Arabia

I agree with everything said by CrazyG and ElliotS. In fact there is a bigger problem. Policy trees are very tied to victory conditions. If you play for the victory - you are pretty much tied to certain choices and certain victory conditions based on what civ you play. Not going for Authority with Zulu is also a dumb move. Playing Tourism with Sweden is a dumb move too. If you play Tourism - you always go for Tradition (with one exclusion - Austria). If you play Diplomatic victory - you always go for Authority. If you play Scientific - you go Authority in 70% cases, exclusions are Korea, Arabia, India maybe some others
 
Not going for Authority with Zulu is also a dumb move. Playing Tourism with Sweden is a dumb move too. If you play Tourism - you always go for Tradition (with one exclusion - Austria). If you play Diplomatic victory - you always go for Authority. If you play Scientific - you go Authority in 70% cases, exclusions are Korea, Arabia, India maybe some others
These are gross oversimplifications
 
These are gross oversimplifications
Are they? I don't think so. How do you define what is optimal and what is not? The only objective measures here are number of techs and policies you have at a certain turn of the game and the turn you won the game. And i am absolutely sure that in any possible situation Authority Zulu will win earlier than Zulu with Progress and Tradition. And it will be Scientific Victory, not Tourism, not Diplomacy
 
The Arabia AI was doing really poorly in recent tests. If the bump is pushing them too far in the other direction we can look at it.
Assuming that Arabia can only be run on tradition is flawed.

In fact when writing my previous message i came up with idea how can all civs be balanced. It is actually called Scientific Method. Science rules! Requires some work though.

The idea is that when civs are tested on a random map with random neighbours and all other random stuff - there are a huge number of relationships and parameters that affect each other and it is impossible to determine what is the decisive factor. You can see the result (one civ wins more often than another), but the cause of such result can not be determined. If you want to understand why one civ wins more than another one - you need test as less parameter as possible.

First of all - you need to test every civ alone several times on a Duel Pangea (one pre-generated map, all civs, all starting policies, multiple tests for each civ on the same map). You need to record parameters several times in a game and check how many techs it have. This way you can compare how every civ behaves and how different policy choices and tech paths affect the game. This way you can define that one civ is stronger than another one on average.

After that you need to create bigger map and test it for every civ playing vs itself. For example Standard Pangea with 8 AIs all playing for India. Again every civ should be tested on one particular map and some parameters should be recorded several times, for example record number of techs and policies for every AI every 25 turns. This way you can define that this particular civ is stronger with some specific choices made (for example you will see that AI Arabia with Tradition will always win all other Arabias. Or maybe you will not)

After that several standard sets of neighbours should be defined and every civ should be tested on a pre-defined map with identical set of neighbours. This way you can see that some civs are very dependent on neighbours (for example, India always wins when it has peacefull neighbours and always loose if it has warmonger neighbours)

If such tests are performed - you can see why one civ is better than another and in what situations. This info will be enough to balance everything
 
Are they? I don't think so. How do you define what is optimal and what is not? The only objective measures here are number of techs and policies you have at a certain turn of the game and the turn you won the game. And i am absolutely sure that in any possible situation Authority Zulu will win earlier than Zulu with Progress and Tradition. And it will be Scientific Victory, not Tourism, not Diplomacy
Where to begin? Authority --> Science victory, what? Also what map?

On Pangea Authority is great just in general but not for other maps. Zulu actually is a good choice for tradition because you can spam high value CS tributes for big hammers and food in your capital, and if you are on a smaller land mass you won't have as much combat. Frankly either way, I wouldn't want every warmonger going authority every game, it really would kill variety.

PS- I wouldn't tell G how to his job/hobby
 
If someone helps me find where the AI personality stuff is located I'll take a look, but if I recall correctly all leaders have a great person emphasis, and Arabia's should be ten.

Where to find the flavors values:

C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid
C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\DLC\Expansion\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid (same file as above)
C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\DLC\Expansion2\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid (just some changes to trade route flavour)


https://forums.civfanatics.com/resources/leaders-spreadsheet.21465/ here a nicely readable format.

