(Testing) Leader Personalities Idea

[meRlinX_AT]

That might actually be a good idea since certain wonders are more useful than others.

FLAVOR_WONDER as far as I can tell is a global wonder flavor that will encourage civs to build them just because they're Wonders. That's why people are noticing Wonder spam coming from certain civs, either Nationalists or Federalists.

i will see all civs are building wonders, if the need them

As far as growth vs. expansion it's more about making sure the AI considers their surroundings. Ex. if they are wedged between multiple CS and rivals and can only expand once, then it's more likely they'll go Tall than wide. If there's land available and their production is strong, they'll settler spam and try to gobble it up. That can backfire, of course, and it should if they did it with a bunch of rivals around them.

this behaviour is good, but a little tuning is needed.

I think the biggest problem with this right now is the vicious circle of low military flavor priority making barbarians more damaging until later on when they aren't a threat anymore, or high military flavor priority choking their economies to death.

I'm personally favored toward the former, because if economy is strong then they can lose a settler or two and still get their 3 or 4 cities in an isolated start, whereas the latter will just choke their economy and make One City Civs that are still in the Classical Era while a player is just getting into Renaissance.

1 of the main problem from the AI is - the did not protect the workers/settlers (and did not try to get them back)
... and the did not hunt/kill barbarians in there own land.
but its nice to see the AI repairs the plundered improvements over and over again

 

[jma22tb]

I can agree with Wonder flavors being allocated based on what they're used for.

The barbarians gobbling civilians is more of an early problem than later in the game, but it's still an important problem to solve because feedback suggests that some players are running away with the game to the point where they don't think the AI is competitive. At least I'm hypothesizing this as the primary cause because their feedback basically amounts to "uh fix this" without providing supporting data lol.

I really don't know how to solve that problem without going into a vicious cycle of broke One City Civs vs. Pushover Start Civs. If there was a way to tell the AI to have at least a Ranged and Melee unit for each city that would probably fix everything but I have no idea how to do that.

 

[Gazebo]

I can agree with Wonder flavors being allocated based on what they're used for.

The barbarians gobbling civilians is more of an early problem than later in the game, but it's still an important problem to solve because feedback suggests that some players are running away with the game to the point where they don't think the AI is competitive. At least I'm hypothesizing this as the primary cause because their feedback basically amounts to "uh fix this" without providing supporting data lol.

I really don't know how to solve that problem without going into a vicious cycle of broke One City Civs vs. Pushover Start Civs. If there was a way to tell the AI to have at least a Ranged and Melee unit for each city that would probably fix everything but I have no idea how to do that.

The barbarian issue is a dll one. I'm working on it.

G

 

[jma22tb]

Oh that explains it. DLL is waaaaay beyond what I can handle, so I'll leave this to a professional haha.

/cheer to Gazebo

 

[jma22tb]

Updated the flavors and personalities. Details in the OP. Both tests were pretty impressive. Settler AI is awesome for protection, but they do settle pretty far off from the capital sometimes. That said, it will certainly ramp up the tension.

 

[vyyt]

I really don't know how to solve that problem without going into a vicious cycle of broke One City Civs vs. Pushover Start Civs.

First of all, thanks a lot for your efforts and trying to juggle the civ flavors into a reasonable balance. I really enjoyed reading the "play-through" descriptions of your early testing games and I think the flavor changes look promising.

I played with the first iterations in the beginning and mid-May. I was writing down notes, but when I was ready to post them, there were already several people reporting the same things and I soon found myself being couple of versions behind. If you are still interested in the notes for the old versions, I can PM them to you, but I think they are worthless now, as the flavors changed a lot since then. The only thing that had not been reported by others was that broke-one-city-civs in my games were falling behind in tech greatly and according to InfoAddict, they had 0 science beakers per turn - I suppose this was caused by their very negative happiness and GPT?

Anyway, while reading the reports from other players and your remark quoted above, I thought maybe we could have both - civs with strong military flavor and not going broke at the same time. I know people usually hate the "cheating AI", but wouldn't it solve your problem and also make the game more challenging/fun if unit maintenace was reduced for AI by - say 50%? Both military flavor and maintenance reduction could scale with difficulty.

And if AIs start producing crazy numbers of units, then there could also be a cap on number of units which would scale with difficulty and number of cities - I think there already is something called unit support limit (I have never reached it), so maybe it could be tweaked/hooked/used in this way ... ?

