Does the AI like Tradition too much?

I've been thinking for a while now that it seems that most AI choose Tradition for their first Policy tree. In an effort to reduce confirmation bias, I've been tracking AI first Policy tree selection. Based on a small sample size of games (I don't have a lot of time), it seems to be the case:

Spoiler Results :

This is with PolicyNumOptionsConsidered set to 1.

In most cases, the civs choosing Authority are ones that should be (Huns, Mongols, etc.) but very few non-authority civs choose Progress.

This means that if I would like to play a peaceful early game I should generally pick Tradition for the "We have similar policies" positive modifier.

Has anyone else noticed this? Is this something that should be adjusted?

 

Yea, especially since progress is a lot better now than it used to be. Based on how I view most civs tradition would actually be the most rare one because of how niche it is, the two other starting trees are more flexible imo.

 

It's funny because I think the AI would do better with Progress in general. Progress does everything automatically, doesn't require moves on the board like Authority, and it also doesn't require detailed and difficult micro of specialists vs food like Tradition. All it requires is building roads, which while it's something the AI often is slow at it it's not hugely slow. Tradition also requires some very specific placements of cities to defend at times which is another thing the AI isn't the best with.

IMO Progress should be the most common AI selection and not the least.

 

In my current game as Corsica (cool design, well done GH) There are 5 trad (germany, morocco, khmer, me, netherlands) and 3 authority (denmark, aztecs, ainu). Khmer and Corsica are like obvious trad-civs, and denmark and aztecs are authority-civs. I dont quite understand why netherlands and germany would go tradition, or why the ainu dont go for progress.

Oh well, at least it made me very happy and surprised when I managed to get Uni of Sankore with 4 other tradition-pickers on deity.

 

I disagree. Progress is very hard for the AI, it's not automatic at all and requires either heavy diplo, full war, or very clever GP play to go science with lots of strong cities, not just lots of cities as AI sometimes does in middle game. I think best AIs and most dangerous long term are tradition and short term authority. I usually see China, Ethiopia, Babylon, Austria, sometimes Byzantium, as top AIs, progress AIs are powerful only when map has a lot of warmongers or lacks strong tradition and has Germany or Siam insttead.

Progress AIs are often aimless, dont capitilize on roads, make few cities, and not getting monopolies fast. I see them on three cities for long. We can't change it fast as these are many small decisions. The strongest progresses tend to be Carthage and Polynesia due to some things automated by their uniques, not progress itself.

Progress gives strong cities early but mid game relies on religious synergies to stay relevant, AI cant be trusted to recognize and execute them, it's not as simple as divine inheritance for tradition or hero worship for authority. Late game progress requires a synergy and clear path to use culture from great writers it unlocks. Culture is only worth as much as something that its unlock benefits you. But you need to know that first. Other policies are more ai friendly cause better at at just get more units, take more cities so you can have all the capitals, or just get more wonders, GPs and theming bonuses. It is the most long term planning focused and tricky tree cause its lack strong focus. Challenge aspect too. You are more often afraid of tradition AI tourism from the other side of the map, or if Sweden swallows your neighboors, have long border and just became hostile than progress I think?

I think spread @Hokiefan00 sees is due to small number of progress so it's as it should be and I'm seeing this too. Only Carthage, Russia, Indonesia, Polynesia, Assyria go 100% progress on deity, Inca 50% in my experience. Second trees are very varied. Whether we should or can change it, is another matter.

I've said my piece

Spoiler :

 

This is very, very small modifier. 20? They will attack still if you are weak or forward settle them and not make them friendly if they don't have any shared interest with you. I can stay at peace as the only progress long usually. When to DoF and not, and where to settle are 100x more important.

 

Very interesting. Thank you. I did not know that Progress was less AI friendly. That’s a shame. I am not nearly a good enough player to determine what, if anything, should be done to make it more friendly to the AI.

 

I must admit in my recent games civs seem to be more 50-50 with Tradition & Progress. In my latest game Shaka has weirdly gone Tradition, with the only two Authority ones myserlf & Genghis both nearby.