<LeaderType>LEADER_HARUN_AL_RASHID</LeaderType>
<FlavorType>FLAVOR_GREAT_PEOPLE</FlavorType>
<Flavor>6</Flavor>

Harun is sitting at average GP flavor like pretty much everybody else. The only civs moving away from the 5-6 middle value are: Zulus (3), Babylon, Venice, Austria (7), Korea (8), Sweden (9 - lol).

With the exception of Sweden (no more Nobel Prize UA), it turns out that those flavor value makes sense even under CPB changes, but altering flavor values to better reflect CPB different UAs should be in order.
 
Last edited:
Zulu actually is a good choice for tradition because you can spam high value CS tributes for big hammers and food in your capital
Any civ can be good in any situation. Every policy, tech or building give something and non of them has negative effect. Question is would you win faster with Tradition Zulu than with Authority Zulu? And i'm confident that the answer is no on any map, even on Archipelago. Same with Arabia, but Arabia wants Tradition. Same with many other civs

PS- I wouldn't tell G how to his job/hobby
This was an advise, not an order and only with the best intentions. I feel like my posts look quite categorical and indisputable sometimes. They are actually not, it is probably due to the fact that English is my 2nd language
 
<LeaderType>LEADER_HARUN_AL_RASHID</LeaderType>
<FlavorType>FLAVOR_GREAT_PEOPLE</FlavorType>
<Flavor>6</Flavor>

Harun is sitting at average GP flavor like pretty much everybody else. The only civs moving away from the 5-6 middle value are: Zulus (3), Babylon, Venice, Austria (7), Korea (8), Sweden (9 - lol).

With the exception of Sweden (no more Nobel Prize UA), it turns out that those flavor value makes sense even under CPB changes, but altering flavor values to better reflect CPB different UAs should be in order.
Great find, thank you.

Would a higher great person bias cause a civ to prefer tradition? If so, just swap Arabia and Sweden, and the issues should mostly resolve themselve without issue (probably good for sweden too)
 
I agree with everything said by CrazyG and ElliotS. In fact there is a bigger problem. Policy trees are very tied to victory conditions. If you play for the victory - you are pretty much tied to certain choices and certain victory conditions based on what civ you play. Not going for Authority with Zulu is also a dumb move. Playing Tourism with Sweden is a dumb move too. If you play Tourism - you always go for Tradition (with one exclusion - Austria). If you play Diplomatic victory - you always go for Authority. If you play Scientific - you go Authority in 70% cases, exclusions are Korea, Arabia, India maybe some others
Progress is viable on every warmonger, and better in many cases. If you're not going to be fighting until medieval then progress will be better in 90% of cases. Progress Zulu for example are very strong. They lose the +15 heal and +10 CS from authority, but gain a much more resilient economy and can still wage war easily because of their unique bonuses.

Also progress or tradition are obviously best for science victory.

When I said Arabia is an exception I meant it. CrazyG and I are on the same page here I think.
If Arabia can do well without choosing Tradition -- and it does -- then it is not downright stupid, shooting itself in the foot, or engaging in piss-poor gameplay. It is just playing sub-optimally in those particular games. I'll accept that as the price for unpredictability any day.

I think hardcoding a single civ for a single difficulty level is... something to be done in the privacy of your own computer.
I was saying that Deity as a minimum is supposed to always make the best choice, and tradition is 100% the best choice. There are a million random and varying things in every game, Arabia always going tradition changes practically nothing on the variety front.

Arabia doing well without Tradition means it's wildly OP with tradition. Which it is. I've never had such easy victories. You're just 100% wrong on this issue imo.

Where to find the flavors values:

C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid
C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\DLC\Expansion\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid (same file as above)
C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\DLC\Expansion2\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid (just some changes to trade route flavour)


https://forums.civfanatics.com/resources/leaders-spreadsheet.21465/ here a nicely readable format.

<LeaderType>LEADER_HARUN_AL_RASHID</LeaderType>
<FlavorType>FLAVOR_GREAT_PEOPLE</FlavorType>
<Flavor>6</Flavor>

Harun is sitting at average GP flavor like pretty much everybody else. The only civs moving away from the 5-6 middle value are: Zulus (3), Babylon, Venice, Austria (7), Korea (8), Sweden (9 - lol).

