How does AI advance so fast in science?

[Delgar]

So I'm playing the latest release v137 and wow, the AI is still just flying by in tech advancing. They don't have as much gold which is good, but they are just advancing so fast. is there anything I can try to change this? I'm only playing on normal difficulty.

How many techs ahead of you is the leader?

 

[apocalypse105]

How many techs ahead of you is the leader?

As I have the Change to say this VEM mod the AI feels like it cheating even more then vanilla.
So much it doesn't feel fair anymore.

I currently play my games on normal difficulty prince ANd I lost all of them because the AI is crazy with science entering the modern era and building spaceships.

Or when I am 10 turns away from the utopia project in the industrial era suddenly siam completes the utopia project themself with 5cities


I don't mind the more difficult AI but this is madness and more of all it takes the fun away.

Some bonusses olso make me mad extra sight for combat units? Really shouldn't the AI fallow the same rules? Whats the use of setting up you're artillery and knowing he can't see me because of the fog of war then?
I don't mind the promotions it makes combat a little bit harder

 

[Txurce]

So I'm playing the latest release v137 and wow, the AI is still just flying by in tech advancing. They don't have as much gold which is good, but they are just advancing so fast. is there anything I can try to change this? I'm only playing on normal difficulty.

They're advancing so fast because they're spending their gold. On another thread we're discussing nerfing the AI's production and science bonuses.

As I have the Change to say this VEM mod the AI feels like it cheating even more then vanilla.
So much it doesn't feel fair anymore.

I don't mind the more difficult AI but this is madness and more of all it takes the fun away.

Some bonusses olso make me mad extra sight for combat units? Really shouldn't the AI fallow the same rules? Whats the use of setting up you're artillery and knowing he can't see me because of the fog of war then?
I don't mind the promotions it makes combat a little bit harder

Explicit bonuses aren't cheating. (I had to say that.)

However, the gold-spending adjustment was like a snake swallowing an elephant. Since we did it one bite rather than several smaller ones (no visuals, please), we now need to make some adjustments. I think some have been made already, and I'm sure more are coming.

The AI has had extra sight for a long time. The goal here was to give the AI more opportunity to not be so stupid. I think it's helped, although obviously not enough. It shouldn't hurt artillery performance. Even if the AI can see you, their cities and city defenders can't reach you. Other units can, of course, but I assume you usually have one melee unit to serve as a shield.

 

[Zaldron]

As I have the Change to say this VEM mod the AI feels like it cheating even more then vanilla.
So much it doesn't feel fair anymore.

I currently play my games on normal difficulty prince ANd I lost all of them because the AI is crazy with science entering the modern era and building spaceships.

Or when I am 10 turns away from the utopia project in the industrial era suddenly siam completes the utopia project themself with 5cities


I don't mind the more difficult AI but this is madness and more of all it takes the fun away.

Some bonusses olso make me mad extra sight for combat units? Really shouldn't the AI fallow the same rules? Whats the use of setting up you're artillery and knowing he can't see me because of the fog of war then?
I don't mind the promotions it makes combat a little bit harder

The mod changes a *lot* of things. It fundamentally changes the way you have to play in a variety of ways. It makes some things that are useless in vanilla actually useful to use. And until you get adjusted to the changes it makes the game harder.

The biggest recent change is that the AI will actually use its surplus gold in mostly intelligent ways. In the v136 game I'm currently playing the AI is battling it out with me on CSes where I have never seen it compete to this extent before. I like the added challenge although other nerfs may be needed to compensate. Which particular feature(s) make you feel like the AI is cheating? We might be able to offer suggestions for overcoming them.

If you're getting behind in science, are you: Making sure to sign a few RAs (preferably with allies) when you have extra gold? Expand and keep your cities productive and growing? Keeping up-to-date on science buildings in your cities (especially early multipliers like the university)? Did you build the NC (probably in your capital)? The NC alone can provide a significant boost in science output. Additionally if you have good production terrain you can get a couple mines up and build the GL early which saves you buying/building a library and provides an additional science influx.

I've been beaten to the utopia project myself and all I can think is to blame myself for not watching the victory progress tally and either adapting my strategy or declaring war to slow them down. Sometimes the AI just gets lucky and will win. It does get frustrating sometimes.

The AI is absolutely horrible in combat in vanilla. Thal figured out that additional sight range and promotions, as well as adding vanguard units helped the AI improve a fair amount, although still not at the level of a human player. Even without the sight range upgrade the AI would figure out where you are as soon as you shoot (or if you're firing at a city it can already see you). Presumably you're defending your artillery with a couple melee units so, so I personally am willing to give the AI a couple advantages to help it compete with me and make the game more interesting.