 

Deity AI policy picks are set in stone. Going by memory there're 20 Tradition, 15 Authority and only 5 default Progress. [Edited after better checking some games]

Spoiler :

Ahmad al-Mansur Tradition
Alexander Tradition
Ashurbanipal Progress
Askia Authority
Attila Authority
Augustus Caesar Authority
Bismark Tradition
Boudicca Tradition
Casimir III Tradition
Catherine Progress
Darius I Authority
Dido Progress
Elizabeth Authority
Enrico Dandolo Tradition
Gajah Mada Progress
Gandhi Tradition
Gengis Khan Authority
Gustavus Adolphus Authority
Haile Selassie Tradition
Harald Bluetooth Authority
Hiawatha Authority
Isabella Authority
Kamehameha Progress
Maria I Tradition
Maria Theresa Tradition
Montezuma Authority
Napoleon Authority
Nebuchadnezzar II Tradition
Oda Nobunaga Tradition
Pacal Tradition
Pachacuti Tradition
Pedro II Tradition
Pocatello Authority
Sejong Tradition
Shaka Authority
Suleiman Tradition
Theodora Tradition
Washington Authority
William Tradition
Wu Zetian Tradition

A bunch of Tradition AIs have very little synergy with GPs or GAs, and in many games they might benefit from a wide progress empire. Or change them for diversity sake. Germany, Celts, Poland, Ethiopia, Portugal, Maya, Inca, Byzantium, Netherlands can run Progress over Tradition and probably do better. Especially Ethiopia and Maya, despite some GP/GA in their kit, work much better as progress due to early science/faith.

Greece picking Tradition instead of Authority instead is something I'm tempted to post as a bug, it goes against everything in their kit. Same Japan, as a player it might work with Tradition but the AI should go and play it straight Authority as well.

Some expansionist AIs that default Authority could also select progress due to lacking an early UU or a warmonger UA: Shoshone, England, America aren't great Authority picks for the AI if you ask me, they usually mass spearmen and horsemen, focus on the bottom part of the tech tree but fail to conquer any neighbour and fall off in the mid game.

 

what difficulty are you playing on? If you play on emperor or lower, there is some randomness on AI policy choices…unless you have tweaked the difficulty file yourself to remove it

 

Set in stone? Policy choice isn't dynamic based on circumstances? In what file is the list of choices defined?

 

@randomnub What you’ve said doesn’t change what I said and I don’t agree with your points and examples. Do you play deity? I do, and Ashurbanipal is always progress, not authority like you said while Babylon is never progress and always tradition. But it’s entirely possible that you play more than I do, I dont play a lot.

I would say the only faction I would change to tradition is not the ones you’ve mentioned but India. Everything in India case points to progress IMO and I would fear progress with strong religion and large cities much more than no religious bonuses and many cities. Food is the last thing you want as tradition. Greece you don’t like as tradition has hills bias, perfect for tradition. You could argue Morocco for progress as it is perfect for it and performs great diplo play, and is prop best or top 3 progress civ in game but you don't. I wonder why?

Again, progress is not AI friendly, it goes well with civs that do automatic stuff like Carthage gold and free city connections, not does automatic stuff itself. I see nothing wrong with tradition Greece, both AI and you can get hoplites, get tributes to get wonders and food in capital and you are free to rush top tech for your UB at drama and poetry which again gives culture. Warring as tradition is powerful as tradition more than as with progress and tradition provides a good early spike, same with Japan who needs tradition to get at all to its uniques, then can get a lot of free cultural GP along with Dojo being great at science and culture tradition wants.

You are wrong, tradition doesn’t require any GP or GA synergy, and that’s its strongest point. Strongest and most consistent top traditions like Ethiopia or China don’t even have GP or GA bonuses at all, negligible faith bonus aside. Changing Ethiopia to progress is just plain mad for me, like you not even know how and why both progress and Ethiopia works. You could argue authority for them perfectly, they get the fastest and most dangerous knights in the game anyway, but never progress. The same with Poland. Tradition Poland is the strongest civ in the game, hands down without even need for UB or UU. Again, authority Poland is brutally strong, only equalled by Songhai in my experience. Not progress Poland. Germany or Celts doesn’t have a problem with tradition, how you can say that for a civ that gets lots of science, culture and GA specifically in the capital, they have a problem with their design in AI hands that will stagnate them whatever they will go for. Portugal again, probably one of the best traditions and authorities in the game, and generally the third most flexible and powerful civ the game.

