Making the AI Better: No City-States

[Bleser]

So just for kicks and because I like to change up options with each of the games I play, I decided to try a game with zero city-states and a few extra AI sprinkled in to makeup the difference. Here's what I found

Negatives:

• An entire social policy tree that I usually take (Patronage) is useless.
• Some civ's city-state bonuses nerfed (Alexander, Genghis)
• No easy food, culture or resource boosts. But see below for counter.
• No buffer between my borders and the AI's. But see below for counter.
• The diplomatic victory type needs to be disabled. But most probably consider this a positive.
• Great Merchant's abilities limited to golden age.

Positives:

• The AI is better at war (both against me and other AI) since all forces are concentrated on the enemy and not at aimless city-states that you might be allied with around the world.
• I saw an AI re-capture their own capital from their foe (Greece took Washington; Washington re-captured 100 or so turns later (marathon speed)). Awesome!
• AI conducted sea invasions of other AI that succeeded.
• No city-state cheating. By this I mean I wasn't able to be allied with every city-state in the game and not have the AI try and steal them back (even though they have 25,000+ gold like in all my other games).
• No out-of-balance (IMHO) maritime city-state bonuses. (+1 per city whether I have one or one-hundred cities). I listed this as a negative as well in situations where you really need it, but I feel overall that it was more fun/challenging w/o.
• No buffer between my borders and the AI's. This made the AI more threatening since they focused on me and not my useless ally.
• Resource and luxury bonuses harder to come by and more necessary to fight or trade for, which added another tough element to the game that was MUCH more rewarding than just dumping 1,000 at a city-state.
• I got nuked multiple times for the first time ever (by Korea). I consider this a positive since it was new and made the game more challenging. This may have been a coincidence but I'm not sure - it was another 1st. Korea even nuked other AIs (Germany).
• Having an AI as an ally was more useful since they attacked the enemy and not the enemy's useless city-state allies.
• Diplomacy was more important for multiple reasons (trade, war allies, zero border buffers).
• Greatly reduced message noise each turn ("AI XYZ now declares it is protecting city-state ABC!").
• Less boring whack-a-mole game play (Give this city-state 1,000 , then that one, then that one, forever).
• Much, much faster turn times late-game since there were sixteen fewer AIs (the city-states) and their units to cycle through each turn.

I'm sure I'm forgetting more, but overall, it was awesome and I can't really see going back. The AI was more challenging, more fun, more engaged, and did things I've never seen before. I lost cities and had to re-take them. I got nuked. They lost their capitals and re-captured them. Having an AI as an ally was actually useful.

Anyway, just thought I would share if you were looking for something new to try!

 

[Aristos]

Aha.

Now increase the Inter-city tiles to a minimum of 4, and you are in for even more surprises. Believe me.

Much better that way, even if you have to sacrifice an entire SP branch.

 

[glider1]

Thanks for the suggestion Bleser. Aristos a minimum of 4 tiles between cities which is vanilla or 5? I kinda like 5. With proper adjustments it allows me to play on larger maps.

Question. In terms of general turn time performance, how many CS's = 1 AI CIV? Each CS is just one city and a few units, so removing 16 CS's = 2 extra AI's? The problem is trying to keep the late game turn times down for me.

Cheers

 

[AxelRhodd]

Need to add 1 more negative - Great Merchant cant raise money from CSs

But overall playing w/o CS is not bad. I play CS free games sometimes and find the gameplay more focused and fun.

 

[gladoscc]

Thanks for the suggestion Bleser. Aristos a minimum of 4 tiles between cities which is vanilla or 5? I kinda like 5. With proper adjustments it allows me to play on larger maps.

Question. In terms of general turn time performance, how many CS's = 1 AI CIV? Each CS is just one city and a few units, so removing 16 CS's = 2 extra AI's? The problem is trying to keep the late game turn times down for me.

Cheers

Each city state is a full-blown AI. There's research calculations, worker management, quests. I'd suggest 16 CS = 4 extra AIs.

If you want to keep the turns down, try small maps?

 

[Mesix]

The AI probably isn't smart enough to realize that there are no CS when selecting social policies.

 

[Aristos]

Thanks for the suggestion Bleser. Aristos a minimum of 4 tiles between cities which is vanilla or 5? I kinda like 5. With proper adjustments it allows me to play on larger maps.

Question. In terms of general turn time performance, how many CS's = 1 AI CIV? Each CS is just one city and a few units, so removing 16 CS's = 2 extra AI's? The problem is trying to keep the late game turn times down for me.

Cheers

Vanilla is 3 tiles BETWEEN cities (popup is confusing). I propose 4 BETWEEN cities; not including the cities.

1 AI = 8 CS, that is exactly my ratio too. An average civ seems to be 8 good cities, so I chose that ratio.

 

[nokmirt]

Vanilla is 3 tiles BETWEEN cities (popup is confusing). I propose 4 BETWEEN cities; not including the cities.

