[NFP] AI crazy about Science

[Infixo]

@MrRadar The Culture Victory works differently than SV. Meaning - you need a different yield. For SV you need Science and virtually only science, which comes mostly from Campuses. This is VERY easy for AI to go for.
For CV you don't actually need Culture - you need Tourism. Culture in fact is needed to defend against CV. And Tourism comes from different places, which AI doesn't handle well... maybe except Wonders. You can easily force AI to build wonders.
AI has troubles building Theater Squares and then activating GWAMs. Lots of them wandering around. Getting CV through the usage of relics - probably not possible atm because it needs cross-coordination with Faith and Religion. Tricky.
In Real Strategy whenever a civ goes for CV, the mod boosts desire for Tourism, Wonders and Great Works. Partially Culture too. The AI can achieve CV this way but it is slower than SV usually.

 

[MrRadar]

@MrRadar The Culture Victory works differently than SV. Meaning - you need a different yield.

Yes, I understand that. AI's best shot for a CV probably comes from spamming Rockbands, so those who have a lot of faith have advantage. The only instances I saw AI to win or even approach a CV was through a spam of very lucky and successful Rockbands. But AI also is very keen to use Music Censorship, so Rockbands have a limited window.
No, I thought more Culture would bring them to advanced governments faster, as they tend to stagnate with T1 governments for far too long, then they usually go Fascism and then they pick something from T4 at the very end. But not much use from that, as there's still that wasting of 30-50 turns while they avoid Nanotech. Do you have some idea, what could be the reason for this? Plastics seems to be the first tech that is kinda cursed for the AI - some will go as far as Future Tech before they start on Plastics.

 

[Kotofei]

@MrRadar I just finished running some similar experiments myself. tl;dr, same conclusions as you. I can nudge the AI toward/away from certain behaviors, but, yeah, it seems hard-coded to throw an interception at the start of the 4th quarter, then fumble the kickoff on its own 20 yard line.

I made a few "personal taste" changes to the Classical section to make the game a little more interesting, but it doesn't change anything seriously.

 

[DizzKneeLand33]

Has anyone had success with programming the AI to have more desire for conquest? I mean serious conquest, not just taking a couple cities. And, does it even declare Formal Wars at this point? I put in an AI favored item for formal wars (at 250 value) and surprise wars (at 25) just to see if there will be more aggression longer term in the game. Don't really know if it will do anything at all, here's to hoping!

 

[aieeegrunt]

Has anyone had success with programming the AI to have more desire for conquest? I mean serious conquest, not just taking a couple cities. And, does it even declare Formal Wars at this point? I put in an AI favored item for formal wars (at 250 value) and surprise wars (at 25) just to see if there will be more aggression longer term in the game. Don't really know if it will do anything at all, here's to hoping!

The AI is ridiculous passive. I can have two archers and a warrior, border on a Trajan who hates and regularly denounces me because reasons with multiple legions visible and he does nothing

 

[Heyesey]

The AI is ridiculous passive. I can have two archers and a warrior, border on a Trajan who hates and regularly denounces me because reasons with multiple legions visible and he does nothing

(They just plain don't like you.)

 

[aieeegrunt]

(They just plain don't like you.)

or they like me too much?

 

[reddishrecue]

I think the devs saw people teaching how to play civilization simply by emphasizing on research before. Seems to me that the devs are teaching the Ai to do their research. Players too can learn from this because they have to stay ahead in technology so that they won't stay behind. Its going to get hard for the noobs but that's why there's lower ranking AI for them so they can learn their research and eventually find a way to victory.

 

[MrRadar]

I think the devs saw people teaching how to play civilization simply by emphasizing on research before. Seems to me that the devs are teaching the Ai to do their research. Players too can learn from this because they have to stay ahead in technology so that they won't stay behind. Its going to get hard for the noobs but that's why there's lower ranking AI for them so they can learn their research and eventually find a way to victory.

Inspired by the above, I thought I'd do another test with a bit more emphasis on science. But, I'm afraid, I've made a typo of my own. Not sure what's happened here, I must've passed out on the keyboard while editing the Victories.xml file... or an elephant (which, for the record, I don't own) sat on the "0" key:

The result was T306 (Deity AI, standard speed size, no modes, GS rules, lasers projects listed as favorites) Kongo SV:

Spoiler :

Although that's a noticeably faster AI finish time, still not sub t300. We'll still need a bit more emphasis on science, I'm afraid.

 

[Gedemon]

Note that there is a probability that this kind of high value is replaced by a default internal value during the game's initialization.

 

[MrRadar]

Note that there is a probability that this kind of high value is replaced by a default internal value during the game's initialization.

What could be that default value?

Yes, I understand that the impact of adding just one 0 there has much bigger effect than all the subsequent 0s, however many. You can't build more than 1 campus per city, after all, so more zeros don't do or matter much.
But the general situation resembled to what's happening now - less units, less wars, less religious units.
And there was a curious anomaly - Norway went for Plastics-Nanotech line first thing, but they never researched Rocketry. They knew how to launch to Mars, but had no spaceports, nor sattelites or Moon landing knowledge. They alone.

And AI is really impotent against barbs. Germany settled a Paititi city. A camp was next door, permanently triggered. So barbs razed the city. After some time, Germany resettled it. Barbs razed it again. Only the third settlement was a charm and survived. City states go after barb encampments with a vengeance, but AI civs seem to completely ignore them. Would it be so difficult to give the AI civs the behaviour towards barbs that City states have?

