Late-Game Research Balance: Savegame and Findings Thread

[Lokki242]

Following up on lots of the discussion around the snowballing around certain Civs in the current late-game, I thought it might be helpful to create a centralized thread where we can discuss our experiences, share information about our games, and upload saves for Leoreth to take a look at to see what's going on under the hood.

Right now I'm starting to play a few games with the express purpose of trying to study + provide examples of the current snowball tendencies.

For my first game, I chose the Turks (Bukhara) at 1700 AD - Normal speed - Monarch difficulty. I thought they'd be interesting as they often remain at the very bottom of the tech ladder on this start.

This game was quite revealing. Over the first century, running as balanced a budget as I could, I researched one tech. I was many techs behind the other bottom-rung civs (Thailand, Ethiopia) entering the 20th century, having researched 3-4 more over the 1800s.

At the same time, Germany and the Netherlands were running away with wonders at the same time (and I confirmed that the wonder rushing is being done when there's still plenty of OTHER buildings available, so dis-incentivizing wonder building may help). Netherlands also beelined for Film, researching it by 1856 (long before any other Civ), and completing Hollywood by 1884 (I think still before any other civs had discovered Film). The Dutch were in general super-powerful this game (as they tend to be in this start), dominating their historical colonies and gaining voluntary vassals in Canada and Scandinavia, with a confident tech lead. Also unique to this game was Germany completely obliterating Russia in an endless war (c.1860-1910 IIRC) before vassalizing them.

Here's a (after my save) look at the tech situation in 1800:

Spoiler :

Here's the Wonder situation in 1836, as well as proof the Germans weren't out of things to do besides spam wonders:

Spoiler :

Wonders in 1868:

Spoiler :

And lastly, the wonder situation in 1939:

Spoiler :

I've also attached a variety of saves from different points in the game that can hopefully assist with investigation; though I want to note that around 1928 my desire to have fun got the better of me and I WB'd 5-6 techs so I had (near, still not equal) parity to Ethiopia. The last save is from 1955!

I encourage anyone who plays games into the Industrial era and onwards who notices the tech/wonder snowball to share their experiences, encounters, and saves here for ease of access!

 

[banzay13]

Tech spread cost. Some times its 0. Or became 0, then entering Renaissance era. So tech are basically very cheap.

 

[Lokki242]

Test game two for my investigation is commencing! Scandinavia at 1700, Normal/Monarch once again. I'll post my saves and observations throughout here!

 

[banzay13]

I'm playing Khmer. Epic/Monarch. So, AI have twice cheaper research cost (don't know, how it must be on Monarch)
But, some intresting fact.

Japan has the lowest value of research cost. Only France has less, when they only have a capital city (one city) left. Considering that Japan's commerce is 830 per turn (like large empires like France before the fall or England) it can discover new technologies every 3-4 turns. So it's not surprising that Japan research Film in 1808 (it happened when they built the Hermitage before me, after which I reload, then they had 900 points in research per turn).
Also, China has quite big tech cost, even under AI control

 

[Temüjin VIII]

Isn't that probably the Japanese UP working as intended?

 

[Lokki242]

Isn't that probably the Japanese UP working as intended?

sort of, though I don't think "discovering film in 1808" would fall under the "intended" definition. I have noticed that certain AI have an absurd upper leg on the player during this Scandinavian game, which I should finish soon to share some more thoughts/findings

 

[banzay13]

Isn't that probably the Japanese UP working as intended?

They are tech leaders, no one didn't have some tech, not saying willing to trade with them
As a player, then I played Japan UP worked as intended. Also, in my game, I had 900 commers with Korea and Philippines, and maybe some China. I'm not sure how they got so big commers with only cites in Japan

 

[Derdan]

AI Japan gets crazy modifiers. If I'm not mistaken they weren't changed since the previous map and with their much, much better land they have now they are incredibly powerful. It would be nice if they had to expand to be competitive with the Europeans. But AI Japan never even attacks Korea. At most they'll nab Nan Madol or one to two cities in Indochina.

 

[banzay13]

Look like i have found problem with Japan. Once they enter Industrial era - tech cost became 2-3 time cheaper for them (Only for AI). Even if no one want trade. Even in Medieval or if they just born. Used WB to figure out it. Since on big map Japan became much more bigger and have richer lands and with new civics (and Education) they can build 2 science monster-city (and more available tech trade opportunity and more use for UP, they can get to industrial quit fast. Than they launch tech rocket to the sky. @Leoreth

 

[Alexius08]

Does the reduced tech costs for Industrial Era affect only AI Japan?

 

[banzay13]

Does the reduced tech costs for Industrial Era affect only AI Japan?

