Why LINCOLN?! Is A Stronger Naval AI Than Ragnar: A Surprising Analysis

TheMapDownloade

Chieftain
Joined
May 10, 2011
Messages
64
Suppose you start a game on Terra and want to pick the AIs most likely to cross the sea and compete with you to colonize the New World. Which AIs should you pick?



Most people would intuitively pick the AIs with a sea-oriented unique unit or unique building (Willem, Hannibal, Ragnar, and Joao). However, this analysis is incomplete.

Becoming A Naval Power

In order to become a strong naval power, the AI must research Optics to unlock Caravels. So which AI is going to get to Optics first?

The AI uses a probabilistic method to decide what technologies to research. The game code defines seven Flavors for technologies (Culture, Gold, Growth, Military, Production, Religion, and Science). Technologies are assigned points in each of these Flavors. For example, Mining is valued at 8 Production, 5 Gold and 3 Science.

The AIs are assigned Flavors as well. Eight leaders have a single-minded emphasis on just one Flavor, such as Isabella (Religion). The other 44 leaders have one Major Flavor and a second Minor Flavor (for example, Willem is a Gold/Science leader).

When the AI has to choose what to research, the leader's values (10 for emphasis, 5 for major flavor, 2 for minor flavor) are multiplied by the tech's corresponding values (and if the tech has more than one matching value the two answers are added together). The tech with the highest total has a greater probability of being researched.


As an example: Hatsheput's flavors are Culture/Religion (aka Culture 5, Religion 2). Let's make her choose between researching Aesthetics and Mathematics. Here are their flavors from the game-code:

Aesthetics: 5 Culture, 2 Religion, 6 Science (a total of 13 points)
Mathematics: 10 Science, 1 Military, 1 Production, and 2 Gold (total: 14 points)

Mathematics has more total points, but Hatsheput only cares about Culture and Religion!

Hatsheput values Aesthetics at 5x5=25 Culture + 2x2=4 Religion = 29
Hatsheput values Mathematics at 0 because it has nothing to do with Culture or Religion!

While there is some randomness thrown in, the AI is likely to pick the highest-valued technology available. Therefore, Hatsheput usually techs Aesthetics in this scenario and ignores Mathematics.


Bee-Lining To Optics

So the question remains, which AIs are most likely to build Caravels and explore overseas? Well, here is a list of all the technologies needed to unlock Caravels (i.e. Optics and all of its prerequisites).



As you can see, not all techs are created equal. The average for ancient and medieval technologies is 18 total points. Some techs have much more, and some much less.

In addition there are technologies with hidden "Power" values. These values are there to make every AI leader prioritize these techs, regardless of their flavor preferences. That's because these techs are so crucial in unlocking important game features or military units. Here are the Power-rated technologies in the ancient and medieval eras, with Optics prerequisites in bold:



With this in mind, let's go back to our list of Optics technologies, and we can see that it is divisible into three groups.

First, Mining and Fishing are starting technologies.

Second come the technologies with high, diverse points or high hidden Power.

Lastly come the technologies with low overall points and low Power.



The limiting factor is really Sailing, Compass and Optics. Most of the AIs will quickly tech Iron Working and Machinery because of their flavor or their hidden Power, so we don't need to worry about that. It's Sailing, Compass and Optics that are standing in the way of the AI discovering the New World.

The value of Sailing, Compass and Optics is overwhelmingly concentrated in Science flavor. Therefore, only Science-flavored leaders have a chance to research Optics early on.

While these 3 techs have some value in other flavors, they're vastly outweighed by the "favored" techs in those flavors. For example, a Military-minded AI like Ragnar is going to look at the Military value of Optics (3) and still prefer the Military values of Horseback Riding (8) and Feudalism (5).

If we look at the 4 additional technologies a leader needs to go from Caravels to Galleons:



It is again obvious that these are techs with far-below-average point totals, and a value overwhelmingly concentrated in Science flavor.

AI: Which Science Guy?


