@VDNKh about those savegames:
Game I'm playing right now, BC-0150 is me right about to capture Angkor Thom, AI refuses to negotiate. [...]
The two where AI Suryavarman said that no peace deal is possible:
In 150 BC, where you had just moved a large stack next to Angkor Thom, there were three problems:
- The AI didn't make a peace offer although it was willing to accept your city of York for peace.
- When updating the safety status of Angkor Thom, all of Sury's units in a 9x9 square centered at the city were counted as potential defenders because I hadn't set a flag for checking their movement speed; only units able to reach the city in a single turn are supposed to count.
- City safety wasn't updated for Angkor Thom when your units moved next to it. That was by design, i.e. my idea was that the AI doesn't need to recognize a threatened city until the invaders are ready to move in. However, it appears that it's actually easy enough to update AI city safety in response to human moves.
With those problems addressed, Sury offers Polytheism for peace. (He still has the upper hand strategically, especially in term of production capacity.)
In 25 BC, you've just moved next to the Khmer holy city. Even with city safety being updated, Sury doesn't assume that the city will fall to your attack, and I don't think it's a safe bet either. He'll finish a Horse Archer at the end of the turn. I did conquer it in two tests, but each time with heavy losses. However, since Sury has suffered some defeats already and the power ratio has evened up, he's willing to give the equivalent of ca. 400 beakers for peace, which is almost the maximum that he'd ever be willing to pay given his current economic output. Two improvements that I'd like to make, but I don't think I'll find the time:
- The AI should not just care about cities that are all but certain to fall. There are already safety states "perfectly safe", "safe", "threatened" and "evacuating". "threatened" just means that the city isn't safe enough as a harbor for non-land units and civilians, so another state, say, "toss-up" would have to be added.
- The limit for what the AI is willing to pay for peace should be increased when the tactical situation (clearly) favors the other side. Normally, it doesn't make sense to pay much for 10 turns of peace, especially not to a human player, but it does make sense to pay extra if it gets the AI out of a bad spot.
For the time being the basic approach remains that the AI can't really evaluate the tactical situation and therefore errs on the side of letting combat resolve before negotiating peace. "If the human enemy could take our holy city at an acceptable cost, they'd probably just do so and wouldn't negotiate peace." (Yep, this reasoning is flawed.)
Then the AD 1450 save – that was mostly about a single AI civ having gotten far ahead of the others. That would seem especially difficult to avoid on an overcrowded map like this. I've counted just 28 cities among 7 players. A player that secures 5 cities at the expense of a neighbor who gets just 3 could already be headed for dominating the game.
can you direct me to that tweak? you know me - i like aggressive ai
@keldath: Sure:
These two commits. In one game, I saw an AI civ with 7 cities assume that all but 5 units would be available for an offensive.
I don't remember where the tech diffusion discussion was. Was your intention to just keep raising KTB, or actually give beakers to techs you aren't researching? If it's the latter, please don't ever give the last beaker. If I'm avoiding Masonry for some strange bulb path, I don't want to get it accidentally because the other 17 civs have it.
Noted. Not sure if that aspect has been discussed; I think my best bet was free beakers but at a diminishing rate as the tech progress approaches some percentage upper bound. So no forced discoveries.
Thanks for the quick feedback. Will be sure to try the new dll.
At your own peril.
Maybe it helps if you go into more detail on the design goals? It seems like the main effect of having to keep units around to prevent revolts in conquered cities is to slow down conquests overall, and to prevent a steamroll effect where the civ with superior army strength just takes their stack from target to target.
I guess it's the "big war" pattern of play that I'd like to discourage to an extent, i.e. players producing hardly any military for a long time (meaning also that opportunities arising from wars between AI civs aren't relevant) while bee-lining to some military tech and then immediately going all-in against whatever opponent(s). No doubt there are numerous ways to discourage this. AdvCiv for example makes it harder to keep AI opponents pacified while working toward a tech lead and directly nerfs a (somewhat arbitrary) subset of problematic units. So the rule changes to revolts are intended as one of several measures. I like them for their simplicity: being outside the culture distance of foreign cities does
not prevent revolts and defeated players culture is
not disregarded – it's simpler than in BtS really. (Making revolts more relevant admittedly draws attention to a set of rules that isn't particularly simple.) Keeping dead culture around is also important to me for narrative reasons.
