Preventing AI from Razing

[sourboy]

Quick question: how can I reduce the likeliness that the AI will raze cities? I know there's an "optimal cities" deal - and a % variable - but how does it all work? Also, the map is US East Coast and UK (300x300) - so capitals are extremely far from each other - and no forbidden palaces are allowed for use.

Can I merely adjust the "optimal cities" % for [Regent] from 90% to 200% or something? or is there another way around this?

Thx

 

[pinktilapia]

Does AI aggression level not have some weight in the decision as well?

 

[sourboy]

That's what I'm trying to figure out. What areas are the triggers, so I can reduce them somewhat.

 

[recon1591]

Ok the only way we've found sofar in the SOE team is to set the "optimal cities" deal - and a % variable to max ( 999 ). Then also uncheck the Pillage box on all units. It'll cut down from say 7 out of 10 cities to maybe 2 out of 10 cities being razed. But we haven't found a way to stop it completely. Hope this helps.

 

[KingArthur]

Adding culture to the cities prevents them being razed. I had to do this in my Zombie scenario to prevent the cities being automatically razed when captured. By adding 10 culture to each city on my map the cities are ny destroyed when they're captured. I didn't have to make any tweaks to optimal city number but this was a very small map.

 

[The Last Conformist]

There's two different kinds of razing; autorazing, which only happens to size one cities with less than ten culture, and deliberate razing, which happens to any city the player (human or AI) feels like. The later is affected by OCN; specifically, the AI is more likely to raze if it's above 2*OCN.

 

[sourboy]

Three great posts - that answers my questions.

All cities actually start with 10 culture, but the map is beyond huge - so I could adjust that for the major cities. The OCN was the presumed culprit - so maybe I will just max the % out. Wouldn't hurt to add a "reduced corruption" deal to keep gameplay reasonable - but in a war for independence scenario, corruption fits. Even the 7/10 to 2/10 razed cities is reasonable. I don't care if Blue Ridge gets razed, but when Boston is torched - that's a problem.

I'll combine the results of your answers & tweak the scenario a bit - I appreciate the help!

 

[Willem]

Another thing you might want to try is to set the flag that keeps all the culture of a newly captured city, available only in Conquests I believe.

 

[sourboy]

I switched the optimal city percentage to the max (999%) - and that solved it. Not a single city was razed the second test. Of course my tests are limited turns, so I guess it went from 4/4 razed to 0/4 after the adjustment.

 

[recon1591]

On a side note, did u uncheck the pillage box on the units as well. ????. I know when i first did it and didn't uncheck the box it took about 20 turns before the AI razed its first city. But with the pillage box unchecked it took around 100 turns before the first was razed.

 

[sourboy]

No I didn't uncheck it. Occasional razings are fine. I just don't want to see more than 1/4 of the cities razed.

 

[hIdDeN_eViL]

i thought pillaging only applies to tile improvements...

 

[sourboy]

I'm willing to bet that's true, but people seem sure it also influences razing. Something tells me that these are the same people that think the AI gets a secret bonus in a battle with a Human controlled unit. Of course, I've always said I would verify the unit thing some day, because there have been cases that were a little too unreasonable.

Hmm...early in the first period - maybe I'll test that out right now, while watching the hockey game.

In fact, I'll test both. Should be easy to set up. I'll post the results in a couple hours.

 

[sourboy]

The pillaging-razing test was inconclusive, but I did have some results on general combat:

Grassland (40) [10% bonus to defense]:
AI attacking Human: AI beat Human 95%. AI won random number 73%
Human attacking AI: Human beat AI 85%. Human won random number 73.5%

Hills (40) [50% bonus to defense]:
AI attacking Human: AI beat Human 90%. AI won random number 70.3%
Human attacking AI: Human beat AI 82.75%. Human won random number 63.2%

 

[hIdDeN_eViL]

does this mean AI gets some unfair advantage?

 

[Figaro]

What difficulty level was this? (*shocked voice: the enemy have been cheating us all along! The Swine!*)

 

[sourboy]

What difficulty level was this? (*shocked voice: the enemy have been cheating us all along! The Swine!*)

Regent, that's the closest thing to even sides.

does this mean AI gets some unfair advantage?

Well, it depends on whether this was chance or thoroughly proven. On one hand, The AI & Human rolled the same on grassland, but the AI's good rolls came in a cluster, thus producing an advantage. I hardly think someone would go through that much trouble to conceal an AI bias, so I feel it's rather inconclusive.

On the other hand, The AI had around a 10% advantage in both regards when attacking a hill (where the defending unit should have a higher defensive bonus). Again, limited testing plays a part - but it at least appears that a human defending in a defensive terrain doesn't get the same bonus as the AI would.

The bottom line so far, is that one thing is definately true. A unit attacking another of the same calibur, regardless of terrain, has a huge bonus. The grassland defending unit should have a 10% bonus, the hills - a 50% bonus. Both cases however, this bonus proved next to worthless. It's as if the attacker gets a 2-300% bonus just for attacking.

I think I'll run another set of tests tonight, and we'll see how they fair. I'll post results in a while.

 

[hIdDeN_eViL]

you'd think that if there was such a large attack advantage, someone would have picked it up by now, or at least it would hve been written somewhere.

 

[sourboy]

Grassland (40) [10% bonus to defense]:
AI attacking Human: AI beat Human 95%. AI won random number 73%
Human attacking AI: Human beat AI 85%. Human won random number 73.5%

Hills (40) [50% bonus to defense]:
AI attacking Human: AI beat Human 90%. AI won random number 70.3%
Human attacking AI: Human beat AI 82.75%. Human won random number 63.2%

Grassland (100) [10% bonus to defense]:
AI attacking Human: AI beat Human 95%. AI won random number 72%
Human attacking AI: Human beat AI 95%. Human won random number 77.4%

Hills (100) [50% bonus to defense]:
AI attacking Human: AI beat Human 79%. AI won random number 64.7%
Human attacking AI: Human beat AI 84%. Human won random number 67.2%

I think it proves that it's a pretty fair system between the AI & Human, but it can go in streaks.

Again, however, the attacker has a HUGE advantage. According to the way they supposively programmed things, the equation must be flawed. It does make the 'defending spearman defeating an attacking tank' scenario all the more incredibile - and it all the more demands an "over-run" feature for Civ4.

 

[The Last Conformist]

What units did you use?