1 AI = 8 CS, that is exactly my ratio too. An average civ seems to be 8 good cities, so I chose that ratio.

Well if you looked at it this way. On a large map, like you said, 10 civs = 80 cities and 20 CSs, so around 100 cities on the map on average. So if you played without city states you could add 2 civs to make 12, with a little room to expand. In my games though, I think 1 civ= 4 CSs for more of a challenge. For a large map 15 civs sounds about right. The fact is most or least enough of those civs will be conquered during the game anyway.

Also, I have my min city distance set at four, which means a cities have to be 5 hexes apart. This really helps unit congestion and improves combat with the AI.

 

[ense7en]

This sounds interesting.
Might have to try this next game

 

[Ruler]

Im going to play next game like that, but I think its easier, AI with lots of gold get extra happiness so easily from CS and REX like mad. Just look when you see next time one of those runaway civs they probably own half CSs.

 

[Ford]

...

Anyway, just thought I would share if you were looking for something new to try!

Ok, that's probably the best explanation I've read contra CS. I've seen others say 'don't play CS he stoopid' which never really made me want to change.
I generally play with less CS than needed for simply buying a diplo victory, so some regular AIs have to be liberated, but I'll give it a try without now I've read this.
If I leave the diplo victory option on, would the AI be dumb enough to still go for it?

 

[ense7en]

If I leave the diplo victory option on, would the AI be dumb enough to still go for it?

One can hope?

 

[Chiyo]

Vanilla is 3 tiles BETWEEN cities (popup is confusing). I propose 4 BETWEEN cities; not including the cities.

Where can I change the maximum distance between cities?

 

[Bleser]

One can hope?

Just disable diplomatic victory as an option type and they should ignore it.

 

[Bleser]

Thanks for the suggestion Bleser. Aristos a minimum of 4 tiles between cities which is vanilla or 5? I kinda like 5. With proper adjustments it allows me to play on larger maps.

Question. In terms of general turn time performance, how many CS's = 1 AI CIV? Each CS is just one city and a few units, so removing 16 CS's = 2 extra AI's? The problem is trying to keep the late game turn times down for me.

Cheers

So this distance helps control the # of cities on the map and thus helps turn times? Simple, and I like it. Thanks.

Need to add 1 more negative - Great Merchant cant raise money from CSs

But overall playing w/o CS is not bad. I play CS free games sometimes and find the gameplay more focused and fun.

Added as a con - good catch that I forgot to mention. But GA is still useful.

The AI probably isn't smart enough to realize that there are no CS when selecting social policies.

Another good point. I can load up my last game and see if any of them took those branches.

Overall, thanks for the comments everyone. I was on a standard map in late-game and turns were very, very reasonable - < 5 seconds I would say. I have a fast computer (Core i5 750 @ 3.8 GHz w/ 12 GB RAM) and with the sixteen city states it would take 10-20 seconds in late-game turns. Plus all my allied CS units that I could "see" aimlessly moving around dragged down the turn times too, which has been eliminated.

I think I added four Civs in the place of the sixteen city-states and two of them were wiped out before I even got a chance to meet them.

And the reduced message noise (losing influence, XYZ protecting ABC) which I mentioned above is the real hidden benefit. Love it!

 

[Aristos]

Where can I change the maximum distance between cities?

In GlobalDefines.xml, there is a variable called MIN_CITY_RANGE or something like that, by default it has value 3, change that to 4 and you will have a minimum of 4 tiles BETWEEN cities (the tooltip is confusing, as it includes the city, so for a value of 3 the tooltip says 4, that's why you see people here referring to range of 5... so when you see people here talking about a range of 5, they refer to a value of 4 for that variable.)

 

[Aristos]

So this distance helps control the # of cities on the map and thus helps turn times? Simple, and I like it. Thanks.

Not only the number of cities, but maneuver room for the AI... you should see slightly better maneuvering with more inter-city space. I have.

 

[glider1]

I played out a complete game the other day on MIN_CITY_RANGE 5. It allowed me to play with a bit more than standard numbers of cities on epic speed on a huge map with reasonable turn times in the modern era. There was plenty of early conflict and basically was fun. What spoiled it was the CS's. I think Bleser is exactly right. The AI waters down and dissipates it's efforts poorly with them in. For example I had China wanting to attack me, but an allied city state was in the way. It could not capture the CS because I was helping the CS to survive, but if it would have been one of my cities and not a CS, it would have caused my own economy more troubles.