 

[Thalassicus]

If you find Prince too difficult, then try Warlord. You're not forced to play on a specific difficulty.

Basically, in vanilla the AI ignores happiness. In vem the AI instead gets combat and per-era bonuses. On Emperor difficulty the AI receives these bonuses:

Vanilla
+6 happiness
-60% unhappiness from cities
-60% unhappiness from population
+10% growth rate
-15% unit costs
-15% building costs
-15% wonder costs
-15% project costs
-20% unit maintenance
-20% building maintenance
-50% upgrade costs
+20% worker rate
+60% vs barbarians
Starts with 2 warriors.

Vem
+0 happiness
-0% unhappiness from cities
-0% unhappiness from population
+0% growth rate
-0% unit costs
-0% building costs
-0% wonder costs
-0% project costs
-20% unit maintenance
-20% building maintenance
-50% upgrade costs
+25% worker rate
+50% vs barbarians
+24 experience + 6 per era (if militaristic)
+2 visibility
+12% production per era
+12% production per era
Starts with 1 warrior and 1 archer.

AIs starting near lots of water get triremes and workboats. "Militaristic" AIs get some extra workers.

 

[Txurce]

Interesting. We've been talking elsewhere about reducing at least the per-era production bonus, but would it perhaps be more incremental and... neater... to first erase the unit and building maintenance bonuses?

 

[Thalassicus]

The maintenance bonuses provide gold for the gold-spending algorithm. I think that's the part I'd want to keep. The AI also favors quantity over quality, so they usually have much higher maintenance than a human player.

 

[Txurce]

The maintenance bonuses provide gold for the gold-spending algorithm. I think that's the part I'd want to keep. The AI also favors quantity over quality, so they usually have much higher maintenance than a human player.

I was thinking that less gold would mean they'd spend what they had on something other than pop/sci improvements, but can see now that wouldn't often be the case. Could the production bonuses be scaled down by era? I've been providing AI pop growth comparisons to help give you a better sense of when they explode.

 

[AlextheGr8]

How many techs ahead of you is the leader?

I just entered the Industrial Era when I get a message "an unmet player has entered the Modern Era"

 

[AlextheGr8]

If you find Prince too difficult, then try Warlord. You're not forced to play on a specific difficulty.

Basically, in vanilla the AI ignores happiness. In vem the AI instead gets combat and per-era bonuses. On Emperor difficulty the AI receives these bonuses:

Vanilla
+6 happiness
-60% unhappiness from cities
-60% unhappiness from population
+10% growth rate
-15% unit costs
-15% building costs
-15% wonder costs
-15% project costs
-20% unit maintenance
-20% building maintenance
-50% upgrade costs
+20% worker rate
+60% vs barbarians
Starts with 2 warriors.

Vem
+0 happiness
-0% unhappiness from cities
-0% unhappiness from population
+0% growth rate
-0% unit costs
-0% building costs
-0% wonder costs
-0% project costs
-20% unit maintenance
-20% building maintenance
-50% upgrade costs
+25% worker rate
+50% vs barbarians
+24 experience + 6 per era (if militaristic)
+2 visibility
+12% production per era
+12% production per era
Starts with 1 warrior and 1 archer.

AIs starting near lots of water get triremes and workboats. "Militaristic" AIs get some extra workers.

I don't find the difficulties difficult at all, for example I can steamroll over a city state, or a neighbor country, usually without any problem. Also I can fight off and beat the AI when they have double what my army has.

I just am annoyed that they are advancing soooo fast in science. I am okay with being able to build a lot of units if it makes the Ai harder to fight against. But building the UN, advancing to the modern era, all before I even enter the Industrial Era, is maddening. I guess I will lower the difficulty and try again.

 

[Txurce]

Being behind by an entire era is not unheard of. Of course if it happens in every game, the either you should adjust, or the mod should. Otherwise, ca you isolate what's making staying even in science difficult? For example, on Emperor the AI's population makes it hard for me to pull ahead in the late stages.

 

[Zaldron]

Being behind by an entire era is not unheard of. Of course if it happens in every game, the either you should adjust, or the mod should. Otherwise, ca you isolate what's making staying even in science difficult? For example, on Emperor the AI's population makes it hard for me to pull ahead in the late stages.

I actually dropped back to king from emperor specifically to hone my strategies and practice because I was getting so far behind in tech. I'd like to hear other experiences too.

 

[Txurce]

I actually dropped back to king from emperor specifically to hone my strategies and practice because I was getting so far behind in tech. I'd like to hear other experiences too.