I didn't checked out, but I think only for them. Couple of messege above I have attached some screenshots with tech cost

 

[banzay13]

Look like i have found problem with Japan. Once they enter Industrial era - tech cost became 2-3 time cheaper for them (Only for AI). Even if no one want trade. Even in Medieval or if they just born. Used WB to figure out it. Since on big map Japan became much more bigger and have richer lands and with new civics (and Education) they can build 2 science monster-city (and more available tech trade opportunity and more use for UP, they can get to industrial quit fast. Than they launch tech rocket to the sky. @Leoreth

And as result

 

[Derdan]

Can't believe I forgot but it probably also helps Japan that they can easily pick up the Shwedagon Paya and Borobudur. Even as the player these wonders can be hard to get to before Japan does. As Burma it's practically impossible to get Shwedagon Paya in my experience.

 

[Lokki242]

Test 2 - POST ONE OF TWO
Denmark (Norse) - 1700 AD - Normal speed - Monarch Difficulty

Alright, game 2 is complete! I played as the norse from 1700 to 2000 (which was perfect timing, as Russia won a space race victory in the year 2000).

I'm going to divide this into three sections, more or less. First are some QoL notes about the Norse in the 1700 scenario, and their specific modern modifiers. Second, I'll go over some more findings/thoughts about balance, and lastly, just some screenshot collections in chronological order. I've also attached a collection of saves, mostly tied to those last screenshots, which Leoreth can use to look under the hood at a given time.
1 - Notes on the Norse (Mostly for 1700)
Some of this may be relevant to the Norse civ overall (like settler regions and modifiers) but a lot of this is specific to the 1700 scenario, which is of course after their UHV and therefore for QOL and historicity rather than 'balance.' The reason I chose the Norse this game is they reliably are the least advanced European civ when played by the AI, often collapse entirely, and are happy to settle Greenland/Iceland while ignoring Narvik annd Finnmark. I'm just gonna hit this stuff point-form as my thoughts are a little scattered:

• When it comes to their low research rate, that was consistent. Even ignoring the snowballing tech leaders, it wasn't until the 20th century running a deficit that I was able to consistently stay above Thailand in tech. Their modifiers are definitely too harsh, perhaps a Viking holdover, that makes modern Scandinavia a total backwater. They probably need a modifier change for late game stuff
• I don't understand how the AI norse collapse so much. Since they only settle Iceland and Greenland typically, 3/5 of their cities are core, which makes me think it most be an economy or civics problem. There is an admittedly realistic synchronicity between tech leads and mega-profits, and I have noticed low tech civs tend to lower stability; may this be a result of economic strain the AI doesn't know how to handle? unsure.
• Speaking of the Norse settlement, it needs some help. I can see why Norway never settles Finnmark: it's not in their settler map. While this is already ahistorical, it hurts the Norse additionally by blocking them from the one iron resource they can grab in the region to offset Sweden's typical three.
  • I suggest expanding their settler map (roughly as shown in the spoilered image) to actually encapsulate the territory--Hammerfest's viability in current terrain is also dubious, and could be changed to a moorland to avoid this.
  • I also personally added a deer resource to represent the traditional sustenance and lifestyles of the Sami people, which I think is minor enough an impact to potentially include.
  • Narvik is almost never settled by the AI; tundra/taiga requires 'island hopping' which stops them the same way it does the Russian's in Siberia.
  • The settlement region in Greenland should possibly also be adjusted--there are settlements in this area. Similarly, Baffin Island (where the capital of Nunavut, Iqaluit, is located) should be Canadian and---oh yeah I still need to do the Canadian north and west geography notes post, one day
    Spoiler Norse Map :

    

    
• why do they start with 0 workers and 0 boats in 1700. that seems a little odd.
• Both Sweden and the Norse struggle form the lack of coal, and AI are not amazing at advocating for trade. This is all real enough, but it did occur to me---should there be Svalbard coal? For the Norse/Russians? A lil island? Probably not. But intriguing.
• Similar to the Jungle/Rainforest resource coverage problem, I've noticed the same thing with taiga/forest (arctic version), where taiga, which can only be cleared with ecology, is covering a resource. In the case of several resources in Northern Canada I assumed this was intentional; but it also blocks the norse copper! Might need a review by someone like the jungle vs rainforest tile problem
• I think representing anthropocentric climate change in the game is good. I think the global warming mechanic that randomly places desert tiles is bad. In this game, I barely got coal power online before getting hit with a desert near Bergen in 1935. By the 1950s, there was another desert. This is borderline comical and I think I'd almost prefer NO implementation over this current system.
  Spoiler Global Warming :

  

  I eventually received another desert 1 tile E from this one in the 1950s.
  

2 - Balance Findings
Just some general notes on balance stuff that occurred to me this run, the chronology is likely more useful for detailed info.

