Statecraft too popular

[saamohod]

I've played several games with the latest version of the game and noticed that AI is overwhelmingly biased towards choosing the Statecraft policy tree.
Roughly 50-70% of AIs have Statecraft as their second policy in each of those games.
Not sure if this can be considered a bug but I will report it to GitHub just in case.

 

[DeAnno]

There's a lesser version of this problem where AI are reluctant to use Progress, compared to Tradition and Authority. Usually among 7 AIs I see 3-4 each of Tradition/Authority and 0-1 Progress.

 

[InkAxis]

There's a lesser version of this problem where AI are reluctant to use Progress, compared to Tradition and Authority. Usually among 7 AIs I see 3-4 each of Tradition/Authority and 0-1 Progress.

Really? In my games, it usually seems like there are too many progress civs. In my last game, there was only one tradition civ (Korea) and one authority civ. Every other AI went progress. And that's fairly usual. I guess it's because there are a lot of civs that are suited to going wide/war so they all would pick progress.

 

[vyyt]

I wonder what is the AI's decision to pick a starting tree based on? Does ratio of number of players / number of land tiles on a shared mass or distance from other civs play a role?

Do the people seeing progress picked more often decrease the number of civs?

 

[InkAxis]

IIRC, a big factor is the AIs trait. So like Dido will pick progress because their UA, Korea picks tradition, Gengis picks Authority. But I'm not sure how more in between civs choose. And if they can override their default choice.

Do the people seeing progress picked more often decrease the number of civs?

I don't usually decrease the number of civs. Sometimes I'll even see a civ go progress but not have many cities.

 

[saamohod]

There's a lesser version of this problem where AI are reluctant to use Progress, compared to Tradition and Authority. Usually among 7 AIs I see 3-4 each of Tradition/Authority and 0-1 Progress.

I cannot confirm this. Progress seems to be relatively popular in my games.

 

[vyyt]

I don't usually decrease the number of civs. Sometimes I'll even see a civ go progress but not have many cities.

I wonder why we see different AI behavior.

 

[Stalker0]

I wonder why we see different AI behavior.

Could be simple confirmation bias (unless you play a large number of games and literally record each game what each AI picks...its easy to notice it a few times and suddenly think that's how it always work).

Also, the AIs do have some randomness in their selection unless on Immortal/Diety….so there is that factor as well.

 

[tothePAIN]

The AI is inconsistent with policy choices at lower difficulties. At higher difficulties, the AI is supposed to make the optimal choices. The AI at higher levels (at least, Deity) is very consistent in its initial policy choice. I played an emperor game and noticed that the AI was picking progress, artistry, and fealty much more often.

For the posters above, to compare apples to apples, we'd have to know which difficulty these observations are seen on.

I will confirm that on deity at least, statecraft is the most popular 2nd policy tree choice, by far. Artistry is rare - sometimes I'll see Arabia or Egypt with Artistry, but they also go statecraft sometimes. Warmongers tend to go authority - fealty, but they also go authority - statecraft.

 

[Recursive]

Ahaha, Gandhi just doing his job - this happens every time someone misspells his name as "Ghandi" (many Diety and Emporer players do this).

Could be simple confirmation bias (unless you play a large number of games and literally record each game what each AI picks...its easy to notice it a few times and suddenly think that's how it always work).

Also, the AIs do have some randomness in their selection unless on Immortal/Diety….so there is that factor as well.

OH NOOOOO! WE NEED TO GET TO THE BLAST SHELTER!

If anyone wants to test and record AI policy choices, set <PolicyNumOptionsConsidered> to 1 in DifficultyMod.xml for the difficulty you're playing on so they always make the "optimal" choice.

A guide on how to do this can be found here.

 

[DeAnno]

To be clear, the majority of my games have been Standard Oval Immortal

 

[azum4roll]

Standard sized Communitu_79a v1.16.1. King difficulty with 8 AIs 0 human. AI starting warrior set to 0 (so same as human start) and all options considered set to 1.

From my observation, authority AIs seem to perform the worst (and often completely wiped out), and tradition AIs tend to win. 75% of the AIs would pick Statecraft, and the others fall behind. The only fealty picker tends to be a non-founder.

Rationalism and Order seem to be the leader's pick, but at the end all 3 ideologies get roughly equal share.

 

[JamesNinelives]

Interesting. I've been playing a lot of games lately myself. Some of my settings are non-standard though: huge map (12 civs, 24 city-states) and epic speed). Map is Continents Plus, events enabled, Emperor difficulty. Other settings are default.

Authority and Tradition both seem to be doing well recently in my games. I have noticed Statecraft being popular but not overwhelmingly so, or those who don't pick it falling behind. E.g. Authority/Artistry/Imperialism Polynesia had a strong lead in my most recent game, with both a Lapis Lazuli and and Whales monopoly and many city-state allies. In the game before that, Tradition Brazil, Authority Sweden and Tradition Babylon were all quite strong (some with Statecraft, some with Fealty).

 

[2506]

Standard sized Communitu_79a v1.16.1. King difficulty with 8 AIs 0 human. AI starting warrior set to 0 (so same as human start) and all options considered set to 1.

From my observation, authority AIs seem to perform the worst (and often completely wiped out), and tradition AIs tend to win. 75% of the AIs would pick Statecraft, and the others fall behind. The only fealty picker tends to be a non-founder.

Rationalism and Order seem to be the leader's pick, but at the end all 3 ideologies get roughly equal share.

Why do you set the AI starting warrior to 0? Didn't you receive a starting warrior?

 

[Recursive]

Why do you set the AI starting warrior to 0? Didn't you receive a starting warrior?

AIs normally start with 2 Warriors.

 

[ElliotS]

99% of complaints are just pure randomness being expressed. I did the math a while ago but long story short if every policy tree has an equal choice to be picked, then it's almost guaranteed that SOMEONE will see every AI pick the same policy for a few games and complain about it. Logs are needed to prove anything.

 

[2506]

AIs normally start with 2 Warriors.

Thanks, I thought it had been equalized after all the recent AI performance gains. Where do I look to see what the AI starts with on each level?

 

[Recursive]

Thanks, I thought it had been equalized after all the recent AI performance gains. Where do I look to see what the AI starts with on each level?

https://civ-5-cbp.fandom.com/wiki/AI_and_Difficulty

 

[JamesNinelives]

99% of complaints are just pure randomness being expressed.

I know this might be intentional hyperbole, but I wanted to point out that if we're talking going to talk statistics, 99% is unreasonably high. Even if 90% were true, that's an important difference.

 

[ElliotS]

I know this might be intentional hyperbole, but I wanted to point out that if we're talking going to talk statistics, 99% is unreasonably high. Even if 90% were true, that's an important difference.

I'm not super active any more but I was for like a year and a half, and not a single patch went by without someone complaining about some policy or ideology being picked by all the AI, and I can't remember a single case where it was an actual bug. I know it's anecdotal, but I would bet money that if I searched through patchnotes and other threads I could find well over 300 complaints and I doubt I'd find a single one with logs proving any sort of bias or overvaluing. If anything I think the real number is 100%, but I figured I might have missed a single patch where some sort of bug existed that actually caused the AI to mistakenly overvalue a choice.