AI's Favorite Techs

DaveMcW

Deity
Joined
Oct 8, 2002
Messages
6,489
These are the 20 favorite techs of the Conquests AI. Never research them if you can trade!

Top 20
1. Rocketry
2. Replaceable Parts
3. Bronze Working
4. Nationalism
5. Flight
6. Warrior Code
7. Invention
8. Feudalism
9. Gunpowder
10. Computers
11. Map Making
12. Astronomy
13. Magnetism
14. Combustion
15. Fission
16. Ceremonial Burial
17. Pottery
18. The Wheel
19. Masonry
20. Iron Working


These are the 20 most avoided AI techs. You can easily get a monopoly, but some of them are pretty useless!

Bottom 20
1. Facism
2. Democracy
3. Atomic Theory
4. Steel
5. Superconductor
6. Ironclads
7. Integrated Defense
8. Smart Weapons
9. The Laser
10. Physics
11. Chemistry
12. Printing Press
13. The Laser
14. Communism
15. Espionage
16. Nuclear Power
17. Recycling
18. Currency
19. The Republic
20. Stealth


Favorite techs by era:

All Ancient Techs
1. Bronze Working
2. Warrior Code
3. Pottery
4. Ceremonial Burial
5. The Wheel
6. Masonry
7. Map Making
8. Mysticism
9. Iron Working
10. Horseback Riding
11. Alphabet
12. Philosophy
13. Writing
14. Mathematics
15. Code of Laws
16. Polytheism
17. Construction
18. Literature
19. Monarchy
20. Currency
21. Republic

All Middle Age Techs
1. Invention
2. Feudalism
3. Gunpowder
4. Astronomy
5. Magnetism
6. Navigation
7. Literature
8. Theology
9. Chivalry
10. Monarchy
11. Engineering
12. The Republic
13. Theory of Gravity
14. Monotheism
15. Education
16. Economics
17. Metallurgy
18. Military Tradition
19. Banking
20. Music Theory
21. Printing Press
22. Free Artistry
23. Chemistry
24. Physics
25. Democracy

All Industrial Techs
1. Replaceable Parts
2. Nationalism
3. Flight
4. Combustion
5. Steam Power
6. Refining
7. Industrialization
8. Electronics
9. Navigation
10. Mass Production
11. Chivalry
12. Scientific Method
13. Motorized Transport
14. Electricity
15. Advanced Flight
16. Economics
17. Military Tradition
18. Music Theory
19. Free Artistry
20. The Corporation
21. Sanitation
22. Amphibious Warfare
23. Espionage
24. Printing Press
25. Medicine
26. Communism
27. Ironclads
28. Steel
29. Atomic Theory
30. Democracy
31. Facism

All Modern Techs
1. Rocketry
2. Nationalism
3. Computers
4. Fission
5. Synthetic Fibers
6. Genetics
7. Miniaturization
8. Space Flight
9. Robotics
10. Ecology
11. Advanced Flight
12. Satellites
13. Sanitation
14. Amphibious Warfare
15. Stealth
16. Recycling
17. Nuclear Power
18. Espionage
19. The Laser
20. Smart Weapons
21. Integrated Defense
22. Communism
23. Ironclads
24. Superconductor
25. Facism


Acknowledgements

Thanks to Ambiorix for the excel spreadsheet I used to calculate this.

Thanks to alexman for explaining the AI tech choice formula.

For more detailed AI research probabilities, check the link above or the attached spreadsheet.
 

Attachments

  • AI_tech_research_estimations_v2.2.zip
    8.8 KB · Views: 381
Looks interesting, but a few of those go against what I've experienced. Is it possible you left out the ability to get a new govt as a reason to research? Democracy & Fascism are the 2 worst, and Communism is bottom 15, but in most games I've played, the AI will go for all 3 of those almost as soon as it has a chance.
 
Nice work :)

This looks promising but I too often find the AI researching many of the bottom techs before they should. Could there be a random element included?
 
Yes, there is a random element. The top 5 or so techs in an era will be researched at close to 100% probablility, but it gets harder to predict as you go down the list.
 
Well, actually, apart from Literature and Ironclads, the AI has a tendency to research Non-needed techs. THeye also research all governments, especially facism and communism, its like tey do a beeline.
 
Yes, I think that you overestimate the AI's tendency to avoid non-required techs. In my experience, while Printing Press, Ironclads and Literature are certainly easy to get monopolies on, The Republic, Democracy, Economics and Espionage are not.
 
