3.10.14 AI-Only Game Statistics - Now With Some Religion/Instant Yield Insights

[CppMaster]

I completely disagree ... I wonder if you even played back then .. they would take damage YES, but they'd also RETREAT if very damaged - which means that a healthy unit would replace it's damaged counterpart - just to continue the process! Sometimes a city fell in just a few turns, if one had failed to build walls. One was never safe!

Maybe on higher difficulties AI should be more risky with their units, because they can have them many more thanks to bonuses.

 

[azum4roll]

The AI shouldn't be factoring in their own handicap bonuses.

 

[rkkn]

they don't need to. they can just be more aggressive on higher difficulties

 

[FinMc]

New Version - 2.7.3 (October 7, 2022)

AI also retreats when a unit is damaged, even by just 1 point.

forums.civfanatics.com

This has gone on for over 2 years now, with idling melee units NOT attacking cities and people ask why there are no domination victories?

The tactical AI has gotten a lot better overall but I remember you posting this and it definitely has impacted the game.

I have started paying a lot more attention to AI vs AI wars in my games. I even build them roads with IGE to help them press attacks and it is still a monumental struggle for them to take cities even against much weaker opponents.

 

[L. Vern]

Azum asked for this on the discord and others may find it interesting, here are the era progression timings (note that these may look a bit slower than you're used to as these games are run with no ancient runes or events)

Spoiler era attainment :

 

[Legen]

Its actually wild how much worse the warmongers are doing (or if they are doing well, its no longer by domination). Look at the zulus, dramatically weaker, and now a lot of their wins are by time....they just aren't able to close out the win and take those capitals.

Not every warmonger is doing worse, Russia and Assyria are examples of warmongers doing better.

My initial hypothesis is that the on kill handicap bonuses have made wars a solid source of science to warmongers, to the point that two things are happening:

1. Warmongers that already had a solid source of science are better able to translate warring into science victories. E.g. Russia, Assyria.
2. Warmongers that don't are overall in worse shape, likely losing games to the warmongers that do have bonus science. E.g. Zulu, Aztecs.

Exceptions are the warmongers that got a major balance change last congress, such as Persia and France, both of which fared better in this latest test and don't have bonus sources of science.

 

[azum4roll]

Those two aren't even your traditional warmongers. They're perfectly fine going fully peaceful.

 

[Legen]

They're hybrid warmongers. Still warmongers.

 

[azum4roll]

If you count everyone with a combat bonus warmonger, Inca, Shoshone, Iroquois, Polynesia are all warmongers.

 

[Legen]

My main criteria for whether an AI counts as a warmonger is the following:

LeaderPersonalitySweeps.sql

-- Catherine (Russia)
UPDATE Leaders
SET
    PrimaryVictoryPursuit = 'VICTORY_PURSUIT_DOMINATION',


-- Ashurbanipal (Assyria)
UPDATE Leaders
SET
    PrimaryVictoryPursuit = 'VICTORY_PURSUIT_SCIENCE',
    SecondaryVictoryPursuit = 'VICTORY_PURSUIT_DOMINATION',
    Personality = 'PERSONALITY_CONQUEROR',

We can debate on whether Russia's and Assyria's kits really fit for a Domination path, but, for the purposes of AI tests, their leader's focus and/or personalities are unambiguously set as warmongers.

 

[ilteroi]

New Version - 2.7.3 (October 7, 2022)

AI also retreats when a unit is damaged, even by just 1 point.

forums.civfanatics.com

This has gone on for over 2 years now, with idling melee units NOT attacking cities and people ask why there are no domination victories?

relax. just last week i implemented a change to make the AI attack cities more aggressively if they heavily outnumber the enemy. but it's not released yet.

 

[CppMaster]

The AI shouldn't be factoring in their own handicap bonuses.

Why not? It could improve decision making. IIRC, in vanilla AI couldn't start a war on the player on Settler difficulty. There was even a case where AI makes a decision out of top K choices, where K is dependant on difficulty. For Deity and Immortal it was K=1, for Emperor and King it was K=2 etc.

 

[Phosphoraptor]

1 stat I would like for clarifying the warmonger performance is score at game end, then we can see if they lose because they were weak or because they cant chase DV.

 

[azum4roll]

Why not? It could improve decision making. IIRC, in vanilla AI couldn't start a war on the player on Settler difficulty. There was even a case where AI makes a decision out of top K choices, where K is dependant on difficulty. For Deity and Immortal it was K=1, for Emperor and King it was K=2 etc.

But they don't think "settling more will give us more bonuses than buildings will give us so just spam settlers", or "we just ram our units into humans because we get yields from kills and our units are cheaper than theirs".

 

[CppMaster]

But they don't think "settling more will give us more bonuses than buildings will give us so just spam settlers", or "we just ram our units into humans because we get yields from kills and our units are cheaper than theirs".

Yeah, currently they don't. I meant that they could if it'd improve decision making.

 

[azum4roll]

They shouldn't make different decisions between difficulty levels when given the same game state. You may argue about the Tactical AI, but it simply thinks deeper at higher difficulties and the decision code is the same.

 

[CppMaster]

They shouldn't make different decisions between difficulty levels when given the same game state.

Ok, but what's the reasoning behind that?

 

[ma_kuh]

Is it just me or are all the high performers affiliated with Great Persons in some way? And except for Great General, Admiral, and Merchant generation, having an ability that enhances Great Person generation in your kit correlates with high winrate.

I guess Celts + Musician is an exception to this hypothesis. Germany is also an exception in the other direction. Brazil is only tangentially related to GPs, in that lower happiness means more specialists can be worked. But the GA focus means they gravitate towards Artistry.

 

[dostillevi]

One change that might help warmonger city capture would be to increase willingness to risk or lose units if they are assaulting an opposing capital, and maybe increase that slightly as they get closer to capturing all capitals. The AI should not be conservatively attacking the last standing sovereign capital. Likewise, if they aren't increasingly focused on capital cities specifically in the late game, maybe that could be prioritized, or prioritized further?

Edit: If they aren't already, warmongers should increasingly be willing to recklessly attack capitals if they believe anyone is getting close to victory.

 

[Anarcomu]

recklessly attack capitals

Or try to vassalage, it also works for victory