[...] A lot of the problem seems to come from how unit strength scales with culture over time. It feels like this is backloaded, where unit strength easily outpaces culture late in the game. Is it part of the design that occupation becomes easier late in the game? Intuitively, that doesn't seem to make too much sense. Maybe it's my playstyle, but the unstoppable conquest train gets going more easily in the Rifle era and later and is actually not that much of a problem in early wars where strengths are often more evenly matched. So if anything, the pushback should be stronger in the lategame.
Yes, perhaps too much so, though "once you get to rifling or so, it magically all goes away" (Elkad) is a ... pointed way of putting it. At any rate, there's is an era factor in the culture strength formula which could be given more impact. My aim has been to make revolts gradually less of an issue because:
* Managing revolts becomes increasingly fiddly as the game progresses.
* Army sizes and unit strength values eventually grow faster than city populations, and it would feel strange to hold down some size-10 city with a stack of 10 Infantry toward the end of the game. Brings to mind drewisfat's K-Mod Deity game where a size-15 capital in AD 1570 had 7% revolt chance against 30 occupying Cuirassiers (
link).
* Quick expansion should be less of a problem in the late game. That's not to say that it can't be a problem too, but I don't think revolts are suitable as the primary solution for it. A player that tries to catch up from behind in, let's say, the Industrial era actually needs to be able to take a lot of territory quickly because there isn't much time left for it to amortize. Also, when one player dwarfs the others militarily in the late game, I wouldn't want to slow down the march toward Domination.
Rifleman isn't quite a late-game unit in my mind. I like to think that the revolt rule change does slow them down – although drafted Riflemen are very efficient at suppressing revolts.
If early conquest is too powerful because it creates a runaway effect, I'd rather tune this on the economic end, i.e. upkeep costs, if necessary. Maybe it also makes sense to limit razing in some way. If you actually have to conquer instead of razing the neighboring early war target's capital, you actually incur long term costs, instead of just knocking out a rival in the land rush.
There is an economic aspect to revolts. Cities can only flip when in the culture range of the revolt player (and even then, the city flips only on the third revolt), so keeping cities at a ~5% revolt chance isn't unsustainable; statistically, it'll lead to fewer yields. I guess it's not ideal that commerce is usually the least affected yield (since cities in disorder cost no maintenance), and losing production yields hampers efforts to construct culture buildings.
[...] It might make sense to toy with the calculations in a way that makes it easier to lower the required occupation power by just accruing a little culture of your own in the conquered city, even if it is still outmatched by the original owner's culture. This still wouldn't be quick, since the city first has to come out of unrest and needs to construct culture buildings, but at least would give an avenue to do something about the problem besides just waiting.
Yes, that's generally a problem with the culture system. It's a long duration from starting to construct a Theater to catching up with another player's city or tile culture. Spreading the state religion (if any and if not already present ...) is pretty effective at lowering the revolt odds. I'd like to give Jail a revolt suppression ability; that's among the many balance changes that I'm holding off on. Factoring building defense into culture garrison strength
has been a consideration. One would think that spies could do something against revolts.
The culture slider should actually already be effective, apart from helping against unhappiness from foreign culture.
The short answer is that I thought base K-Mod had the balance between rewarding war vs peace about right. AdvCiv has been jarring to me because 1. early conquest has been made much harder, both because of the slight reduction in strategic resources near the capital and the culture issues and 2. peaceful growth seems more powerful.
I guess there are indeed fewer strategic resources near capitals. Unrevealed resources are disregarded by AdvCiv when evaluating city sites. Maybe that's something I should reconsider. I do dislike how BtS makes players uneasy about moving their Settler when a starting location looks poor and may therefore have a hidden Copper/Iron/Horse resource.
I rather doubt that the revolt rule changes are the primary issue when it comes to breaking out of a box. Usually, that's just a matter of conquering a couple of cities. Which can be tough (or even hopeless) when the only available target is a tough one and the available units mediocre, but having units tied down after the war doesn't seem like such an impediment. Well, as with most balance and AI problems, I think it would help to analyze specific situations. Obviously, I'll try to keep the issues and ideas raised here (thanks!) so far in mind when I finally get around to play-testing. In any case, as for intentions, I don't mean to hamstring players in a difficult starting position, and any kind of early rush should be effective under the right circumstances, for all I care even the odd Warrior rush.