Not only are techs not created equal, but AI flavors are not created equal as well. The playing field is HEAVILY tilted towards Military, and Science is actually the least popular flavor among all the AIs:



The AIs are a bunch of jocks! There is no overwhelmingly Science-focused leader and only 8 leaders place any value on Science at all. Here they are (and I've calculated the total value these AI leaders assign to Sailing + Compass + Optics in blue):



We have successfully narrowed it down to 4 AIs who place a high value on the Optics tree and begin with one or more prerequisite technologies:




Why Lincoln And Joao Rule The Seas

The Carrack: Joao has the ability to carry Settlers and military units in his Caravel UU, the Carrack.

Tech Trading: Peter is the most willing to trade techs with the human and the other AI leaders, followed by Joao and Lincoln. By contrast the notoriously isolationist Tokugawa is unlikely to trade at all.

Unit Spamming And Warring: Building large militaries and conducting wars slows down the AI's technological progression. Tokugawa and Peter are rabid warmongers, while Joao is neutral and Lincoln is inclined to peace.

For these reasons, Lincoln and Joao are the most likely AIs to discover a new continent.
 
I wish this analysis would have been made for which Civ goes first for Liberalism.

Or:

What are the chances of an AI building Oracle!
 
Oracle is pretty easy. It requires Priesthood and either Meditation or Polytheism. All three are exclusively Religion+Culture technologies so the top AIs are those that focus on Religion (Isabella) Culture (Hammurabi, Ghandi) or both (Hatsheput). You could also look at the Industrious leaders - since they build wonders twice as fast - and see which ones have a particular inclination towards Oracle. There are three: Huayna Capac starts with Mysticism. Louis XIV and Ramesses have Culture as their major flavor (but Military and Production respectively as their minor flavors, not Religion). So one of these 7 AIs is probably gonna build Oracle.

Liberalism is a lot harder because it is so far down the tech tree. However, if you look at the major prerequisites (Education, Philosophy, Paper, Theology, Civil Service, Drama, Code Of Laws) they tend to be Culture, Religion and Science technologies. So if I were you I'd look for a Financial or Philosophical leader with a flavor emphasis on Culture, Religion, or Science. These are:

Pacal II - FIN with Culture flavor
Suleiman - PHI with Culture flavor
Peter & Lincoln - PHI with Science flavor
Elizabeth - FIN/PHI with minor Culture flavor
Gandhi - PHI with high Culture flavor

In general though... your #1 threat for Liberalism is probably just going to be whoever is the #1 techer on your map for whatever reason. This analysis is a lot better for predicting early game behavior when all the civs are on a level playing field technologically.
 
Suppose you start a game on Terra and want to pick the AIs most likely to cross the sea and compete with you to colonize the New World.
AI's are pretty bad at large-scale post-Astronomy colonization in general, though.

Also, note that Joao is Imperialistic, which adds his colonization attempts.

Anyway, technical questions:

Hatsheput values Mathematics at 0 because it has nothing to do with Culture or Religion!
Isn't it: "gives the tech an additional value of 0"? I thought that the AI tries to evaluate actual benefits from a tech, and then adds the flavor numbers? If it's as you described (pure flavor + randomness) I am quite disappointed.

In addition there are technologies with hidden "Power" values. These values are there to make every AI leader prioritize these techs, regardless of their flavor preferences.
I thought these techs affect the Power rating, not AI preference, since there's already the generic AIWeight tag (0 for every tech in base game)?
 
I wish this analysis would have been made for which Civ goes first for Liberalism.

Or:

What are the chances of an AI building Oracle!

I've lost count of the times I've seen Isabella complete Oracle around 2500BC in Deity attempts.

Do not put her in your game if you are trying to Oracle something late like Civil Service.


Which civ usually techs Liberalism first?

In my experience, whichever civ techs Education 1st will almost always tech Liberalism as soon as possible afterwards.
 
I've lost count of the times I've seen Isabella complete Oracle around 2500BC in Deity attempts.

Do not put her in your game if you are trying to Oracle something late like Civil Service.