With the exception of Sweden (no more Nobel Prize UA), it turns out that those flavor value makes sense even under CPB changes, but altering flavor values to better reflect CPB different UAs should be in order.

Great job! Looking at it I think Arabia thinks it's a science-victory orientated civ! Arabia should have a great person focus of 12 (like Gandhi for nukes) and a focus on cultural victories, not scientific. They should also be made a bit friendlier and more looking for trade.
 
Where to find the flavors values:

C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid
C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\DLC\Expansion\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid (same file as above)
C:\Steam\steamapps\common\Sid Meier's Civilization V\Assets\DLC\Expansion2\Gameplay\XML\Leaders\CIV5Leader_HarunAlRashid (just some changes to trade route flavour)


https://forums.civfanatics.com/resources/leaders-spreadsheet.21465/ here a nicely readable format.

<LeaderType>LEADER_HARUN_AL_RASHID</LeaderType>
<FlavorType>FLAVOR_GREAT_PEOPLE</FlavorType>
<Flavor>6</Flavor>

Harun is sitting at average GP flavor like pretty much everybody else. The only civs moving away from the 5-6 middle value are: Zulus (3), Babylon, Venice, Austria (7), Korea (8), Sweden (9 - lol).

With the exception of Sweden (no more Nobel Prize UA), it turns out that those flavor value makes sense even under CPB changes, but altering flavor values to better reflect CPB different UAs should be in order.

Ehm, correct me if I'm wrong, but those you listed are vanilla values, and a lot of flavors changed for VP, in fact if I remember it correctly someone actually went through all AI flavors and changed them up to function better.
 
But Arabia is losing in those games right? A loss is a loss.

So far no one seems to challenge the idea that Arabia is much better with tradition. I really think the Arabian UA needs a nerf, otherwise he is too powerful on tradition. An extra 2 gold per historic event, on top of what was already a borderline broken unique ability? I think this is a false dichotomy, if Arabia was say twice as likely to take tradition as he is now, would that really make the game so much less predictable as to decrease your enjoyment?

Also you are essentially advocating intentionally poor AI decisions. I'm not saying that hard-coding tradition is appealing, I think either path has a pill to swallow

Losing isn't the same as performing poorly; they perform very well overall. Again, I see a difference betwen "poor" and "sub-optimal."

But as I said somewhere above, I have no problem with nerfing the UA (and buffing elsewhere) to address yours and ElliotS's issue. All I'd keep in mind is that it shouldn't make Arabia much stronger overall.
 
I couldn't disagree more. The AI being 'unpredictable' is only fun if it's still smart.

If the AI decided to mix it up a bit and start feeding it's units one-by-one into your cities while embarked it would certaintly be random, and considered a bug or terrible design. I don't like randomness that means the AI shoots itself in the foot. Choosing different religious beliefs? Sure. Choosing Industry over Rationalism for a science victory? That's debatable. Going anything but Tradition as Arabia? That's just downright stupid.

If I wanted AIs to make random dumb choices that lead to piss-poor gameplay I'd just uninstall VP. If we're going to make the AI smart and competitive I think we need to avoid leaving in terrible AI choices for 'muh randomness'!

I also can't see any other civs as restricted as Arabia. Tradition Aztecs can be really good, Authority Venice is downright menacing and many classic warmonger civs can perform very well as progress. (I've gone Progress Denmark before and it was the right choice.)

I urge G to hardcode choosing tradition, at least for Deity.

You and I both know you are using an absurdist conclusion (AI feeding units to city one by one) to push an argument. That’s not an effective use of our time.

G
 
But Arabia is losing in those games right? A loss is a loss.

So far no one seems to challenge the idea that Arabia is much better with tradition. I really think the Arabian UA needs a nerf, otherwise he is too powerful on tradition. An extra 2 gold per historic event, on top of what was already a borderline broken unique ability? I think this is a false dichotomy, if Arabia was say twice as likely to take tradition as he is now, would that really make the game so much less predictable as to decrease your enjoyment?

Also you are essentially advocating intentionally poor AI decisions. I'm not saying that hard-coding tradition is appealing, I think either path has a pill to swallow

I will increase the chance for civs like Arabia to take tradition, but I won’t force it.

