've read that you actually gave the barbs some settlers to start with, I'm admiring this scenario even more - one of the best I've ever seen. Again, thanks!


That the AIs tend to favor attacking the human player more than other AIs is indeed a design flaw, but I think it's not only an indirect consequence because the human will inevitably end up being the weakest in hard difficulties. There are some direct design decisions that affect it too, namely the fact that the player can (is forced to!) accumulate diplomatic mali like "trading with worst enemy", "refused to give us tribute", "refused to help us" and so on that the AIs will not have. Thus, the AIs will always feel worse about the human than about each other. I consider myself very good at managing diplomatic relations with the AIs, but it annoys me to no end that I have to keep this asymmetry in mind when dealing with the AIs.

You also mentioned that you need a tech lead to be able to bribe other AIs to help you in a war, meaning in a situation when you don't need help in the first place. I'm not sure about that, and in fact had planned to try it out if I had found the time to finish my Epic 4 game. My plan was to shut down research once I had bombers/tanks, and to use massive amounts of money to bribe the AIs. I think that *is* possible. It means you will fall behind in tech even more, but it should be worth it in cases like this. It would have been a end-game, mop-up war which I hate, but still.

-Kylearan

[roland of gilead]

you should be proud of this senario i had never played deity before and decided to give it a go and enjoyed this imensley.

i think your senario worked out perfectly in my game quin and ghandi went to war on and off for centuries in my game. ultimately i was able to befrend both and a stalemate occured where we all became to stron for each other to attack.

just trying to out tech the barbs was a joy to do and a real dificulty.

this senario has really encouraged me and am lookin forward to taking part in some of the adventures and epics coming up.

i think if you had made friends with ghandi from the off you would have found it played more like you thought.

and puting issy and louis behind the mountins was a nice move you have no idea how long i sailed roung that island before i cottond on to what was happening.

i am sure others really enjoyed this game i hope someone won as that will be very interesting to read about.

thanks sirian.

ps i am surprised you never got your prets out earlier my next tech after bronze was iron and my 2nd city went right on top of the iron it allowed me to be more strategic with my city placemet their after.

ie not on the copper so i got the corn as well.

my cith in the south between furs and silver was also very productive and got an academy in the end.

one last thing i poped compas from a hut now i got writing once and thought that was good wonder what the best tech anyone has got from a hut is.

[Jabah]

Wow, Gandhi sneak attacking you, how often does that appear !

Religion probably played a huge role in your game compared to mine.
with Isa/Louis getting all the 7, you had (almost) no way to be 'friend' with Qin and/or Gandhi, while I ha almost the perfect distribution with Gandhi/Qin 'discovering' 2 religions (and sending me missionaries), but both adopted the same official one.
Net result was Isa/Louis were their worst ennemies for obvious religious reasons and I was able to adopt their religion for a BIG bonus AND trade with both. And was later able to sign a Defensive Pact with Gandhi to make sure Qin had no stupid idea. I was pretty safe with Qin being 'only' with Infantry while Gandhi was ModernArmor & MechInfantry (And I was with Rifle/MachineGun).

Jabah

[mostly_harmless]

Hi,
in my game China and India were at war with eachother around 1200AD or so. All religions went to the mountain keep, so no holy war. The outcome of this war back then was crucial for the development of the complete game (see my report, which I will post later today).
In my game I was backward tech wise of course, but Qin would have joined (not started!!!) a war with India for a meagre 1000gold.
Just to add some info to your assessment.