I can top that! I just played 100 turns at Prince, to see what the difference was. Using Korea, I was way ahead in tech, but that wasn't too far off the Emperor norm at that stage for Korea. I was in the ballpark with population, despite having missed the HG and not having a ton of farmland in my first two cities. (I had four). The AI expanded more slowly on Prince. I quit after the Ceremonial Rites bug.

On Emperor, using Korea, the AI catches me in beakers between t250-300. Unless v138 has some fundamental changes - in which case I'll try Korea again for comparison - I may go with America or the Aztecs for a wide, pop-heavy civ, or try France or Songhai: I did really well in science going wide with them earlier. I'm skipping over Siam because I'm pretty sure I'd hold up with them. I expect to play India in the featured game.

 

[Zaldron]

On Emperor, using Korea, the AI catches me in beakers between t250-300. Unless v138 has some fundamental changes - in which case I'll try Korea again for comparison - I may go with America or the Aztecs for a wide, pop-heavy civ, or try France or Songhai: I did really well in science going wide with them earlier. I'm skipping over Siam because I'm pretty sure I'd hold up with them. I expect to play India in the featured game.

I'm actually playing a France "wide" (only room for 8 cities) science game right now and it looks like even with dedicated research and keeping all RAs fully active through 2/3rds+ of the game, I'm going to lose to a utopia project. I'll probably finish the game tomorrow.

 

[pthmix]

Yeah, the AI is a very fast at teching (Emperor). I dealt with the problem by always building the Great Library but this "strategy" is really boring after a while...

Is it possible to adjust the tech-discount you get when another civ has researched a tech that you have not researched yet? I think this could help.

 

[wobuffet]

Yeah, the AI is a very fast at teching (Emperor). I dealt with the problem by always building the Great Library but this "strategy" is really boring after a while...

Is it possible to adjust the tech-discount you get when another civ has researched a tech that you have not researched yet? I think this could help.

Agreed that needing to grab the Great Library just to keep pace with the leading AI civs in Science is no fun at all.

Something like the catch-up mechanic you describe, however, is already in the game: see here for details.

The relevant value is TECH_COST_TOTAL_KNOWN_TEAM_MODIFIER in
...Steam\steamapps\common\sid meier's civilization v\assets\Gameplay\XML\GlobalDefines.xml , set to 30 by default. I think corresponds to a beaker discount of up to 1 – 1/(1 + [7/8]*30/100) = 20.8% on a standard size (8 player) map if every civ other than you has already researched a technology.

I'd be more than okay with experimenting with bumping this value up, although I doubt this alone will satisfactorily solve the AI runaway Science problem.

[edit: If you don't believe me, you can verify this by loading a game in progress and seeing that techs in the same column often have slightly different beaker totals.]



In any case, I think by this point (extensive discussion in two 3-page threads) it's safe to say there's a consensus that VEM has a problem with runaway AI Science leaders, at least at higher difficulty levels. Washington, in particular, seems to always have an incredible population and beakers per turn in my games.

I don't mind having one or two leader Civs that I have to watch out for if not deal with directly (i.e., cripple or eliminate) each game, but when I consistently have no chance at all at nabbing the late Medieval and Renaissance World Wonders (e.g., Notre Dame and The Forbidden Palace), the game's simply less fun to play.

 

[pthmix]

I know that's in the game.

And it is really not surprising that Washington has such a high beaker per turn given his unique building.

 

[wobuffet]

edit: Incorrect math corrected by by Seek in spoilers:

Spoiler :

One of the issues is that Public Schools are simply way too powerful. Consider the Science output of a satellite City with 17 and 3 from one Scientist.
With a Library (+1 per ): 20
...and a University (+33% ): 26.6 [a +33% increase]
...and a Public School (+3:c5science, +1 per ): 53.2 [a +100% increase]

I'd propose something more reasonable like +1 per 3 for the Public School. This looks terribly weak, but would still actually raise the City's yield by a healthy amount (26.6 -> 34.1, a 28% increase). This kind of increase, as we would want, would still very easily be worth it right away for any Civ prioritizing Science, and eventually worth it for virtually all Civs.

Obviously mid- and late-game tech costs would have to be rescaled after such a change, but I think the gameplay result would be worth it. Not all of the AIs prioritize Scientific Theory and Public Schools above all other options. This shouldn't cripple them. Similarly, beelining through the tech tree for Scientific Theory is almost always the best choice for humans who want to keep pace with Tech leaders, which is no fun.