 

[Gedemon]

What could be that default value?

can't say without the source code, for civ5 this kind of value had a maximum of 32767, and the default value was usually the initial value in the DB from the base game (at release, unsure, but I think it was not updated with balance patches)

also no idea if setting a value > max will use max or default.

 

[Kotofei]

Although that's a noticeably faster AI finish time, still not sub t300. We'll still need a bit more emphasis on science, I'm afraid.

This looks like a fluke. No other AI has launched the Mars Colony at T300, and those science yields are high but not outrageous, considering it's past T300.

I ran a similar test, trying to make the AI a yield powerhouse. tl;dr I could make the AI put big numbers on the yield ribbon, but can't change behavior.

Edit:Yields aren't the problem. The AI usually beats me to Smart Materials. It's all the awful decisions it makes that you already observed.

 

[Aristos]

can't say without the source code, for civ5 this kind of value had a maximum of 32767, and the default value was usually the initial value in the DB from the base game (at release, unsure, but I think it was not updated with balance patches)

also no idea if setting a value > max will use max or default.

or go overflow.

 

[Tuvok694]

"AI crazy about science" - I can absolutely confirm that. I am playing on level immortal or emperor (mostly unsuccessful). In the last few days I started about 20 games as Shaka and I could not win any of them. I tried to use Shaka's special abilities by proritizing military. Always when I thought, my troops were stromg enough to attack a neighbor civ, they had fewer but way more modern units, and I could not conquer any cities. When I had swordsmen and archers, they had men-at-arms and crossbows. When I had men-at-arms, impis, crossbows and trebuchets, they had musketmen, field cannons AND renaissance walls.

 

[Warmoonger]

Just want to ask about there being two victories.xml.

There is one in configuration -> data and one in gameplay -> data. I edited the one in gameplay - is that correct?

Please could someone explain why this file occurs in two places.

 

[Laurana Kanan]

The Victories.xml in the "..\Configuration\Data\" folder is only for the game Advanced Setup screen, where you can toggle the individual victory conditions on or off. The Victories.xml in the "..\Gameplay\Data\" deals with the actual mechanics of the victory conditions in-game. There's nothing in the former xml file other than attaching names and descriptions to the victory conditions on the setup screen. Depending on what you are changing, the latter xml is the one you'll want to look at.

 

[Ash plays civ]

@Infixo I really appreciate and enjoy your mod. Thank you for sharing it. I just wanted to ask how I can tweak it have the AI place higher priority on improving and repairing their worked tiles?

 

[MrRadar]

I was thinking again what could cause AI to leave the Plastics-Nanotechnology line until the very last thing in the game and looked at the victorie.xml files again. And then I realised that they probably forgot to update AI priorities for victory techs.
Vanilla Victories.xml files contains AI project and victory techs priorities:

But in Expansion2_victories.xml they've only updated victory projects:

and requirements.
But I've found no trace of victory techs update, so AI must be still going on vanilla priorities, however GS projects are no longer unlocked by the the above victory techs.

So, by the above analogy, I updated the victory techs like this:

and ran some test games under the same settings as before, hoping for a visible AI improvement in teching paths for the SV, however the results were somewhat lackluster.

AI teching choices changed, at least they no more went for a Future tech before going for Plastics, however, in a majority of cases science leaders still went for Smart Materials first, before going for Nanotechnology, and then for Offworld Mission.

And they will not build any laser projects unless they have aluminium. The only way to make them build terrestrial lasers is to include terrestrial laser project as AI favored item, but not Lagrange. Then they build it immediately and like crazy, everywhere. If you include Lagrange as well, no terrestrial lasers at all.

AI victory times with the updated victory techs priorities were as follows:
Test 1 – current official Victories.xml file with +150 science bias, no priorities for laser projects: t333 SV, Mvemba won, he also built one or two Lagrange lasers.
Test 2 – science bias reduced to +15, campus and TD priorities for all civs in Classical deleted, both laser projects prioritised: t364 SV, Frederic won, nobody built any lasers, as they everybody used up their Al for planes ,etc.
Test 3 – gradual science bias increase, as in the file attached somewhere above in the thread, only terrestrial laser prioritised: t319 SV, Mvemba won, he achieved 11 ly speed with terrestrial lasers.

No luck in bringing AI SV under t300 so far.

Btw, is the deforestation factor formula available somewhere in the open in some files? Or buried in the .dll? I was actually looking for this first, when I had the idea to take a closer look into the Nanotechnology avoidance.
I vaguely remember one test game a month or so ago, where after jumping first to 50% in one turn, eventually it started dropping: went to 30% and then to 10%, if I'm not imagining things. Could it be that they reversed some signs in the formula somehow in the NFP?

 

[Pfeffersack]

(...)
Btw, is the deforestation factor formula available somewhere in the open in some files? Or buried in the .dll? I was actually looking for this first, when I had the idea to take a closer look into the Nanotechnology avoidance.
I vaguely remember one test game a month or so ago, where after jumping first to 50% in one turn, eventually it started dropping: went to 30% and then to 10%, if I'm not imagining things. Could it be that they reversed some signs in the formula somehow in the NFP?

Not sure if it helps, but JNR (Steam name, not sure if active here) has done a mod changing deforestation:

https://steamcommunity.com/sharedfiles/filedetails/?id=1868147338&searchtext=deforestation

Following the mods description I don't think that the changes made would be possible without having access to the formula, but I'm no modder