Expand or DIE - Concept Unknown to AIs

[bcaiko]

Folks -

I've played multiple games now, and the AI seems to fully misunderstand that a strong military and early expansion is in their best interest. Most AIs I've met have kept to 3-4 cities - well under the suggested 6-7.

I think the universal consensus on these forums is more cities is better. But the AI seems to fundamentally misunderstand this. I've even met Civs who never expanded past their first cities. Some of these AIs have been hampered by barbs (the fact that this happens is yet another example of poor AI behavior - why aren't their settlers escorted?). In my latest game, Germany had lost a myriad of workers and settlers to the Barb, and by the time I got there, his capital was under assault and nearly taken by normal barbarians (they had spawned some siege weapons). I saved his bacon and then took his capital for myself as payment.

But that's not always the case. My neighbor, Victoria, in my first game never expanded once even into the Medieval Era. I finally just walked in and removed her from her poor existence. To paraphrase Shakespeare, there's a lot rotten in the AI behavior code.

Thoughts?

 

[skallben]

Agreed. They are very poor at expanding. Also they leave settlers unprotected all the time.

 

[Matthew.]

On a slightly related note, this game really needs to hold off several mechanics for like the first 60 turns or so.

What I mean:

--Warfare. It can be okay for a Civ to rush a near city-state for flavor reasons and interesting story lines, but as a default it is only hurting AI performance. Many Civ's cannot recover from failed early military play. The wasted production on crap units which they often cannot upgrade (don't have iron for the warrior spam), the wasted opportunity cost of not focusing on infrastructure and city expansion. The game often feels over by turn 60, because the AI cripples itself early on. I know Civ 5 had some code to prevent the AI from declaring war too early, and with good reason. It is far better for them to get some core cities and infrastructure down before engaging in warfare.

--Agendas and diplomacy. It really cheapens the entire idea when these are implemented from turn 1. Yes, my culture sucks. I haven't had time to even build a monument. Yes, my expansion sucks, I haven't built a settler yet. Navy? Nope, not yet. Every AI you meet and engage with is overly hostile because you simply haven't had time to establish anything, so regardless of what their agenda is, you have already failed to meet it. The only exception is good ol' Teddy, who is glad you haven't started any wars on your continent within the first 10 turns of a game. Bully for you!

In short, there really needs to be a starting phase of the game where factions establish themselves before a lot of these mechanics (warfare, agendas) start, because currently it is just poorly implemented.

But yes, as to the OP, there needs to be a greater focus on AI's building up their empires in general. Both expansion (early and still into mid-game), but also within their empire (improving districts, actually upgrading military, maybe even implementing code to have them delete obsolete units instead of warrior spamming without iron,etc.)

 

[CaiusDrewart]

--The AI needs to be programmed to expand much more aggressively, and also to ALWAYS escort its Settlers. The AI should NEVER allow a Settler to leave a city without at least a warrior attached to it. This should be hardcoded. In Civ IV it was--on high levels an AI Settler would never venture forth without 2 archers. This was a good design decision. It's absurdly easy for players to capture AI Settlers right now, and even barbarians are having a pretty easy time doing it. This really, really cripples the AI.

--If you can get the AI to claim more land, they will probably get more strategic resources, which will help with the unit upgrading thing. Part of the problem here also seems to be that the AI isn't prioritizing military techs enough. Some preferences need to be changed so that the AI cares about always having a modern military.

--The AI needs to be more aggressive in clearing barb camps, and needs a larger combat bonus against barbs. I keep seeing AI core cities absolutely ravaged by barbs well into the Medieval Era, and sometimes even later. Barbs are in the game to pose a threat for the human player, not to stunt the AI's growth. Give the AI some more help here.

 

[teakbois]

--The AI needs to be programmed to expand much more aggressively, and also to ALWAYS escort its Settlers. The AI should NEVER allow a Settler to leave a city without at least a warrior attached to it. This should be hardcoded. In Civ IV it was--on high levels an AI Settler would never venture forth without 2 archers. This was a good design decision. It's absurdly easy for players to capture AI Settlers right now, and even barbarians are having a pretty easy time doing it. This really, really cripples the AI.

--If you can get the AI to claim more land, they will probably get more strategic resources, which will help with the unit upgrading thing. Part of the problem here also seems to be that the AI isn't prioritizing military techs enough. Some preferences need to be changed so that the AI cares about always having a modern military.

