[1.19a] Strategic AI Analysis

from Elemental Forums

Purpose: Analyze strategic AI behavior to help identify issues and hopefully help improve the game.

Scope: Basic block and tackle strategic actions such as exploration, expansion, city improvements, etc.

Scenario:

  • Large custom map built from "Ovia" - created an island so I can cycle turns, reduced starting positions to 10 and spread them out
  • Ridiculous across the board
  • Max players
  • Normal pacing
  • Custom sov
  • Playing as Capitar
  • Used Ctrl-U cheat to see entire map
  • Used Ctrl-Z cheat to cycle turns

Saves:

Faction Concerns:

  • Pariden
    • S12 FP5 (1 city) - housing, goodie huts, unit queue
    • S24 FP5 (1 city) - goodie huts, unit queue
    • S36 FP9 (1 city) - goodie huts
    • S48 FP19 (2 cities) - goodie huts
    • S60 FP24 (2 cities) - goodie huts
  • Kraxis
    • S12 FP5 (1 city) - housing, goodie huts, unit queue
    • S24 FP9 (1 city) - goodie huts
    • S36 FP12 (2 cities) - goodie huts, no unit defending additional settlement
    • S48 FP21 (3 cities) - goodie huts
    • S60 FP47 (4 cities) - goodie huts
  • Gilden
    • S12 FP5 (1 city) - housing, goodie huts, unit queue, improvement queue
    • S24 FP5 (1 city) - housing, goodie huts, unit queue
    • S36 FP24 (1 city)  - goodie huts
    • S48 FP47 (2 cities) - none
    • S60 FP57 (2 cities) - available champions
  • Resoln
    • S12 FP5 (1 city) - goodie huts, unit queue
    • S24 FP10 (1 city) - goodie huts
    • S36 FP11 (1 city) - goodie huts, pioneers (pathing?)
    • S48 FP27 (2 cities) - goodie huts, military units (pathing), available champions
    • S60 FP33 (3 cities) - goodie huts, military units (pathing), available champions
  • Yithril
    • S12 FP3 (1 city) - goodie huts, unit queue
    • S24 FP5 (1 city) - goodie huts, unit queue
    • S36 FP10 (1 city) - goodie huts
    • S48 FP10 (2 cities) - goodie huts, available champions
    • S60 FP33 (2 cities) - goodie huts, available champions
  • Magnar
    • S12 FP5 (1 city) - goodie huts, unit queue
    • S24 FP10 (1 city) - goodie huts, unit queue
    • S36 FP10 (1 city) - goodie huts, pioneers (not building)
    • S48 FP12 (2 cities) - goodie huts
    • S60 FP30 (3 cities) - goodie huts, available champions
  • Tarth
    • S12 FP3 (1 city) - housing, goodie huts
    • S24 FP7 (1 city) - goodie huts
    • S36 FP15 (3 cities) - goodie huts
    • S48 FP31 (3 cities) - goodie huts
    • S60 FP38 (3 cities) - goodie huts, available champions
  • Altar
    • S12 FP5 (1 city) - housing, goodie huts
    • S24 FP6 (2 cities) - housing, goodie huts, unit queues, no unit defending additional settlement
    • S36 FP18 (2 cities) - goodie huts, no unit defending additional settlement
    • S48 FP24 (4 cities) - goodie huts
    • S60 FP39 (4 cities) - goodie huts
  • Umber
    • S12 FP4 (1 city) - housing, goodie huts
    • S24 FP6 (2 cities) - none
    • S36 FP17 (2 cities) - none
    • S48 FP36 (5 cities) - none
    • S60 FP69 (6 cities) - none

Summary of Observations:

  • Neither soveriegns or spouses ever leave capital to explore, level, or grab goodie huts - the only factions grabbing goodie huts are those who hired champions. Recommend giving sovereigns and spouses xp per turn or have them explore the world (being careful of course)
  • AI doesn't seem to weight housing as much as it should - i.e. on ridiculous why does it need a workshop or other improvement before housing when it has 10+ prestige a turn and all the other bonuses? Recommend AI on ridiculous is hardcoded to build x number of housing units or get population to y value before building anything else.
  • AI seems to randomly leave the unit or improvement queues empty on occasion - they should be building nonstop on ridiculous due to the resource buffs. Recommend some kind of check is done to ensure those queues never remain empty.
  • AI is not hiring all available champions - Umber, however, is hiring like crazy (it's also building like crazy). Recommend the second a champion is available to hire, they're hired (and kept geared of course).
  • Resoln had some army pathing issues - maybe due to map location as Gilden was having a similar issue in the same location for my 1.19 analysis. Recommend exploring why that map location seems to cause the pathing concerns.
  • All in all the fixed xml tags have certainly improved the AI over 1.19. Good job Stardock!

[Edit] Brad - If you choose to implement some or all of these changes you might start hearing how the ridiculous AI needs a nerf. That would be a good thing. :grin:

For 1.19b analysis see here: https://forums.elementalgame.com/405148

6,976 views 3 replies +1 Loading…
Reply #1 from Elemental Forums Top

thanks for doing this...  i hope these suggestions don't fall on deaf ears.

Reply #2 from Elemental Forums Top

Quoting Trojasmic, reply 1
thanks for doing this...  i hope these suggestions don't fall on deaf ears.
End of Trojasmic's quote

You're welcome! Brad challenged us to provide details and be specific - trying to meet those deliverables. :grin:

I think this weekend I'll analyze the same save files to determine whether the AI is properly gearing their sovereign, spouse, and champions (for those who have hired). I'll note the gildar that they have, current tech levels, and report the results. I didn't pay attention to these things in the first analaysis - wanted to limit my scope to whether they were performing the fundamental strategic decisions.

Star Control: Origins
Reply #3 from Elemental Forums Top

I have noticed that it seems like the AI never uses consumable items if their sovs do pick something up. I.E. books that give + to atts, exp potions etc... 

It also seems the AI doesn't put much much importance on leveling sovs and champs. When I'm running around with level 6-8 they usually are still at 2-3 and with no armor.

 

Great post btw