Does the AI play to win?

[Kyroshill]

War mongering seem to be the prime paradigm these days. I'm ok with that.

But Civ nor any of the games based on Civ are really meant to be "war" games. They have a military aspect.... but obviously the war engine is not remotely correlated to any recognized military startegy or tactics.

The point of the game is to explore options and possibilites to grow your culture from stone age backwardness into raging profitability and happiness for you and/or your constituents.

More and more though this game is turning into massive warrior spam, and then ax spam, etc... until all enemy cities have bit the dust.

More builder love dammit!

That is all!

 

[Mesix]

Micromanaging is a piece of cake for AI and computer logic driven systems. It seems hard to us because it is usually boring and we do not like to do the same thing over and over, but computers love itterations, and they are pretty fast at it too. So, this is not a problem for the AI.

My conclusion so far, is that the AI plays for you to lose or win. Not for it to win. But I am sure this will be taken care of at some point in the mod's development.

By experience, I can tell you that the first thing you try to do when programming a game, is to make it work in an acceptable level for the human player. After this is done, you then try to make the AI competitive, but not invinsible, which is a lot harder than make it either pathetic or Unbeatable.
I guess that the greater weakeness of all strategy games I have ever played( and I have played a lot), is just that. The AI programming is too complex if you want to maintain a challenge, and whatever you do, humans always find a way to adapt to how the AI plays. The most difficult thing is to introduce versatility and adaptibility to AI players in a strategy game. This is doable, but most strategy game producing companies haven't thought that it does worth the time needed to invest.
So, Kael and the team, have an AI inherited from a game, and I am sure they will try to do their best to make it better, but I wouldn't expect too much.

The point I have been trying to make is that the AI they inherated does play to win with the original game design. With all of the content added (and some taken away like Cathedrials and Wonders) in FfH, the AI seems only able to win by force (and it does that poorly). It would be nice if some of the AI players were coded to go for an Altar win while others were going for a Tower win and still others might be going for a Culture win. It would be nice to lose an occasional game to the AI when they win rather than just losing an occasional game when I lose my last (sometimes only in the early game) city.

 

[Asthix]

Its because of all the content added that the AI the team inherited has problems with FfH2. Also, the team is on the record (I think) as saying that AI tweaking is a necessarily slow process in that each tiny change requires extensive playtesting. We've already seen AI improvement recently though with magic use, upgrading units, worker logic, etc. I also think we're lucky to have a team that does this because they like it, so I bet we'll see more improvement sooner rather than later.

 

[Thunder_Gr]

The point I have been trying to make is that the AI they inherated does play to win with the original game design. With all of the content added (and some taken away like Cathedrials and Wonders) in FfH, the AI seems only able to win by force (and it does that poorly). It would be nice if some of the AI players were coded to go for an Altar win while others were going for a Tower win and still others might be going for a Culture win. It would be nice to lose an occasional game to the AI when they win rather than just losing an occasional game when I lose my last (sometimes only in the early game) city.

Well, I did read your first message carefully. The inherited AI was never to consider upgrading a unit to level 6, then make it a shade, then have it to join a city as a free specialist. Just an example was that. In addition, the inherited AI cannot first move and then consider a combat action. It even cannot handle structure creating combat units. Under BtS the AI can simply claim a cultural or space victory passivelly. While in FFH2 this is not doable. FFH2 forces energetic play for every type of victory, thus, the passivelly taught inherited AI has problems to adapt.

I am sure they will do it though, now that the major features programming is finished and they can move focus from the Human element to the Computer element, in order to have the AI adapt to those changes. Do not forget that we, humans, have already lost many cool gaming features because the inherited AI cannot use them...

 

[Valerarren]

In the very first FFH game I played, the AI won an altar victory (this was before BTS). I didn't really understand the victory conditions, and I had been winning in the millitary/domination style. I learned about the wonders of putting mages with fireballs on boats, and even got a few archmages with meteors that were happily going around capturing coastal cities.

I kept hearing about these altars being built (in a very far away land), but then suddenly the game was over.

 

[Jabie]

If there are AI improvements coming up in patches, may I recommend that the developers teach the AI to play to win.