--The AI needs to be more aggressive in clearing barb camps, and needs a larger combat bonus against barbs. I keep seeing AI core cities absolutely ravaged by barbs well into the Medieval Era, and sometimes even later. Barbs are in the game to pose a threat for the human player, not to stunt the AI's growth. Give the AI some more help here.

When Im spying late game it seems the AI doesnt build enough districts either. So im not really sure what the AI is using hammers on.

Funny thing i saw last game was Lisbon with a settler. Im guessing it went AI > Barbs > City State. i should have levied them to see what would happen if I placed it.

City States have little problem with barb camps (and really enjoy stealing my slinger's kill). i think the AI just doesnt scout properly/pay attention to when there are new spawns.

 

[Matthew.]

I wouldn't be surprised if it is half the time trying to build wonders in cities without the production to do so, so spends 40-50 turns of the game with a world wonder in production and not doing anything else.

 

[dwarfling]

They also seem to love build holy sites, campus, theater district, but barely any industrial district to be seem.

 

[Magil]

A lot of the time when they do expand they place cities in stupid places too. Like, say, in between your cities.

Sometimes I do see the AI with 8+ cities. But then sometimes you see an AI stuck on 1 city. Sometimes in the same game, this happened in the last game I played. Hojo and Pericles expanded quite well, but Hojo was also settling cities in marginal spots in between my cities annoyingly. Meanwhile for the entire game, 500 turns of epic speed space race (with the 8 ages of pace mod), Phillip only had one city.

I find Victoria falls into this trap a lot. She wants to settle other continents, but gets stuck until she's capable of doing so.

 

[elitetroops]

When Im spying late game it seems the AI doesnt build enough districts either. So im not really sure what the AI is using hammers on.

There's a log file (City_BuildQueue.csv) that logs all builds of all cities. I've noticed the AI (on deity) pretty much always starts by building a monument, then some other non-military build, like builder, granary or scout. The exception is city states, which usually start by building a warrior. I think they have less of those at the start of the game. Later in the game builds seem to be pretty random. I haven't noticed any difference in AI builds for when they are at war compared to when they aren't.

Here's an example from a deity game. About 5 turns ago Trajan declared a surprise war on me. I've slaughtered most of his backwards army and my infantry is marching on his cities. What is he building?

[table=head]GameTurn|City|Production/turn|Build|ProductionInvested|BuildCost
156|LOC_CITY_NAME_ROME|20.0|BUILDING_WORKSHOP|122.0|175
156|LOC_CITY_NAME_AQUILEIA|12.0|DISTRICT_HOLY_SITE|290.5|297
156|LOC_CITY_NAME_ANTIUM|49.4|BUILDING_POTALA_PALACE|1048.3|1170
156|LOC_CITY_NAME_LUGDUNUM|10.0|UNIT_CROSSBOWMAN|20.7|180
156|LOC_CITY_NAME_ARPINUM|6.0|UNIT_KNIGHT|97.9|180
156|LOC_CITY_NAME_PRESLAV_1|20.0|BUILDING_WAT|96.0|170
156|LOC_CITY_NAME_VELITRAE|54.0|UNIT_SPY|257.0|400
156|LOC_CITY_NAME_RAVENNA|12.0|UNIT_SPY|48.0|400[/table]

I can't get tables to work in this new forum format, but that should be read as GameTurn, City, Production/turn, Build, ProductionInvested, BuildCost. Production/turn includes whatever modifiers they have, production invested goes up by that amount every turn.

Moderator Action: Table formatting applied.

Only 2 out of 8 cities are building units. Those two are crappy tundra cities that make 6 and 10 production/turn. The crossbowman is scheduled to be ready in 16 turns, the knight in 14.

This pretty much confirms what I observed so far in the game. The AI has their carpet of old outdated units and once you've dealt with that, there are no more units coming and you can take their cities without much opposition.

Edit: They do seem to buy units in threatened cities though. Maybe the thinking behind the coding is that they buy units when needed and build buildings, which in theory is a good idea since they seem to have enough gold. If they only knew how to also keep up in military tech and how to produce units that are useful.

 

[bcaiko]

I find Victoria falls into this trap a lot. She wants to settle other continents, but gets stuck until she's capable of doing so.

That's often been my experience. In my last game with Victoria, though, she did actually expand because her continent was connected by land nearby London. But she didn't settle the area around London and was surrounded by French and American cities.

 

[Roller123]

they probably still have the Civ5 AI and are trying to play Tradition games