The only reason that I won COTM 13 was that the AIs spent the large (20-30-40) number of turns by which they beat me to the IA on Nationalism and the optional techs stemming from it.
 
Have to agree. The lists sounds highly accurate, with the huge exception of Communism and Fascism. In my experience, the AI will always go Nationalism-> Comm - > Fascism before they research Steam Power.
And IMHO Nat is the top choice number one - way before Rocketry and BW. You will also see that in mods - it's the Drafting, MPPs and Mobilization that makes a tech so appealing for the AI. Much more than the Alu from Rocketry. And both give a resource-free defender/attacker.

There also seems some dependance on the level: The research choice for the leading AI is one thing; but once one of them has Nat/Fasc/Comm, the costs drop, and all others will follow instead of getting Steam. The level matters here because if the human has the tech lead, other techs will be researched first (especially Steam), making them cheaper for the AI.

Edit: Actually, after looking at Alexman's :worship: tab again, I really wonder how you came to rating Rocketry above Nat:

Nationalism: 259 (conscription) + 259 (mob) + 198 (defender) + 134 (attacker) + 8 (MPP) +2 (specialist) =860
Divide by 1.5 (optional) = 573.3333333

Rocketry: 198 (defender) + 134 (attacker) + 4 (Jet Fighters, CM) + 16 (alu) = 352

And the optional Gov techs Comm and Fasc score 261 (Police Stations)/259 divided by 1.5 = 174/172.666666

Steam: 16 (Coal) + 1 (RR) + 6 (Iron Works) = 23

Of course, the tech costs need to be figured into the calculation. But that depends on the number of Civs who already know that tech - and it for sure doesn't equalize a 174:23 ratio...
 
I agree that the AIs will always beeline Commie & Fascism. They're almost obsessive about it. In COTM14, everyone researched these - which of course they're reluctant to trade around once they have them - slowed the AI help I'd had in the middle ages from a veritable fountain, to a crawl (all I got from them was Industrialization, as opposed to 4 or 5 techs in the middle ages).

I'm wondering if a better human strategy would be to research all three (Nationalism, plus the two govts), and gift them to the AI, in order to set them on the right path?!

I'd certainly gift them as soon as I can trade for them, if I think a particular AI is going to be helpful (again, COTM14 - by the time I'd finished Flight+Mech Transport, not one single AI out of those remaining, all of them by then Industrial, had researched Medicine :crazyeye: )
 
Doc Tsiolkovski said:
Have to agree. The lists sounds highly accurate, with the huge exception of Communism and Fascism. In my experience, the AI will always go Nationalism-> Comm - > Fascism before they research Steam Power.
And IMHO Nat is the top choice number one - way before Rocketry and BW. You will also see that in mods - it's the Drafting, MPPs and Mobilization that makes a tech so appealing for the AI. Much more than the Alu from Rocketry. And both give a resource-free defender/attacker.

There also seems some dependance on the level: The research choice for the leading AI is one thing; but once one of them has Nat/Fasc/Comm, the costs drop, and all others will follow instead of getting Steam. The level matters here because if the human has the tech lead, other techs will be researched first (especially Steam), making them cheaper for the AI.

Edit: Actually, after looking at Alexman's :worship: tab again, I really wonder how you came to rating Rocketry above Nat:

Nationalism: 259 (conscription) + 259 (mob) + 198 (defender) + 134 (attacker) + 8 (MPP) +2 (specialist) =860
Divide by 1.5 (optional) = 573.3333333

Rocketry: 198 (defender) + 134 (attacker) + 4 (Jet Fighters, CM) + 16 (alu) = 352

And the optional Gov techs Comm and Fasc score 261 (Police Stations)/259 divided by 1.5 = 174/172.666666

Steam: 16 (Coal) + 1 (RR) + 6 (Iron Works) = 23

Of course, the tech costs need to be figured into the calculation. But that depends on the number of Civs who already know that tech - and it for sure doesn't equalize a 174:23 ratio...

Remember, the bonus from government itself depends on the cost, (so you would divide the 259 by ~4-20 depending on the size of the AI) Meaning that it does equalize a 174:23 ratio..if it takes the AI about 8 turns to get Com or Fasc.
 
Did you alter the research rate (or whatever the best name for it would be) as you went through the technologies, or did you leave it at one value (for example, the default 10) for all technologies? That could cause the discrepencies in your list and game experience. In particular, communism and fascism would be much more desirable (due to the government bonus based on tech cost).
 
