Archipelago AI Expansion

[donZappo]

Can you guys comment on how effectively the AI is at expanding on Archipelago / Small Continent maps using GEM?

As a bit of background, these land masses are my favorite to play on. I've been using the CivUP but not the GEM as I've just found time to get into Civ V (after being a grizzled vet of all prior Civ games) and was concerned that some of the changes in GEM were too far from the standard game. I was an active member of the Apolyton University back when that was still a thing, so I like the tweaking mods to make everything more useful but GEM seemed a much larger diversion from what I was comfortable with. I planned on eventually transitioning to GEM but I wanted to get a number of un-modded games under my belt just to have a valid frame of comparison to know if GEM is for me for the long-term.

Anyways, with that being said with my games on small continent and archipelago maps (King difficulty recently bumped up to Emperor difficulty) these games have been a cake-walk. I am definitely an above-average civ player, but by no means am I as good as these games are indicating. Doing some google searches I found some posts from 2010 saying how the AI is terrible on these maps but I don't know if Firaxis did anything to try to address this over the past three years to make it more challenging. I see some tweaks were done in GEM to give more things like triremes (and free embarkment?) to civs that appear on the coast. Combined with the AI emphasis on settlers that appears to be perpetually ongoing, it seems like these factors may address the expansion issues...

Does anybody play on these map types and notice the AI actually able to compete? If so, that may be the defining factor into initiating me into the GEM club from this point forward. I just don't want to have to resort to playing all of my games on pangaea maps... bleck. Also, in case it matters, I play all of my games on epic speed.

Thanks for any comments on this! You guys can definitely save me some time running some test games as experimentation.

 

[mystikx21]

Some other points in their favor in GEM
1) Islands are generally improved city sites which helps the AI economy a bit.
2) Naval units are more important, more powerful, and more likely to be built by the AI, which allows it to fight a little more effectively for the islands. AI combat isn't great still, but it's usually better if it has more units floating around that you must kill to do anything or protect your empire.

The other stuff, like improving their gold buying AI or improving their settler AI/push help a bit too.

I'd say it should do better on archipelago as a result, but it won't be suddenly awesome either in my experience. It's still a work in progress to balance some of the issues with the economy and map types like archipelago can skew the results a bit.

Civ5 in general seems to be about 1-2 difficulty levels easier than previous versions. GEM has in the past been about 1 difficulty level back down but I'd say hasn't quite gotten back there yet since the GK expansion.

 

[Ahriman]

I find the AI unable to compete in the latest versions of GEM on any map type - I think this is largely due to removing the AI's bonus happiness. Without this (and maybe because of the super-expensive Arena and circus?) the AI seems unable to expand effectively.

I find in general that the AI performs better on land maps than it does on island maps though, it still isn't very effective at naval warfare , on settling on other land masses or on managing inter-island invasions.

 

[donZappo]

I find the AI unable to compete in the latest versions of GEM on any map type - I think this is largely due to removing the AI's bonus happiness.

Is that with the Steam Workshop GEM or with the beta GEM? I haven't found a changelog or updated list for the beta GEM so it's hard for me to tell which persistent problems have been fixed. Overall looking at these forums there appears to be a lot of issues with both the stock and beta GEM versions and it makes me leery to check them out.

Do you guys find GEM playable in its current state, or are you optimistically awaiting this upcoming "fix" to get GEM back to its proper state?

 

[mystikx21]

The beta GEM made some significant changes to AI happiness. The buildings changes (in cost) would be in both.

I think the stable version is playable, but lacks some fixes from the beta and new changes to leaders. The beta isn't because of the happiness changes.

I think the question is whether the AI does better than in vanilla. Which I think it did but whether it does "well" is not necessarily the case, yes.

 

[Ahriman]

I think the stable version is playable, but lacks some fixes from the beta and new changes to leaders. The beta isn't because of the happiness changes.

Agreed.

I think the question is whether the AI does better than in vanilla.

It's been a long time since I've played much vanilla, but I'm guessing that the answer is no. It turns out that the AI was very reliant on that free happiness.
I'm not sure what the best fix is. I don't like how it floated huge amounts of happiness before, but perhaps there is some middle ground change we can make that gives the AI a smaller amount of free/bonus happiness or reduced unhappiness per pop that is scaled by difficulty level, so it still gets no bonus on lower difficulty levels but still gets significant bonuses on ~King and above.
I might also be wrong that happiness is the main problem (they might be using too many specialists, not working enough food tiles, not using enough farms to grow their cities, not building enough arenas or aqueducts - which are also now very expensive to build), but that seems to me to be what is limiting AI expansion.