• Indicating the wonder snowball problem, Chateau Frontenac was already completed before the Canadian spawn. Like Buddhist wonders in SEA, there might need to be some rethinking about how capable civs are actually of building their own iconic wonders
  • This is an aside, but I wonder if civs should have a single UW like they do UBs.... much to consider.... could avoid some of these problems....
• The AI occasionally uses voluntary vassalization to climb the tech tree (though I don't think it realizes it's doing this per se.) Not once but twice this playthrough, Canada voluntarily vassaled to another civ, was freely granted all their tech, then promptly declared independence. Check out the screenshots from 1950 to see how that plays out with the UK, for example. Definitely busted their own tech climb.
• Again, this is normal. Snowballing is worse on higher difficulties. By 1955, Germany had completed the Internet, and England had researched Fusion.
• Something Leoreth has asked about in regards to wonders snowballing is whether the AI is building enough buildings. Other than super-producer cities, I think I can confirm, no they are not. Here are some 1930s examples;
  Spoiler Buildings, or lack thereof :

  

  

  

  

3 - Chronological Screenshots
Pay close attention to the absurdly fast rate of the English in particular. Also to Canada--you can see when they willingly vassalize, are tech gifted, and then declare independence in their tech rate.

• 1842 - Note the incredibly low rate of the Norse v. other Euruopean civs, and the very ahead of time wonders under construction
  
  Spoiler 1842 :

  Spoiler Tech :

  

  Spoiler Wonders :

  

  
• 1860 - first wonders screen, already crazy
  Spoiler 1862 :

  

  
• 1900 - I think here you can see that hoarding of wonders is as much if not more so the problem than the tech rates, at least on the wonder side of things....
  Spoiler 1900 :

  

  
• 1919
  Spoiler 1919 :

  

  
• 1929
• Spoiler 1929 :

  
• 1950
  Spoiler 1950 :

  
  
  
  I have reached the image attachment limit... to be continued.... Second post will have my savegames!

 

[Lokki242]

Test 2 - POST TWO OF TWO
3 - Chronological Screenshots continued

• 1970
• Spoiler 1970 :

  
• 2000 - Endgame. I want to note (and include) that England completed the mission to mars in 1993, then Russia won the Space Race this turn wowie zowie. Also, the UN seemingly exists to limit all civics options successfully, and fail to do anything else, player-controlled or otherwise. This means certain parts of the world consistently end up nuclear wastelands by endgame. The US still fails to settle California, Russia fails to get Siberia entirely (but snowballs out of control anyway), and the British and Ottoman Empires seem like to never collapse.
  
  • Japan still hates Sapporo!
  • I've included Britain and Russia's tech screens here as well; horrifying!
    Spoiler 2000 - THE END :

    

    

    

    

    

    

    

    

    

    

 

[Lokki242]

Oh I have a few more saves - sorry to double a few but this is all of them for posterity, not limited. just in case specificity becomes of use later

I also forgot to mention two techs the AI beeline for seem to be Ballistics and Film, hence consistent early Hollywoods in the latter case. For the former, I found every civ in Europe except me had riflemen by 1810, a few turns after I had discovered replacement parts.

 

[banzay13]

I played several late games with Russia and Rus (Epic/Monarch) - in 600AD game England no so good, as in 1.17, mostly coz strong Mughals (or just born Tamils), some time it's really bad

Spoiler Unthinkable :

Also i can tell, that Tech not cheap for player, but for AI - it very cheap. Even with BIG Empire as i build with Rus and Russia - i stole about 1/3 of Industrial tech and 1/4 Global Tech from France/ Poland (Poland with 2 city could research Assembly Line in 7 turns, and me - with full development Rus+Levant+Mesopotamia+Balkans with Istanbul and running 100% science for 7 turns), if not - i would fall behind to far. It remind me 1.17 games with Indonesia or India or someone else with bad research modifiers. Bur Rus and Russia has very good modifiers. Actually, if i didnt stole tech a lot - i could be in historical timeline, but too outdate compare others superpowers.
Usually we have now 2 tech leaders - France and Japan. Sometimes Dutch and England. But they dont skyrocketing atleast. Japan skyrocket then enter industrial, i think it needed to remove some resources (Gold?Silver?) from Japan until 1800 year(?)
What to do with France - I dont know. And France don't want settle Canada now. They have no enemies now, they like England in 1.17
And some thoughts about Spain. Then they get conquer event - they dont have Heavy swordsmans, no Bombards. Sometime i saw even catapults(!)
Toledo such poor land, no wonder they just cant research something. Maybe adding Silver and some food resources witch will exhaust in 1100AD? Or they should get Bombards without having Gunpowder

 

[banzay13]

Oh I have a few more saves - sorry to double a few but this is all of them for posterity, not limited. just in case specificity becomes of use later

I also forgot to mention two techs the AI beeline for seem to be Ballistics and Film, hence consistent early Hollywoods in the latter case. For the former, I found every civ in Europe except me had riflemen by 1810, a few turns after I had discovered replacement parts.

Of coz AI go for Ballistic ASAP, or they will be wipeout by super advanced Natives, Barbarians and Independent