Here's an alternate list, based roughly on 10-turn research. The research multipliers (which when multiplied by the base cost equal the research time) were 3 for the first-tier ancient age techs; 2 for the second-tier techs; 1 for the third-tier techs; .3 for the middle age techs up to education, music theory, printing press, and gunpowder; .1 for the rest of the middle age techs; .05 for the industrial techs up to refining, steel, electricity, and scientific method; .04 for the rest of the industrial techs; and .03 for the modern age techs. Some of those values are probably off, but changing them didn't significantly change the research order, so I left them where they are. In particular the third-tier ancient age value is probably too high.


1. Rocketry
2. Replaceable Parts
3. Nationalism
4. Flight
5. Bronze Working
6. Warrior Code
7. Invention
8. Feudalism
9. Space Flight
10. Computers
11. Synthetic Fibers
12. Gunpowder
13. Astronomy
14. Superconductor
15. Magnetism
16. Combustion
17. Map Making
18. Satellites
19. The Laser
20. Steam Power
21. Fission
22. Robotics
23. Communism
24. Scientific Method
25. Mass Production
26. The Corporation
27. Industrialization
28. Banking
29. Medicine
30. Fascism
31. Motorized Transportation
32. Sanitation
33. Democracy
34. Philosophy
35. Refining
36. Free Artistry
37. Electronics
38. Pottery
38. Ceremonial Burial
40. Metallurgy
41. Chemistry
42. The Wheel
43. Iron Working
44. Theory of Gravity
45. Electricity
46. Physics
47. Navigation
48. Construction
49. Espionage
50. Amphibious Warfare
51. Mysticism
52. Ecology
53. Genetics
54. Recycling
55. Steel
56. Miniaturization
57. Writing
58. Economics
59. Ironclads
60. Nuclear Power
61. Masonry
62. Atomic Theory
63. Smart Weapons
64. Stealth
65. Military Tradition
66. Horseback Riding
67. Engineering
68. Advanced Flight
69. Theology
70. Code of Laws
71. Polytheism
72. Monotheism
73. Mathematics
74. Chivalry
75. Integrated Defense
76. Education
77. Literature
78. Alphabet
79. Monarchy
80. Music Theory
81. Currency
82. Printing Press
83. The Republic

The previous list was made using the research multiplier when the tech could be researched; the following lists use the lower multiplier of the next level for older techs (such as optional techs in the next age). These lists show the order the AI would research the techs in.

Ancient Age:
1. Bronze Working
2. Warrior Code
3. Pottery
3. Ceremonial Burial
5. The Wheel
6. Iron Working
7. Masonry
8. Mysticism
9. Horseback Riding
10. Alphabet
11. Writing
12. Map Making
13. Philosophy
14. Mathematics
15. Construction
16. Code of Laws
17. Polytheism
18. Literature
19. Monarchy
20. Currency
21. The Republic

Middle Ages:
1. Feudalism
2. The Republic
3. Engineering
4. Invention
5. Gunpowder
6. Monotheism
7. Theology
8. Chivalry
9. Education
10. Astronomy
11. Banking
12. Printing Press
13. Democracy
14. Free Artistry
15. Music Theory
16. Chemistry
17. Metallurgy
18. Physics
19. Magnetism
20. Theory of Gravity
21. Navigation
22. Economics
23. Military Tradition

Industrial Age:
1. Nationalism
2. Navigation
3. Economics
4. Steam Power
5. Military Tradition
6. Communism
7. Industrialization
8. The Corporation
9. Medicine
10. Fascism
11. Sanitation
12. Electricity
13. Replaceable Parts
14. Scientific Method
15. Refining
16. Espionage
17. Steel
18. Combustion
19. Flight
20. Mass Production
21. Motorized Transportation
22. Ironclads
23. Amphibious Warfare
24. Atomic Theory
25. Electronics
26. Advanced Flight

Modern Age:
1. Rocketry
2. Space Flight
3. Computers
4. Satellites
5. Fission
6. Superconductor
7. Ecology
8. Synthetic Fibers
9. Recycling
10. Advanced Flight
11. Miniaturization
12. Genetics
13. Nuclear Power
14. The Laser
15. Robotics
16. Smart Weapons
17. Stealth
18. Integrated Defense

Some civs may, due to their UU, have a different priority. Those with significant differences are Rome, Japan, India, Spain, and Maya. Significantly missing is in that list is Babylon, whose bowman is not defensive in C3C.