I disagree. I'd far rather see the AI role-play. If I'm building the Tower of Mastery, am of good alignment, have neighbours with Good alignment and have carefully built strong diplomatic relations with my neighbours through years of peace, trade, assistance, etc, I wouldn't expect them to suddenly declare war on me because they had been programmed to win at all costs.

 

[Arcite36575]

If you want to see the AI "play to win" then:

1) up the difficulty
2) agressive AI
3) pangea map
4) double the # of civs (or more for fun)
5) enable vassal states

The game gets very nasty, quickly, with world peace being just a pipe dream (fortunately). The option of vassal states makes alliances makes very very big world wars break out by mid to late game, and by the end of the game, you'll have usually two very big "factions" (hopefully I'm the leader of one of them), and maybe a few irrelevant tiny countries.

 

[Monkeyfinger]

Lakes >>>>>> Pangaea for that kind of bloodbath game.

 

[Arlborn]

The AI won't win in BTS by "accident". In BTS the AI is actually programmed(by Blake) to win a cultural victory. They normally decide it by the time the last religions are found(also, only AIs a bit more peaceful go for it). They actually decide and know which 3 cities will be the cultural ones.They start build wonders and using Artists on those, and when they are near victory they up the cultural slider, build culture in those 3 cities and defensive units in all the others. That is nice programming.

The AI is also programmed to win Space Race. In fact that was so since Vanilla, but the code was probably improved further with Warlords and BTS.

Those 2 are the only ways the AI is actually programmed to win. Obviously this mod is too different from the normal game, so yea, AI needs to be changed.

 

[Emptiness]

Cultural victory on the other hand, just requires cities with a lot of culture. As long as you build at least a monument in each city, all that's required is time. If no other victory happened, it's inevitable that someone would eventually win culturally. It's one of the few ways to win that's not too hard to do by accident (which is why I always disable it.)

There just aren't enough turns in the game to win a cultural victory with monuments alone. Sources of culture don't double in value over time, and there seem to be fewer sources of culture available than in BtS. I've won a cultural victory in FfH2, and it required a great deal of focus on that goal. The ai just isn't capable of that, in its current form. I doubt that you've been in danger of cultural defeat by the ai. I've never seen the ai achieve legendary culture in even one city.

 

[Kyroshill]

I've never seen the ai achieve legendary culture in even one city.

As I said above.... in my last completed game I was beaten by Ethne who won a cultural victory.... and all three of her cities went legendary in fairly short order..... seemed well planned to me..... of course she was happily isolated and pretty much stayed that way

 

[WarKirby]

I disagree. I'd far rather see the AI role-play. If I'm building the Tower of Mastery, am of good alignment, have neighbours with Good alignment and have carefully built strong diplomatic relations with my neighbours through years of peace, trade, assistance, etc, I wouldn't expect them to suddenly declare war on me because they had been programmed to win at all costs.

This idea is best.

I really don't like the whole "wining" part of Civ/FFH. I enjoy the playing, not te winning. Journey, not the destination, etc.

I'd rather see all AIs designed to properly use the strengths of their civilisation, play in character, and just see what happens.

For example, Ethne/Eionion are too aggressive, IMO. I think they should only ever declare war on an evil civ, if they're right next to them. But otherwise, their designed to be defensive, and they should play like it, by not declaring war, and also by encouraging other players to make peace.

That makes me think of something, actually. Although there are many,MANY times an AI asks me to go to war with their enemy, not once, EVER have they asked me to STOP a war against someone. That's something the Elohim should really be doing, I think.

Doviello/Clan of Embers fit just fine with the warmongering AI as it is now, so not much change needed there.

I think Sidar should be very unwilling to have open borders, though, like tokugawa in vanilla to reflect their secretive nature.. They also need to be weighted a lot more towards building asassins, and trying to found council of esus.

Just some examples.
When I play FFH, I'm not really looking to just win a game against competitors. I play it to enjoy the feeling of erebus, I like the flavor of each race, and I really just wish they would play howthey ought to, and not really worry too much about winning or losing.

As is, the AI all seem to mostly research the same military techs. Quiteoften in mid/late game, I can look around and see 6 AIs at war with each other, all of which have exavtly the same techs. Warfare, Iron working, Feral bond, bowyers, etc. They don't seem to give a damn about unique flavor much, and jut go for the most powerful generic military path.

