[1.19n] Strategic AI Analysis

from Elemental Forums

Prior 1.19m analysis here: https://forums.elementalgame.com/407804

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
  • Normal sov
  • Playing as Pariden
  • Used Ctrl-U cheat to see entire map
  • Used Ctrl-Z cheat to cycle turns

Saves:

Faction Concerns:

  • Yithril
    • S12 FP5 (1 city) - housing, unit queue
    • S24 FP5 (1 city) - unit queue
    • S36 FP10 (1 city) - none
    • S48 FP15 (1 city) - none
    • S60 FP34 (2 cities) - none
  • Resoln
    • S12 FP5 (1 city) - spouse on a city but not in a city
    • S24 FP11 (1 city) - unit queue, SPOUSE DEAD
    • S36 FP30 (1 city) - SPOUSE DEAD
    • S48 FP44 (1 city) - SPOUSE DEAD, PIONEERS NOT PATHING
    • S60 FP47 (1 city) - SPOUSE DEAD, PIONEERS NOT PATHING
  • Gilden
    • S12 FP3 (1 city) - spouse on a city but not in a city, housing, improvement queue
    • S24 FP7 (1 city) - none
    • S36 FP17 (2 cities) - housing
    • S48 FP29 (3 cities) - available champions
    • S60 FP43 (4 cities) - available champions
  • Umber
    • S12 FP4 (1 city) - spouse on a city but not in a city
    • S24 FP9 (1 city) - spouse on a city but not in a city, unit queue
    • S36 FP12 (1 city) - none
    • S48 FP37 (2 cities) - none
    • S60 FP69 (2 cities) - none
  • Magnar
    • S12 FP3 (1 city) - none
    • S24 FP12 (1 city) - none
    • S36 FP17 (2 cities) - none
    • S48 FP43 (2 cities) - none
    • S60 FP60 (3 cities) - none
  • Altar
    • S12 FP5 (1 city) - none
    • S24 FP9 (1 city) - none
    • S36 FP26 (2 cities) - none
    • S48 FP33 (2 cities) - none
    • S60 FP43 (3 cities) - available champions
  • Tarth
    • S12 FP5 (1 city) - none
    • S24 FP9 (1 city) - none
    • S36 FP17 (1 city) - none
    • S48 FP19 (1 city) - PIONEERS NOT PATHING
    • S60 FP24 (1 city) - unit queue, PIONEERS NOT PATHING
  • Kraxis
    • S12 FP5 (1 city) - housing, improvement queue, unit queue, spouse on a city but not in a city
    • S24 FP6 (1 city) - spouse on a city but not in a city
    • S36 FP12 (1 city) - spouse on a city but not in a city
    • S48 FP16 (2 cities) - spouse on a city but not in a city
    • S60 FP26 (3 cities) - spouse on a city but not in a city
  • Capitar
    • S12 FP5 (1 city) - housing, unit queue
    • S24 FP5 (1 city) - housing, improvement queue, unit queue, POP LOCKED
    • S36 FP5 (1 city) - housing, improvement queue, unit queue, POP LOCKED
    • S48 FP5 (1 city) - housing, improvement queue, unit queue, POP LOCKED
    • S60 FP6 (1 city) - housing, improvement queue, unit queue, POP LOCKED

Summary of Observations:

  • URGENT PROBLEM: Brad, something is desperately wrong with Capitar's AI as also noted in the prior 1.19m analysis. Can you please fix this? He's completely pop locked. Fun factor -10 when an enemy faction refuses to play the game.
  • URGENT PROBLEM: Multiple spouses did not occupy their capitals and instead sat on top of them. Resoln's apparently was attacked and killed by wandering monsters. Kraxis' spouse continued to teabag her capital.
  • URGENT PROBLEM: Pioneers are being created by both Resoln and Tarth but are not pathing to create new settlements. Fun factor -3.
  • Umber and Magnar are playing solid well-rounded games.
  • Looks like gearing spouses and sov's is better than prior builds - but still not using max tech/gildar to rock 'em out.
  • SOV's need to use items they find such as books and stuff, appears they don't.
  • Spouses and sov's should get some xp and level as well.

Can you please plug the holes? Thank you. :cylon:

9,285 views 8 replies
Reply #1 from Elemental Forums Top

Some of you guys really do alot to help the devs, just wanted to say it's appreciated.

