Adding Policy Priorities

[EmperorPinguin]

The AI all pick the same policy trees regardless of preferred victory types, which looks odd to the single player human.

I here added the following policy preferences: Cultural: Tradition-Aesthetics-Freedom Scientific: Liberty-Rationalism-Order Diplomatic: Liberty-Patronage-Order Domination: Honor-Commerce-Autocracy I set Exploration as the 4th preferred policy for all, as the AI loves to settle coastal cities.

[SeventhM]

I'm not so sure Liberty-Rationalism-Order makes sense for scientific. Order makes sense, I guess, but Liberty+Rationalism is a bit off and Order+Rationalism often unnecessary. I'd imagine something like Tradition-Rationalism-Freedom or Piety-Rationalism-Freedom makes more sense

[EmperorPinguin]

According to Lekmod players, Liberty is reliably at least 5 turns faster science victory than Tradition, so I think it makes sense to go for that one. Rationalism seems like a logical follow-up tree, as there's a lot of science on it. I'm not entirely sure about Order, but it's the preferred ideology in competitive Lekmod for science victory, and the AI is not exactly great at using Freedom specialists.

[EmperorPinguin]

Commerce is quite often used for domination victories, especially with Autocracy bomber rushes, but not necessarily as a follow-up to Honor. A common type of domination victory is actually Tradition-Commerce/Piety-Autocracy for artillery rushes, but that doesn't match with the Domination AI play style.

[EmperorPinguin]

Diplomatic AI in unciv likes to play very wide, so Liberty and Order make sense, in addition to Patronage of course.

[SeventhM]

Hmm... Maybe you're right. Except I have a few problems with that viewpoint of it

1. When it was stared that Order is 5 turns faster, it was also somewhat clear that it's felt that, generally speaking, Order was just all around better than Freedom and Liberty was all around better than Tradition. If that's the context, I'm not entirely sure I trust the idea there that necessarily Freedom is worse at science victory than it is just Freedom is worse overall
2. Being faster by 5 turns isn't saying a whole lot tbh. If the argument was more that Workers Facilities + Communism + all of the production benefits was better for science victory than all of the benefits to specialists, then sure, I can buy it. But I'm not sure I buy in a vacuum that being just supposedly 5 turns faster is a big deal
3. The more important point is that if we're talking priorities, then we're talking AI civ behavior, not actually easiest way to achieve the victory. From this perspective, a lot of the Civs that wants wide play (Rome, America, Moors) are just Civs with generic benefits while the scientific Civs that wants tall (Babylon, Korea, Bulgaria) are also Civs that are a bit more... Obvious that they're science focused. And probably even more of the Civs with explicit science benefits are more Civs with more obvious tall benefits over wide. And if we can argue it's rough equal regardless, might as well focus it more on Tradition-Rationalism-Freedom

For that matter... Starting to think maybe Scientific should want Commerce more than Exploration just in case they can't get Rationalism in time

[EmperorPinguin]

I just ran a test with a few games, the lessons: 1) The AI doesn't build a spaceship, but dozens of inquisitors instead to block the way for spaceship parts. 2) It appears that Tradition-Rationalism-Freedom is approximately 10 turns slower than Liberty-Rationalism-Order.

The Scientific AI always builds enough cities (6-13 self-founded cities in my testing, on medium maps) to justify going Liberty over Tradition, and the AI is just terrible at using specialists. They're filling up maybe a third of the available slots (even with maxed Korea/Rationalism/Freedom bonuses (and half of those are merchants, not scientists)), I think that's why the science multipliers from Order are just more significant than the specialist-related bonuses from Freedom.

I get why Commerce is probably better than Exploration, I'll change that. Maybe too for Diplo and Culture civs?

[SeventhM]

2. It appears that Tradition-Rationalism-Freedom is approximately 10 turns slower than Liberty-Rationalism-Order.

Hmm... That's interesting. What's your test conditions?

For the record, I'm testing games with both by copying the scientific win condition multiple times. I need to do some more tests (and I want to do some stuff that puts merging this on hold for now), but it seems more to me that Tradition isn't mattering too much in terms of how fast a civ gets to spaceship parts. If anything, the actual advantage of liberty here seems to be they just have a larger military. Then again, this could just be the civs I'm testing with (Moors, Rome, and Germany was given Liberty, Bulgaria, Babylon, and Korea was given Tradition. I'll switch after the next test game I do). If Order has any serious advantage here, it's that the AIs seem to pick policies at random so it's more likely that the AIs pick a particularly useful policy with Order than with Freedom. And tbh, I don't want to base the priorities that way because then we might as well turn all of them to Liberty/Piety and none of them to Order. Also, yes, the AIs are pretty bad with specialists though observing what the governor does for players when set to balanced does show the governor does start to behave if you adopt Treaty Organization... Which the AIs aren't doing

For the record, If I understand the behaviors correctly, the only difference between scientific AI and normal AI is scientific AI wants science building a bit more. That's... it, so again, since this is more of a focus to define AI styles and not necessarily win conditions, I want to be more hesitant to just go "this wins more often, so this is their priorities". Also, even if you went with that mentality, I accidentally broke the culture AI so they are actually production heavy AI, not culture heavy AI, and likely wouldn't do as well at culture victory even if you explicitly set them to Aesthetics.

[SeventhM]

Ok, decided to swap which played which branch early, and I think I've come to a simple conclusion: it truly doesn't matter. Perhaps the issue is my choice of Civs to test with (Germany, Moors, and to a lesser extent Bulgaria are always the fastest 3 at most tasks). That said, I need more games, but I definitely feel like Bulgaria is worse on Liberty than on Tradition, and Germany still outpaced everyone on science despite some of the worst random choices on Freedom. Perhaps the issue stems from the AIs not using some mechanics well or at all, but I think I have my opinion on how things are going

I need more test games, and I am curious on your test conditions, but I might just prefer finding a way to get more preferred victory types than there are actual victory types over just taking your word that Liberty/Order is just better. Again, we're arguing AI and civ behavior, not necessarily optimal play

[SeventhM]

Rejecting now that https://github.com/yairm210/Unciv/pull/10939 is merged. I would rather tailor this to particular Civs rather than to particular wincons.