 

[GePap]

I think that aspect of FfH, that all the civs are so different, is what makes the AI a challenge. After all, in regular civ there are some basic aims all civilizations want to aim for, because while there are basic differences between all the civs, the commonalities are far greater. For example, all religions are basically the same thing, and tanks are clearly better than cavalry.

In this mod, each civ is very different, so its hard to balance between basic common needs, like setting up a sustainable economy, with the more particular aims of each distinct civ, so that the AI can play to that strenght. I have mentioned before for example the retisance of Bannor AI to call a crusade, whih makes sense in a way, given how radical a step that is, but also takes away the most powerful mehcanic in Bannor hands from the AI. Or the Sheaim chosing to follow any religion other than AV, which lowers greatly the growth of the AC and thus invalidates their great Planar Gate strategy.

In the end, those civs whose particular goal dovetail nicely with some basic goals, like the Calabim, Baselraph, and elves, whose aims almost always enebale them to build a sustainable economy, end up doing very well.

 

[Lone Wolf]

For example, Ethne/Eionion are too aggressive, IMO. I think they should only ever declare war on an evil civ, if they're right next to them.

For some reason, they are set to be backstabbers who are able to consider declaring war on someone with whom they are pleased with. Most FFH leaders are coded that way through, with the exceptions being a totally random group of:

Capria
Decius
Amelanchier
Arturus Thorne
Cassiel
Jonas Eldain
Tebryn Arbandi
Dain the Caswallawn

And some minor leaders.

As is, the AI all seem to mostly research the same military techs.

They nearly all have strong "militiary" flavours set.

 

[Imuratep]

If it would be possible to make the Elohim try to end wars diplomatically this would be fantastic. So let's extend this thought. Which leaders should do what flavorwise?

 

[WarKirby]

Ljosalfar are generally known for being peaceful too, are they not? I'm sure they would like peace.
Kuriotates are known merchants and traders, so they have a more peaceful slant too.

The svartalfar would probably want peace too, not because they're nice, but because covert activities are easier with the illusion of friendship. Hidden nationality units are pointless when you're already in open war.


Maybe cassiel should be more aggressive towards the founders of large, sucessful religions, (>25% world dominance or so ?) and have a relations bonus with any (non illian) civ that has no state religion.


Doviello/clan/calabim/sheiam/bannor etc, are all fairly classic warmonger civs, so less AI changes wold be needed from them.

I'd like to see the Sidar AI focusing on unit preservation, valueing the lives of any unit > lv1 far more, and only risking them in battles where victory is fairly likely (60%+) where possible. And of course, waning the units once they reach lv6.

The amurites really need to focus more on a magical path, and specialising their mages. I also think they should do what I do when playing them, stationing an adept in every city as that city's "guardian", giving them Mind I, Spirit II, Enchantment I and II, and Earth I, essentially focusing them purely on supporting the city, and remaining there permanantly.

The infernals need to be a lot more aggressive on the whole. Every game I've seen them AI controlled in, they got planted in one little corner of the map, and just sat there permanantly doing very litle. Although their power did rise steadily from manes, they never expanded beyod 2 initial cities, even though the amurite empire was right next to them, and I crushed them hundreds of turns later

 

[WarKirby]

For some reason, they are set to be backstabbers who are able to consider declaring war on someone with whom they are pleased with. Most FFH leaders are coded that way through, with the exceptions being a totally random group of:

Capria
Decius
Amelanchier
Arturus Thorne
Cassiel
Jonas Eldain
Tebryn Arbandi
Dain the Caswallawn

This really doesn't make much sense. It'd be nice to hear from Kael, why this is the case.

 

[Monkeyfinger]

I've never, ever seen either Elohim leader declare a war when controlled by AI. Go on the offensive and beat the piss out of someone who declared on them, sure. But they were always provoked into doing that. It was never the result of a war of aggression started by the Elohim.

 

[Lone Wolf]

I've never, ever seen either Elohim leader declare a war when controlled by AI.

That' because they are peaceful in general and a high peaceweight. But I've seen pleased Ethne declaring war on me when I was Cardith.

 

[WarKirby]

As above, I've been the victim of their strange idea of "peace". They've never had a problem declaring war on me.