I remember having epic battles with AI in March and April (Emperor, Epic, huge crowded Communitas) when AIs were swarming me with huge armies, forcing my smaller elite forces to tactical retreats, healing, regrouping and then executing concentrated pincer attacks which was a great fun. I have not had a chance to play with the latest changes yet, but it would be a shame if we had to choose between militarily strong, but broke and technologically backward AI or successful, but militarilly harmless AI ...

Would do you guys think?

 

[jma22tb]

Things are coming together and the AI is improving. To put it in perspective, the sheer volume of changes done in CBP and CP have made a new game.

It's frustrating to see progress made and then reality setting in about how much it actually was instead of what it looked like from a few perspectives in testing.

We'll get it

 

[WileyNg]

Jma,


Is there any particular flavour tweaked concerning the vassal system?

Say Federalists are more likely to become a vassal while nationalists, though ignorant of diplomacy, take a liking to create vassals of their own and shore up their influence in the region.

 

[sc369]

This screenshot is from my latest game (Germany Standard Standard Deity Pangaea), using 6-2 version with flavors from this thread. I had a slow start (my capital was completely surrounded by forest with only sugar and copper resources), but I still expanded fast compared to the AI. I have just conquered Poland on turn 114. From here on it will be easy to win the game.

GOOD

AI built Stonehenge early, beating me by 3 turns
AI also beat me to Petra
AI had a good mix of melee and ranged units, and had a reasonable number of them.
AI settled second cities fairly early, at the same time as me. (around turn 35 for most civs)
AI kept pace with me in tech
AI Zulus and Poland declared war on me together around turn 75, probably because I settled too many cities and ignored Poland when they asked me to stop settling near them.

BAD

AI settled third and fourth cities too late. I settled 3rd city turn 42 and fourth city turn 49, at which point most civs were still on 2 cities. Poland never settled a 3rd city.

There were two great city spots near Poland, one with Mt Sinai and one with Mt Kilimanjaro, both with nice river locations and resources. I beat him to both of them - he really should have got at least one.

Zulu AI declared war on me but never sent a single unit my way

Poland AI attacked me but divided his forces between two cities. If he had concentrated them all in one place he might have taken one of them. At most 2 units attacked each city at the same time, while he probably needed at least 4 to be successful.

Poland AI did not show any signs of preparing his forces to invade. It was several turns after he declared war before any of his units entered my territory. I had time to rush walls, after which he had no hope of capturing a city.

Poland AI let his archers die easily away from his spearmen. He would also send only one unit at a time within range of my city and archer, letting me slowly shoot him to death.

OVERALL COMMENTS

Poland needed either to settle more cities and faster, or he needed to declare war earlier (maybe around turn 50), having focused solely on military. If the AI builds a lot of units and attacks the human before walls are up he has a chance. Once walls are up it's very hard for the AI to win.

I went full economic strategy up till about turn 65, settling 6 cities, then went full military, and the AI didn't punish me until it was too late.

It seems that you are not giving the AI a free settler on Deity. I think that for the AI to be competitive against a human, he needs the free settler to get an economic head start.

 

[Gazebo]

OVERALL COMMENTS

Poland needed either to settle more cities and faster, or he needed to declare war earlier (maybe around turn 50), having focused solely on military. If the AI builds a lot of units and attacks the human before walls are up he has a chance. Once walls are up it's very hard for the AI to win.

I went full economic strategy up till about turn 65, settling 6 cities, then went full military, and the AI didn't punish me until it was too late.

It seems that you are not giving the AI a free settler on Deity. I think that for the AI to be competitive against a human, he needs the free settler to get an economic head start.

First off, I'd like to reiterate this point regarding the AI: one game, or one scenario, are not (individually) good indicators of AI performance. There are so many functions and calculations behind the AI that it is impossible to say that 'x is the problem, and y is the solution' after one game. I don't doubt that you are a great player, however I do believe that more testing is in order before we decide what the issue is.

Second, the AI will never think like a human. You noted that the AI 'should have reached out for one of those natural wonders,' however the AI was most likely mobilizing for war, a strategy that reduces the AI's desire to expand (as they don't want to build a city just before they go to war).

As far as the free settler is concerned, I'd like to get the early game AI in a place where that is unnecessary. I hate those early game cheese buffs.

G

 

[Gazebo]

Also, jma, you've got a database error in your latest leader values:

[8668.835] near "WHERE": syntax error

there's a missing value for theodora in the major_civ_deceptive slot.

G

 

[Finarvi]

[...]