One problem on MCR=5 is that at the end of the game I revealed the map to find that quite a number of civs had huge underpromoted navies that they did not deploy. It was an ugly site for a fan of the civ series. It might be because of a bug in the naval AI with capitals that are spread a long way from each other. Perhaps it disables the use of navies but the AI still builds them. I think I know why the AI had huge under-promoted navies that it did not use. On maps like Perfect World, tectonics, fractal, the AI sometimes does not find an attack target for it's navy and so it does not deploy it's navy and does not promote it. The situation is not as bad on traditional maps like continents. However the real bug is that it at least should put some of that navy to work in any case, and probably stop building so many. Sigh, add that to the list of tactical bugs that we beg Firaxis to fix. The absolute worst navy bug at this stage is the ghost navy where one human frigate can mow down an entire naval fleet because the AI's navy is not under any control and does not defend itself. It's almost as if the AI navy logic is disconnected from the strategic AI. That bug is so bad that it actually has basically stopped me playing civ until it is sorted out.

The other big problem with MCR=5 is as Aristos says, the settler escort code is bugged, and the AI can loose high numbers of settlers to barbarians it's a shame to see. So perhaps more conservative settings in the distance between cities is a good way to go at 4:

<Update>
<Where Name="MIN_CITY_RANGE"/>
<Set Value="4"/>
</Update>

Another reason why I try to increase inter city spacing and play on huge maps, is that the AI formation code works like this:
1) AI decides on a target city
2) Moves attack division unit by unit towards a chosen position short of the city, range tending to be at the rear but quite disorganised.
3) Units stop short of city at this position and reorganises into a flimsy formation, waiting for the lagging units to catch up.
4) The distance between formation and city attack target can be up to 5-10 tiles from each other (the naval code works the same way).
5) At some point the formation then attacks.

The thing is that the period where the AI is waiting for the formation to form, requires neutral land (the equivalent of open water for navies) so that this process is not disturbed. That is why we go for bigger inter city spacing on larger maps, to generally create more neutral land for the AI's flimsy attack strategies. If the AI is attacked while it is forming it's army (by range fire from a city for example), it's efforts are totally destroyed because the AI code never is able to reorganise the formation. The formation is spoiled and the units then work independently like chooks on the barn house floor.

But Bleser has given me more hope, with the idea to disable the CS's for a better AI. I think it will help to create a bit more competitive games. Essentially with the tactical AI the way that it is, when we are in conflict with the AI and the AI has sufficient numbers of units, the tactics can still be quite complex. It's more like a one-sided puzzle mini-game rather than the cause and effect tactics between two opposing armies in battle. Tactics in Civ5 against this AI have very little to do with simulating war. But this light weight puzzle quality to tactics against the AI in Civ5, can still be quite a bit of fun if the AI throws enough units in.

Cheers

EDIT: Sorry Aristos. I'm not awake and I hit the post button after the first sentence. I've completed the post now.

 

[Aristos]

I played out a complete game the other day on MIN_CITY_RANGE 5

interesting. did you see noticeable different gameplay from the AI?

Did you check the early game to watch settlers behavior? That was one weak point of >4 range.

 

[nokmirt]

In GlobalDefines.xml, there is a variable called MIN_CITY_RANGE or something like that, by default it has value 3, change that to 4 and you will have a minimum of 4 tiles BETWEEN cities (the tooltip is confusing, as it includes the city, so for a value of 3 the tooltip says 4, that's why you see people here referring to range of 5... so when you see people here talking about a range of 5, they refer to a value of 4 for that variable.)

Yes that is exactly right I am changing that now. It means a space of a minimum of 5 hexes between cities.

In my current game, which will be in CiV Stories and Tales, because I need to start doing some writing again. These are my changes.

First of all, I am going to play a domination game on the YAHEM TSL large Europe map on King level and starting in the ancient era. Mods are info addict, agS resource panel, End at Renaisscance, YAHEM, and a CS leaderhead mod.

XML changes

1.) 12 civs and 8 CSs all civ leaders have had warmonger hate reduced to 0.

2.) All civs except England, Ottomans, Danes, and Spain have had their naval flavors increased by 1.

3.) In the GlobalDiplomacyAIDefines.xml Warmonger opinion weight has been reduced to these values.

<Row Name="OPINION_WEIGHT_WARMONGER_CRITICAL">
			<Value>50</Value>
		</Row>
		<Row Name="OPINION_WEIGHT_WARMONGER_SEVERE">
			<Value>35</Value>
		</Row>
		<Row Name="OPINION_WEIGHT_WARMONGER_MAJOR">
			<Value>20</Value>
		</Row>
		<Row Name="OPINION_WEIGHT_WARMONGER_MINOR">
			<Value>8</Value>
		</Row>
		<Row Name="OPINION_WEIGHT_WARMONGER_NONE">
			<Value>0</Value>

4.) In Global Defines.xml MIN_CITY_RANGE has been changed to 4.

This should be quite interesting. Now I have to think of some story lines that I can apply as I go along. I will be playing as Napoleon. Since I am studying him in real life anyway.