Reply #2 from Elemental Forums Top

Quoting Grizzyloins, reply 1
Some of you guys really do alot to help the devs, just wanted to say it's appreciated.
End of Grizzyloins's quote

Thanks and you're welcome!

Groupy 2 - Organize multiple applications into grouped tabs on your Windows desktop
Reply #3 from Elemental Forums Top

ALM - Does Capitar somehow start in the exact same place in these tests?

Reply #4 from Elemental Forums Top

Quoting RexDaddy, reply 3
ALM - Does Capitar somehow start in the exact same place in these tests?
End of RexDaddy's quote

Nope, same map but different starting locations between m and n.

[Edit] I certainly do believe the pathing issues can/are specific to starting locations as I've seen consistency there. However, Capitar in m was in upper right hand corner of map and in n he was in bottom left (going off memory).

Reply #5 from Elemental Forums Top

Wait, does this suggest that the different factions have different AIs attached to them? That's kind of what it sounds like from this post, and I didn't think that was the case.

Reply #6 from Elemental Forums Top

Quoting LightofAbraxas, reply 5
Wait, does this suggest that the different factions have different AIs attached to them? That's kind of what it sounds like from this post, and I didn't think that was the case.
End of LightofAbraxas's quote

I'm not sure to be honest. But Capitar is consistently pop locking on the ridiculous level so the evidence, to me at least, indiciates the factions have their own threads. Actually, they must have their own threads, right? We can choose different difficulty levels per faction so the rules that govern each faction can be different...

I really have no idea other than for two builds in a row, same map, different starting locations, Capitar just sits there and does nothing.

Reply #7 from Elemental Forums Top

Quoting AlLanMandragoran, reply 6
I'm not sure to be honest. But Capitar is consistently pop locking on the ridiculous level so the evidence, to me at least, indiciates the factions have their own threads. Actually, they must have their own threads, right? We can choose different difficulty levels per faction so the rules that govern each faction can be different...

I really have no idea other than for two builds in a row, same map, different starting locations, Capitar just sits there and does nothing.
End of AlLanMandragoran's quote

Hmm.. I don't know? Isn't it possible that the same code is running on individual threads for each faction? What about custom factions controlled by the AI? What faction's AI would they use? Anyway, sorry... I don't want to derail your thread!

Reply #8 from Elemental Forums Top

Quoting LightofAbraxas, reply 7
Quoting AlLanMandragoran, reply 6I'm not sure to be honest. But Capitar is consistently pop locking on the ridiculous level so the evidence, to me at least, indiciates the factions have their own threads. Actually, they must have their own threads, right? We can choose different difficulty levels per faction so the rules that govern each faction can be different...

I really have no idea other than for two builds in a row, same map, different starting locations, Capitar just sits there and does nothing.

Hmm.. I don't know? Isn't it possible that the same code is running on individual threads for each faction? What about custom factions controlled by the AI? What faction's AI would they use? Anyway, sorry... I don't want to derail your thread!
End of LightofAbraxas's quote

Naw, good questions! I'm dying to know how it's coded as well.

I went back and looked at my prior analysis threads (for the 1.2 beta) regarding Capitar, here goes:

  • 1.19n - pop locked (not building any improvements or units) all the way to season 60
  • 1.19m - didn't found city until between season 24 and 36, pop locked between 36 and 60, finally was snapping out of it by 60 (building a hut)
  • 1.19g through l - didn't officially test but I do remember running into some pop lock issues with Capitar in my real games
  • 1.19f - ok
  • 1.19e - ok
  • 1.19d - didnt found city, dead from monsters between season 12 and 24
  • 1.19c - ok
  • 1.19b - pop locked all the way to season 60
  • 1.19a - unknown (ran the test as Capitar)
  • 1.19 - ok

Other factions had pop lock problems throughout these tests...but the results above are for Capitar only.

So it's a mixed bag of results but spread over 15 tests 5 confirmed problems... probably closer to half if I tested g through l. Brad mentioned a couple of months ago that the pop lock thing (when the faction isn't doing anything) is caused by broken xml tags. I think as they're tinkering with memory fixes and such it can break the AI which is why I started doing these analysis threads in the first place.