Anyway, while reading the reports from other players and your remark quoted above, I thought maybe we could have both - civs with strong military flavor and not going broke at the same time. I know people usually hate the "cheating AI", but wouldn't it solve your problem and also make the game more challenging/fun if unit maintenace was reduced for AI by - say 50%? Both military flavor and maintenance reduction could scale with difficulty.

And if AIs start producing crazy numbers of units, then there could also be a cap on number of units which would scale with difficulty and number of cities - I think there already is something called unit support limit (I have never reached it), so maybe it could be tweaked/hooked/used in this way ... ?

I remember having epic battles with AI in March and April (Emperor, Epic, huge crowded Communitas) when AIs were swarming me with huge armies, forcing my smaller elite forces to tactical retreats, healing, regrouping and then executing concentrated pincer attacks which was a great fun. I have not had a chance to play with the latest changes yet, but it would be a shame if we had to choose between militarily strong, but broke and technologically backward AI or successful, but militarilly harmless AI ...

Would do you guys think?

I completely disagree. Firstly, AI don't need more cheating. If you feel like it's too easy for you, try playing higher difficulty or give yourself some handicaps. Secondly, AI spams so huge army anyway that fighting with it may become boring at some point. Sure, leveling units and waging war for half of game is cool but I'm sure I'd get bored after few games. And it takes lots of time. I don't wan't AI to be allowed even larger army

 

[jma22tb]

Updated the files to fix that error and moved Expansion to 9. Took the CoreDefines off of the RAR because the changes I added to it are already in anyway. Also made all the 0 flavors 2 for Personalities because I don't know if that adversely affects diplomacy. If it does then that issue is alleviated with lower priority, yet still present capacities for those lower priority behaviors.

Since Growth is valued high enough to ensure cities will build Granaries, Aqueducts, etc. I want them all to aggressively pursue the land grab early phase. Hopefully that alleviates the problem sc369 mentions about it being too easy to gobble up land and passing up opportunities. Since the new fix addressed settlers going too far off, hopefully we'll see civs expanding to 3 or more cities in the first 100 turns as a standard, potentially 4.

Something I want to recommend also is adding more City-States to your maps. The tension between major civs should increase if there is less space overall and Nationalists/Imperialists have to conquer CS, pissing off the Federalists and Hegemony powers.

 

[sc369]

First off, I'd like to reiterate this point regarding the AI: one game, or one scenario, are not (individually) good indicators of AI performance. There are so many functions and calculations behind the AI that it is impossible to say that 'x is the problem, and y is the solution' after one game. I don't doubt that you are a great player, however I do believe that more testing is in order before we decide what the issue is.

Second, the AI will never think like a human. You noted that the AI 'should have reached out for one of those natural wonders,' however the AI was most likely mobilizing for war, a strategy that reduces the AI's desire to expand (as they don't want to build a city just before they go to war).

As far as the free settler is concerned, I'd like to get the early game AI in a place where that is unnecessary. I hate those early game cheese buffs.

G

I don't consider myself a great player. On vanilla, I can win maybe 25% of the time on Deity. I'm certainly not in the Marbozir league. (Incidentally, everyone should watch his videos if they want to improve. It never occurred to me before to chop down forests to create better fields of fire for my ranged units!)

As far as feedback is concerned, jma was saying that players weren't giving specific enough feedback, so I was trying to give some definite ideas based on one game. I would say that, in my experience, the AI settles its 3rd and 4th cities too late nearly every game that I've played, although settlement has definitely improved since previous versions. If the AI is not going to settle early, I feel it needs to declare war earlier, perhaps with an opportunistic theft of a worker or settler. (I admit I'm very ignorant about the way the AI actually works).

I understand that your mod philosophy is not to buff the AI too much in the early game, and I appreciate that: it's certainly fun to build early wonders or found a religion, which is often impossible in Deity vanilla. I'm just not sure that however good the AI is, it will ever be able to match a human without buffs. If the buffs are delayed until later in the game, the human may be in an unassailable position.

In any case, the mod is very enjoyable and it would be hard to go back to vanilla. The increased yields are more satisfying, and I especially like the new farm rules, which, combined with trying to get villages on connection routes, really make you think about where to chop and which tiles to buy. Figuring out optimum city placement has never been so fun! Many thanks for all your work.