Netherlands, again, I see only tradition or authority if you want to maximise their stuff in synergy as bonus culture is affected by GA and overkill as tradition, coupled gold for a tradition production starved cities. Byzantium and Maya I see no problem with, they are many times the strongest civs in my games and tradition is the right call for them. Cause of faith, science and very strong, early UU they can both double as powerful authority. Shoshone and England are garbage that have no no uniques basically. Shoshone are beyond saving so authority is the right call, it’s hit or miss for them. England has better offensive navy, UB that doesn’t even require philosophy but provides science and gold (both most needed as authority), and earlier spies are only useful to gain science and military CS allies, both strongest as authority, so authority is making the best of bad situation. England could get Longbowman, it would be unlocked the same with UB a great offensive unit it needs. Ranged ship are bad, and indomitable is awful cause you have bonus moves to fire and hide between melee. America should be tradition, has early wonder mechanic, defensive UU with GA points, and unique hermitage that all about tradition, recent changes did little to change that.

Any bonus culture or GA is best with traditon and authority as fiinishing them ASAP is very powerful, while progress is better with science and production cause getting instant culture from building and tech, not affected by GA and doesnt give nothing important by later policies and finisher until later with writers so your earlier culture has lesser effects. Same with statecraft, you even get mechanism to avoid policies requireement as big boost for progress.

@Hokiefan00 , you can enable transparent diplomacy in options on new game. It will show you wats desirable modifiers are. I especially try to get you fought the same enemy when asked by AI to go to war, but it requires actual killing units, not just declaration. It’s something like 60 or 80.

 

Honestly I think there should be more Progress AI simply for diversity's sake, it's just not as fun when practically every peaceful civ is Tradition and every warmonger is Authority. And I agree it makes no real sense to have Japan and especially Greece be Tradition, a human can make them work but Authority just naturally plays better for a warmonger/expansionist AI.

Progress does have a bit of an issue in that while Tradition gives defense bonuses and can do well even if it doesn't expand alot, Progress AI's that don't expand quickly tend to suffer quite a bit and make easy targets. I think Progress AI should put a greater priority on quick expansion and buying units//walls when threatened.

At the end of the day whether AI's do Progress or Tradition better is not totally clear, but I think the best way to find out is have more AI's go Progress and then see what happens.

 

I think you underestimate how hard is to change AI policy choices. And AIs are right at the core choice. There's no simple switch. Simpliest way would be to... change UA as policy decision is complex and dependent mostly on UA. It's not simple to tell progress AIs when to do what again.

I would like to see some civs as progress if it's possible, I think diplo are the best to start so Morocco, Siam and India too. Pretty sure they can be beasts as progress, especially Morocco, not the ones suggested before. Progress is not cure for Celts, England, Japan, Shoshone uniques, they will be weak as AIs forever. Apart from civs, I wonder if its possible that AI can consider sea resources as progress weighing?

 

I only play on Prince level, so perhaps they are more balanced on lower levels. My latest game has 5 progress, 1 tradition, & myself (Mongols) & Sweden Authority. Aztecs are progress.

 

How to you know that? @Recursive could you co firm?

 

@CppMaster AIs analyze all or at least many policies choices same as build or tech everytime it takes any of those actions. There is no define to stick to any policy in the files. In vanilla it was assuming culture victorry, then modified by flavors and how close others are etc then choosing policies to victory goals. It is set to only choose the best choice at deity, so we see consistency that is not 100% if we run 100 autogames, but it makes calculations every time for a lot of things, like terrain or neighboors, that's why some people say ai is "set" to be one policy. It may appear that way but it isn't. If you don't want to read files, you can check it by easy making a small change to UA in sql, favoring different aproach. AI will automatically respond. It is also seen by rest of policy choices. You can see Carthage going both imperialism or rationalism, artistry or statecraft on deity for example. I've seen Germany taking one industry then rationalism, cause of trade routes. I believe UA, then UB and UU is the most important factor for first tree, that's why thhere's so much consistency in the first tree choice.

Changing Flavors doesnt seem to change much? But even slight change to ua can in my tests. It can be used it to "switch" Japan for authority on deity. Japan doesn't require big push, AI considersc Uu and UB already as weighing for authority.

Also there are personalities but I haven't tested them. I believe they wouldn't matter, as they seem to govern diplomacy only, probably also as one of the factors?

 

Hmm, I always thought that AI chooses a policy mostly based on it's personality and current situation, but it's not smart enough to analize it's UA.