I can also move the gold off the UA and to the UB if desired. I still don’t see Arabia run away ever, so I personally think it is not needed
 
Ehm, correct me if I'm wrong, but those you listed are vanilla values, and a lot of flavors changed for VP, in fact if I remember it correctly someone actually went through all AI flavors and changed them up to function better.

Yeah those are not current values.

G
 
If you hard code that, and the UA changes later, you'll need to hard code it again. That's why it feels wrong. And I agree. You need the scripts to decide for themselves when something works best for the current AI, not telling a civ to play in a specific, approved by humans, way. If they don't work properly, fix them if you can, unless it's too much CPU load.

I don't quite understand what G means by AI not having enough data to make informed decisions when picking first tree. Doesn't AI look into the tree for future synergies? If it has an UA enhancing great people and tradition has 4 policies related, more or less, with great people, then it should put higher the priority on tradition, with a small open window in case Authority is the best response to aggressive neighbours.
If AI can only look at the immediate next policy, then I can see the problem. But G said that AI have a long term strategy (sometimes picked a bit randomly, sometimes it may change its priorities, but there's one), so there must be some long term logic that look at the whole tradition tree and find sufficient synergies to increase the chance of picking the tree.
 
If you hard code that, and the UA changes later, you'll need to hard code it again. That's why it feels wrong. And I agree. You need the scripts to decide for themselves when something works best for the current AI, not telling a civ to play in a specific, approved by humans, way. If they don't work properly, fix them if you can, unless it's too much CPU load.

I don't quite understand what G means by AI not having enough data to make informed decisions when picking first tree. Doesn't AI look into the tree for future synergies? If it has an UA enhancing great people and tradition has 4 policies related, more or less, with great people, then it should put higher the priority on tradition, with a small open window in case Authority is the best response to aggressive neighbours.
If AI can only look at the immediate next policy, then I can see the problem. But G said that AI have a long term strategy (sometimes picked a bit randomly, sometimes it may change its priorities, but there's one), so there must be some long term logic that look at the whole tradition tree and find sufficient synergies to increase the chance of picking the tree.

If you want to be the first person to create reliable long term logic for video game AI please, by all means, I would appreciate it. :)
 
If you want to be the first person to create reliable long term logic for video game AI please, by all means, I would appreciate it. :)
:goodjob: It would be cool, ain't it?

Right now they need generative algorithms to achieve such a complex thing. Plus, they don't work well in my machine :D. But there are workarounds. I don't evaluate policy by policy what can work best for me, but I sum up all the policies in a tree, compare that with my current situation and uniques, label them in a few tags (building, fighting, great people, religion, capital focus, seafaring, trading, golden ages, wltkd), and the tree that matches better is my choice pick. Joining together a few of the same tags leads to synergies.
Anyway, what I'm trying to say is that if you can improve AI logic within existing code, that would be the best approach. You surely have an idea why Arabia doesn't usually go tradition, when the synergies are obvious.

You could manually increase the likeliness of Arabia taking tradition, but if you ever change its UA, then you will need to manually decrease it again. Such things make more sense in personalities, if there is a personality that favours going GP.
 
The only other changes I could find to leaders flavors are in ...\MODS\(2) Community Balance Overhaul\Balance Changes\AI\Flavors\LeaderFlavors where FLAVOR_GREAT_PEOPLE has been set to 10 for: Pacal, Pedro, Sejong, Wu Zetian.

Anything else?
 
I can also move the gold off the UA and to the UB if desired. I still don’t see Arabia run away ever, so I personally think it is not needed
Want one of my saves?

I understand you are really pushing some fine tuning among AI performance, but while doing that you can't make a civ too strong in human hands. Also Arabia should have a 10 for great people, but Wu Zetian probably doesn't need it to be that high
 
Want one of my saves?

I understand you are really pushing some fine tuning among AI performance, but while doing that you can't make a civ too strong in human hands. Also Arabia should have a 10 for great people, but Wu Zetian probably doesn't need it to be that high

I'm not saying that Arabia can't run away, just that it isn't a commonality at all for me. But that's why I need all of you to report things! :)

I'm going to try shifting 1 of the gold to the UB and remove the gold from the UA.

I've also spent some time over the past few weeks chipping away at a new AI policy selection modeler- we'll see if it works better.

G
 
Top Bottom