 

[sc369]

Since Growth is valued high enough to ensure cities will build Granaries, Aqueducts, etc. I want them all to aggressively pursue the land grab early phase. Hopefully that alleviates the problem sc369 mentions about it being too easy to gobble up land and passing up opportunities. Since the new fix addressed settlers going too far off, hopefully we'll see civs expanding to 3 or more cities in the first 100 turns as a standard, potentially 4.

If civs are going Liberty, I suggest that 6 cities by turn 100 is a good number to get the AI to aim for. More cities means more free production, more gold from pop growth, and the extra culture from buildings more than covers the increase in policy cost. All the excess gold can then be used to purchase or upgrade military units to defend the cities.

Speaking of pursuing land grab, is it possible to teach the AI to change its cities to production focus when building settlers? Excess food is converted very inefficiently to the production of settlers.

 

[jma22tb]

If civs are going Liberty, I suggest that 6 cities by turn 100 is a good number to get the AI to aim for. More cities means more free production, more gold from pop growth, and the extra culture from buildings more than covers the increase in policy cost. All the excess gold can then be used to purchase or upgrade military units to defend the cities.

Speaking of pursuing land grab, is it possible to teach the AI to change its cities to production focus when building settlers? Excess food is converted very inefficiently to the production of settlers.

Production is at 9 as well. Growth is valued below both. Also it's just an estimate, if Expansion is ranked highest then it's likely they'll grab whatever they can.

 

[jma22tb]

Jma,


Is there any particular flavour tweaked concerning the vassal system?

Say Federalists are more likely to become a vassal while nationalists, though ignorant of diplomacy, take a liking to create vassals of their own and shore up their influence in the region.

I'm not sure how to make that happen to be honest. The diplomacy interactions are something I haven't dug into that much.

 

[sc369]

So my latest game is as the Netherlands, Deity Standard Pangaea, using the most recent flavors from 6-3. Overall I thought this was the best performance since 5-15.

There have been no wars so far. On turn 100, most AI's have settled quite well. Greece, Ottomans and Shoshone have 4 cities. Shoshone also managed to build 3 wonders along with them. Germany has 3 cities and Spain has 2.

Spain built Stonehenge early and then finished Great Library on turn 65. However, they had still not built a settler. They didn't settle their second city until turn 80.

Only two religions have been founded, by me and Spain.

The tech race is pretty even. All civs are between 15 and 17 techs. I am on 16.

AIs did well building wonders, leaving me with only Zeus and Artemis.

Culture is pretty even. I have 5 policies, and all AIs have either 4 or 5.

However, I am well ahead in the score charts, with 7 cities and 36 population. My closest competitor is Shoshone with 4 cities and 25 pop.

 

[jma22tb]

Yeah I'm seeing a lot better performance too. Economic growth is great, military power progresses, trade routes go up fast, policies are looking good, and Wonders are fought over hard.

That said, the economic growth and capability of a civ does improve the more cities they have. The Spain situation sounds like a fair exchange of two Wonders for expansion but their economy will definitely not be as good as yours.

Perhaps Expansion can move to 10 in its own tier. If the AI is this capable of handling its business economically, then indefinite expansion might actually be feasible on their end. If they have no space, they'll just stop and I'm pretty sure there are failsafes in the AI economic management if their happiness tanks hard.

I'll upload a test RAR with Expansion at 10. Let's see how crazy the Settler Spam gets lol.

 

[vyyt]

I completely disagree. Firstly, AI don't need more cheating. If you feel like it's too easy for you, try playing higher difficulty or give yourself some handicaps. Secondly, AI spams so huge army anyway that fighting with it may become boring at some point. Sure, leveling units and waging war for half of game is cool but I'm sure I'd get bored after few games. And it takes lots of time. I don't wan't AI to be allowed even larger army

I am not a big fan of cheating AI either, I was merely commenting on that jma's remark about the difficulty of heaving both prosperous and militarily strong civs at the same time. As I stressed out, I have not had a chance to play the latest versions, so maybe all is good now and my comment is pointless. But I saw people still reporting that AIs declare war and then just sit with their few units around their cities waiting to be destroyed. If it is not the case anymore, then great and ignore my comment. But IF and AFTER trying all the options we come to the conclusion that by tweaking the flavor values we either get militarily strong and broke/backward civs or prosperous but militarily weak civs, then I would rather have them cheat or go back go the version before the new flavors.

Anyway I do not want to sound like a downer, I do understand this new flavor system is still being tested and tweaked and I do appreciate the efforts to make the AI civs play smarter greatly. I will play with the newer versions when I have time see if AIs will be more active.

By the way at what difficulty do AIs spam huge armies for you now?