Caster of Magic II, AI elimination mechanic tests

[Seravy]

Test 6 : Expert, Large

Elimination mechanic enabled fully :
   
Elimination mechanic enabled in diplomacy only, disabled in combat:
   
Elimination disabled :
   

[Seravy]

Test 7 : Lunatic, Tiny, Islands

Elimination mechanic enabled fully :
   
Elimination mechanic enabled in diplomacy only, disabled in combat:
   
Elimination disabled :
   

[Seravy]

Summary :

   



Assuming we want 6 wizards surviving at the end of the game on average, that would mean a total of 42 surviving wizard in 7 games.
The first column - using both elimination mechanics - is reasonably close - without the low difficulty test run where the AI had no troops to effectively fight, we'd be quite close to it.

The second is somewhat worse, with 10 more surviving wizards, out of which 1 difference came from the low difficulty game.
The third is much worse, having 22 more surviving wizards out of which none come from the low difficulty.

The "weak AI' column difference specifically shows the opposite : removing the combat mechanic resulted in 12 more surviving wizards in this category vs only 18 more if both mechanics were removed. Meanwhile the average wizard column stays nearly unaffected by the removal of the combat mechanic.

What this tells us is, the diplomacy part of the feature is necessary, as it guarantees the wars that result in reducing the number of average AI wizards, while strengthening the top ones.
What this also tells us is, we can remove the combat mechanic entirely, but we do need to have a replacement that specifically triggers on already weakened wizards and makes them somehow disappear.

The possible options we have are :
-Make the wizard turn neutral (retire, die in an accident, get destroyed by a revolt, etc)
-Make the wizard reduce garrisoning requirements - for example 6 in fortress, 5 in other cities, and use the rest of the units to build stacks and try to conquer.

I think we should give up on the other ideas though (random event or monsters killing the wizards, wizard surrendering to other players, wizard failing to pay maintenance).

The second feels less artificial but the risk there is, if they don't have more than one city and their units get killed regularly, then they have no opportunity to actually build a stack so the garrison stays maxed. Maybe I should try running some tests with that next.

[Intaka]

I don't mind the diplomacy mechanic. Feels real enough. Eat the weak.

The neutral city thing is much easier to adjust to your end goals. After year X, any AI wizard this ratio below top AI wizard goes neutral at a 5% chance per turn. Done. They will fall out in a seemingly natural way, and it is easily adjustable.

Just for clarification, this will never cause the last AI wizard to retire, correct? That would be a bummer.

[Intaka]

You could even factor game difficulty into a neutral retirement forumla like the one above.

[Seravy]

The elimination can never reduce the number of wizards below 4 so no, it won't affect the last wizard.

The neutral retirement is easy to code but I feel the AI simply moving their garrisons out a few units at a time matches the game better, as it leaves what happened to the player's imagination. Maybe they got desperate, maybe they just have some civilian issues to deal with like organized crime, maybe they are running low on food or something else because the siege has been going on for years, etc.
The other advantage of this solution is, if the AI manages to build up a stack that way, which is not impossible if they are in the human player's territory, far out of reach for other wizards, then they can actually attack the human player with that stack or steal their lairs and nodes.

[Suppanut]

How ai retirement works in CoM I? I have it twice when I collapsed ai's economic support and lost most of units via desertion.

I think both voluntary retirement of weak wizard and lost power by rebellion if unrest of the empire out of control (or even uprising events lead by heroes units) are good idea for me and should implement both.

[Seravy]

Ran the tests on all 7 games with the garrison mechanic added and it doesn't work unfortunately. We most likely will need the retirement feature.

In CoM I you're given the option to win the game if your total power is more than twice all other wizards combined, but there is no way for one specific AI to retire alone.

[Anskiy]

Ran the tests on all 7 games with the garrison mechanic added and it doesn't work unfortunately. We most likely will need the retirement feature.

What happened when you implemented the ungarrisoning mechanic? Was the AI too reluctant to make their troops step out of their forts?

[Seravy]

Ran the tests on all 7 games with the garrison mechanic added and it doesn't work unfortunately. We most likely will need the retirement feature.

What happened when you implemented the ungarrisoning mechanic? Was the AI too reluctant to make their troops step out of their forts?

Not sure, I only saw the resulting charts. I'm guessing the cities were still too hard to conquer with 2-3 units missing from them.

I'll try this mechanic next :

-Wizards still in play 36 turns after being selected for elimination will check for retirement conditions :
Can't be at war with the human player
Can't be Guardian
Must own no more than (Land Size setting)/2 +1 cities
No random events happened this turn
If all conditions are satisfied, there is a 20% chance for the wizard to retire and their cities turn neutral.