I can attest to this. I once lost Oracle to her in 1900 BC on NOBLE difficulty in which oracle usually goes out in ~600 BC.
 
Which civ usually techs Liberalism first?

In my experience, whichever civ techs Education 1st will almost always tech Liberalism as soon as possible afterwards.

If they already have Banking when they get Education (they often do,) they seem quite likely to go for Economics rather than Liberalism, as long as no-one has already got the free GM.

I'm wondering if the AI also has a preference for researching techs that lead to their UU, UB or favourite civic. I've noticed Justinian often goes for Aesthetics much earlier than most other AIs and guessed that it might be because it's on the path to Drama which unlocks his UB.
 
If they already have Banking when they get Education (they often do,) they seem quite likely to go for Economics rather than Liberalism, as long as no-one has already got the free GM.

I'm wondering if the AI also has a preference for researching techs that lead to their UU, UB or favourite civic. I've noticed Justinian often goes for Aesthetics much earlier than most other AIs and guessed that it might be because it's on the path to Drama which unlocks his UB.

I am pretty sure it is true. usually Hannibal is the first leader to HBR in my games, due to financial trait even earlier than Mongolia
 
I notice folks usually pick FDR when going after TGL, how often does the AI for his leader-head emphasize that wonder?
 
Great article with a fascinating insight.
 
I noticed Darius beelining Liberalism from quite early in my current game. I was semi-isolated with him, so all my espionage points were going against him and I could see his tech path. I traded him Code of Laws, and after that he went straight for it - Philosophy -> Civil Service -> Paper -> Education -> Lib.
 
I noticed Darius beelining Liberalism from quite early in my current game. I was semi-isolated with him, so all my espionage points were going against him and I could see his tech path. I traded him Code of Laws, and after that he went straight for it - Philosophy -> Civil Service -> Paper -> Education -> Lib.

Yep, I'm always concerned about Willem, Asoka, Elizabeth and Darius because they are pretty single-minded about getting to Free Religion. Of course, these AIs are strong techers in their own right.

Interestingly, they all like to try to build the Shwedagon Paya and run FR that way. Sometimes this seems to make them care a lot less about Liberalism since they are already in FR! :D

Honourable mentions also go to Hannibal and (of course) Mansa, who go after Economics for Free Trade and often pick up Liberalism while they're at it, since they have Education.

Obviously, the best techer on the map is still the number one threat, but in a much more haphazard way. Kind of like the way the number one techer can win a space victory out of sheer inertia. They just randomly tech stuff, build the projects in an aimless and unfocused sort of way, and then find themselves in Alpha Centuri.
 
Based on a thread of Seraiel's where she illustrated a "Paper block" http://forums.civfanatics.com/showthread.php?t=507901 (scroll down a bit or search for "paper-block"), AIs can't beeline straight to Lib. They can only "see" two techs ahead, so their research choices are influenced by the mechanic mentioned here, rather than a specific tech way down the tree.
 
I think that this post goes to show why an AI will never beat a human who understands what they are doing. The AI does things because they are told to do so by programming not because they understand why. They spam cottages not because they understand that this is the best way to get commerce and thus establish a lead. That is why I wish the game worked on a less rigid basis for techs AI researched. For example the AI will research a tech depending on whether certain conditions are met. For example if it starts with a gold mine in its initial fat cross it will research mining or if it has five or more flood plains, pottery would become a priority, rather than this permanent bias towards certain techs according to arbitrary values. Anyways thanks for the analysis. Next time I'm playing a Portugal game on Terra with Darius I'll make sure Lincoln tastes my axe before he reaches the new world. Or perhaps he should taste Monty's axe paid for by me.:)
 
I had Isabella build the Oracle on 2400 BC on Immortal and hate having her in my games now. A lot of people weren't / still aren't aware of how strongly she and some other civs strongly favor that tech line and wonder.

I used to think that it was Civs that start with Mysticism that get that upper hand, bit they also need the right bias too. Culture and Religious bias civs, and especially is they start with Musticism will beeline Oracle ASAP.
 
Top Bottom