There are some indications of problems with this list, such as literature before monarchy, military tradition being last in the middle ages, medicine (and several other techs) before fascism, and computers and fission being so low in the modern age.

P.S. DocT, the values for conscription and mobilization are divided by the number of turns.
 
P.S. DocT, the values for conscription and mobilization are divided by the number of turns.
Sure. But so are the values for anything (unless I completely missunderstand the whole topic).
And I was using that mainly to adress the Steam Engine - Communis issue, with the same tech cost modifier. Rocketry for sure is the most desirable tech for the AI in te Modern Era, and in my experience trading for the conscription tech is always the most expensive trade - that's why I figured Nat would come for Rocketry in an overall ranking.
However, the real big issue why experience doesn't match the proposed list: Research time.
In the levels I usually play (Emperor and above), the leading AI(s) will be able to reseach Nationalism and the next Gov techs in 4 turns - and after they're done with it, the costs for the midsize AIs has dropped enough to allow them a 4-6 turn research as well.
So, they may occasionally go for Steam earlier - but they NEVER will prefer Medicine (as you already mention).
And, the equitation doesn't explain why they'd ever research Ironclads - any other choice at that time (before only AT is left as next mandatory tech) would be more desirable (the 4 IA techs rated even lower could be the best choice at one point; 2 are mandatory, and the other 2 considerably cheaper than Rocketry, so the research time modifier comes into play).

And, in the Middle Ages list Navigation simply doesn't work. The AI will almost always go there first before Physics and often before Metallurgy.
Also, Polytheism is, from experience, a beloved AI tech - much more than CoL, for example.

Something else: Certain Civs prefer to research different techs. That is absolutely reasonable if it relates to their UU (Rome). But I cannot see anything explaining why America and Ottomans head to Polytheism for about 99% of the time. Before Alpha.
It could be that traits affect research choices; other IND Civs have different priorities here (Persia - IW, Maya - WC/BW, France/China/ Carthage - BW, Egypt - Wheel/BW). Of course, America wants BW as well - but will either get it from a hut soon, or due to the easier contacts, trade for it.

Bottom line: I don't think the whole subject is fully understood ;).
 
Just found this. Great discussion.

However I too find a great disconnect between the statistical analysis of what they should research, and the practical reality of what they DO research.

My observations:

Natioanalism. Number one. Communism and Fascism. Always.

Rocketry - almost always.

Steam Power and the path to ToE....almost never.

Medicine..never.

Democracy - very good chance. Free Artistry and music theory - high choices. Printing press low until they get banking.

Literature - never. Philosophy - low. Road to monarchy - high. Road to construction - high. Map Making - moderate.

Ironclads and Economics - be certain at least one AI civ will research one of these in a game. Seems to be random which one.

Modern era - they all drive towards Space Flight and will pretty much ignore the rest of the tree - but if they get the chance to be on it (Computers free tech for sci civs) they will beeline for Genetics.

This list is from my personal experience only - not great, but it does seem to contradict the statistical choices for whatever reason. Perhaps there is a scripted sub-routine for the AI to follow for techs that activates sometimes....
 
Very nice! The list is quite useful and the discussion as well. I've noted differences myself between AI behavior and Dave's list, but I'd agree with about 80% - 85% of it. Very funny that Nationalism is an optional Tech but practically the #1 AI research priority!!
 
from my experiences, the AA techs researched vary for nearly every civ. I've found that the Byzantines go for Philo most of the time, because it has a wonder which starts their golden age, Greece seem to go straight for Constuction and Currency, and the Japanese seem to go for Mystism very early.
 
I found the smae as azzaman. I found that the AI like Nationalism even though it isn't required everyone jumps for it first.
 
If I may, I think the lists would be more relevant in a more number-wise elaborate fashion: adding both the numbers for priority value and cost; setting a few more scales for comparison (early, mid, late-AA, etc.). That way it would be easier to see why the AI researches AA government techs before literature so often, or how it would deal with researching maths when MM is available for research and when it is not.

Edit: it's already being done, I just read the other thread about calcs :blush: It does make for odd stuff vs game reality though, esp. govts, but nevertheless 80%+ accurate at least.
 
There are reasons to research pottery/ceremonial burial if you don't start with them, even if they are favorites of the AI.

Pottery can get you granaries early, which are very useful.

If you have SGL's on, pottery is cheap and if no one else has it, you can get an SGL for it easily, as the AI will be researching WC/BW/CB/The Wheel.
 
Top Bottom