Here's a list of changes I think could help Science balance, esp. with respect to runaways:

• Make Public Schools less powerful. (edit: probably less of an issue than I thought)
• Change one of Rationalism's + policies (Free Thought [+1 on Villages, +17%:c5science from Universities], perhaps?) to something like "+20% from Public Schools," to reduce the raw beaker advantage of Civs who choose Rationalism.
• Add something like "+20% in Cities with a National Wonder" to the Tradition finisher, to help tall civs compete in Science. (The current finisher's +25% surplus could be moved to, say, Landed Elite, replacing the +2 per City. But it might be too easy for wide players to cherrypick the +25% then, so Ceremonial Rites could be added as a prerequisite).
• Reduce the Pioneer Fort's + bonus. (Has anyone else noticed Washington is often a population & Science powerhouse?)
• Lastly, I suspect the -% Building maintenance for AIs has a snowballing effect for expansionist/wide AIs: they get just as much Income but pay less maintenance for each new City, meaning they can buy out lots of CSs for the luxiries/.

 

[Seek]

One of the issues is that Public Schools are simply way too powerful. Consider the Science output of a satellite City with 17 and 3 from one Scientist.
With a Library (+1 per ): 20
...and a University (+33% ): 26.6 [a +33% increase]
...and a Public School (+3:c5science, +1 per ): 53.2 [a +100% increase]

I'd propose something more reasonable like +1 per 3 for the Public School. This looks terribly weak, but would still actually raise the City's yield by a healthy amount (26.6 -> 34.1, a 28% increase). This kind of increase, as we would want, would still very easily be worth it right away for any Civ prioritizing Science, and eventually worth it for virtually all Civs.

Obviously mid- and late-game tech costs would have to be rescaled after such a change, but I think the gameplay result would be worth it. Not all of the AIs prioritize Scientific Theory and Public Schools above all other options. This shouldn't cripple them. Similarly, beelining through the tech tree for Scientific Theory is almost always the best choice for humans who want to keep pace with Tech leaders, which is no fun.

Here's a list of changes I think could help Science balance, esp. with respect to runaways:

• Make Public Schools less powerful.
• Change one of Rationalism's + policies (Free Thought [+1 on Villages, +17%:c5science from Universities], perhaps?) to something like "+20% from Public Schools," to reduce the raw beaker advantage of Civs who choose Rationalism.
• Add something like "+20% in Cities with a National Wonder" to the Tradition finisher, to help tall civs compete in Science. (The current finisher's +25% surplus could be moved to, say, Landed Elite, replacing the +2 per City. But it might be too easy for wide players to cherrypick the +25% then, so Ceremonial Rites could be added as a prerequisite).
• Reduce the Pioneer Fort's + bonus. (Has anyone else noticed Washington is often a population & Science powerhouse?)
• Lastly, I suspect the -% Building maintenance for AIs has a snowballing effect for expansionist/wide AIs: they get just as much Income but pay less maintenance for each new City, meaning they can buy out lots of CSs for the luxiries/.

I think you forgot to count base beakers from population, wobuffet. I'm getting different numbers:
17 +3 = 20
With Library = 37 (85% increase)
With Uni = 49.21 (33% increase)
With PS = 74.82 (52% increase)

I really don't think the issue lays with the Public School, but the fact that RAs will Allies jumped from being 50% to 200% better than normal RAs in v131, and it's this which is causing the runaway problem. (There is some confusion here, because the 131 patch notes say that the bonus went "from 150%" but I see no mention of RAs in any patch notes after v108.2 beta, which listed the DoF/Allied bonus at 50%. So there could have been an undocumented interim jump from 50% -> 150% along the way. Speaking of undocumented changes, I'm not sure what is intended: 1 science per pop or 2 - after the science scaling in v131.1 I was seeing two/pop, but in 137 it's only one.)

Aside from nerfing Public Schools I have no problem with the other suggestions above.

 

[wobuffet]

Thanks for the correction, Seek!
A 52% increase from the Public School still seems like a lot to me, for what it's worth. (I assume every AI gets a Library reasonably soon in every city, so the first 85% doesn't worry me too much.)


Re: RAs, unless I'm misunderstanding something, I don't think Research Agreements are the crux of the runaway problem – or at least not all of the story. For example, in my current game (on 6 difficulty), InfoAddict (which AFAIK doesn't include Income from Open Borders or Science from RAs) tells me that Washington is, domestically, researching as fast as pretty much the rest of the known world combined:

Spoiler :

(I'm China, by the way. I built the Great Library but didn't expand past two Cities until far too late.)

That said, I'd support reducing RAs by a third to 2% of combined output (4% with an alliance), at least experimentally/to see how much it helps/hurts/changes things.