I wonder if all the social policies that give happiness are also a problem: the AI doesn't know it needs to prioritize them, and then it might not recognize that it should concentrate on constructing buildings that now give happiness because of social policies.
The AI also isn't nearly as good as the human at using religion for happiness.

 

[mystikx21]

I think in the non-beta stable version it still gets a lot of free happiness, where it should be doing better still while it has been crippled a lot in the beta right now. (I did say it "did" ).

I think some AI happiness bonuses (per difficulty level) are appropriate but I'd prefer not to see it getting 40-50 in the classical age without something really strange going on.
I suspect it's more early per city problems than high growth that's causing it most of the problems if it tries to expand it cripples itself and if it just grows up they stagnate. A combination of both as bonuses would help but not very much. I don't know if it's possible but giving it a +.5-1 per luxury for free bonus might also help if per city or per population bonuses are hard to balance without changing them too much (it should be just somewhere in between what it got before and nothing).
I think adjusting the wonder/building cost balance, and in particular lowering some of the happiness building costs (at least in production time/rush buy cost) would help the AI also. My economy mod-mod changes I'm pretty sure was lower on most of these costs at least already but I haven't tried it with the beta much yet to see if the combination of changes could help at all.

I'd assume it can prioritize policies/beliefs for happiness, but perhaps it does so insufficiently yes.

 

[Ahriman]

Agree with all this, except that I thought the AI picked policies and beliefs based on flavor ratings, not on their actual needs? I'm not sure about that.

 

[Shadzy19]

Both versions have their problems.

In the latest stable the AI civs still have the extreme amount of happiness (tall empires easily get 60-90 happiness and wide 15+ city's with 30+ happiness left on immortal) on top of that there are anoying bugs (for me at least, not sure if everyone gets them) with both science and faith.

There seems like something hidden is given the AI's free techs and free faith or something, in my last game (immortal / epic / cont plus) me and 3 civs have about 300-400 science, yet a civ with 60 science income is tech leader (35 techs into the game)
All the civs never get more then 2-3 techs behind the tech leader no matter what the science incomes are (got the same results in 10~ games)

With faith its the same, if u get a faith ruin u can get a pantheon first, otherwise u always get 6th or 7th pantheon, every time u get 2/3 turns of founding a pantheon someone else gets it.
Founding religions is very hard aswell, once the first is found u can count on it that the next religions will continue to be found every 3-5 turns.

I tried a game going completly on faith, popped a faith ruin to get the ancient / classical wonder bonus (only gives 15% in stead of 25% though?) I got the first great prophet and got the faith for world worlders (had 6 wonders alrdy) 3 turns later 5!! religions are founded at the same turn.
Long story short i had 30~ faith a turn after, 20 turns later i enhanced mine and indeed 2 turns later 5!! civs enhanced theirs aswell...

U can still play a fun game in the latest stable as long as u dont play a civ that focuses on science, they will keep up no matter what u do and if u can live with the weird faith problems.
For the rest this version is great and very challenging and the AI is far better, dunno if its much better on archi maps though.


For the beta i only played beta 2, the problem there is the AI really isnt competive, I tried 5 / 6 games on immortal / deity and by rennaisance im always 3x better in everything.
They get a strong army but their economy is horrible.
U might be able to get more competition if u want by altering some stuff in the handicaps (I havent tried)
The science and faith bugs are not here, some ppl have some problems with the city screen info not showing up, i dont have this bug.
There is a new bug though (again for me,not sure if everyone gets them), your happiness isnt counted as it should be and specially with mercantile city-states it become a pain in the ass.
It happens in every game and breaks the game for me.

It messes up with the national happiness, all the nat. happiness u get gets put in your capital and something gets messed up there, this includes the happy from city states /opportunity's and probably a lot more things.

In my last game my happiness showed as +41 yet my city's didnt grow and production was reduced, i checked the golden age bar and it showed i was losing 28! towards the golden age counted because of unhappiness (this is what the game thinks your happiness is and when city's stop growing etc)
I went ahead to count all my happiness and unhappiness and i should be at +26(so both of them are wrong, the diffrence with the happy shown is with the city states)

I wouldnt recommend playing the beta in its current state unless your trying to help and test.