Campaign/Diplomacy AI

[nikolai1962]

Personally I'm still having trouble with Sicily getting 'stuck' on Tunis if the Moors get it first Sicily will usually just sit there until thier entire starting force goes rebel, and by then even hard level AI is so far in debt it cant do anything for a very long time. However if they dont get stuck, or the Moors dont get Tunis first, Sicily spreads like a deamon. One thing I've done that helps them (and Russia and Scotland, most factions that have the tendency to get 'stuck' early actually) is have a lower invade priority for the first 3 or 4 turns, then boost it up.

Noticed the same thing and can happen with sicily over durazzo too. I'm thinking it is a side-effect of forced invasions maybe? If they're assigned a target this way and then don't take it they seem to lose their way completely.

As a side-note to the second part the AI does seem to be more *efficiently* aggressive sometimes when they're a bit less aggressive overall. I've been thinking of having some early game mini-profiles for specific factions which get switched to the main one after x turns. On the other hand i've been thinking of giving factions like egypt, which have no neighbours and lots of rebels, an early game mini-profile which is extra-aggressive towards rebels. W

With the settings I'm currently using the AI defends the frontline (defend_fortified) unless...

Will defo try those out.

Currently going backwards to try and nail down some actual numbers and using a stripped-down testing mini-profile where it is easy to change all the settings. I tried the invade_oppurtunistic idea but very weird things happened so i thought it was time to stop editing my main profile and write this instead as otherwise i'm probaly getting wrong results through typos etc.

I want to see if i can pry out some approximate numbers relating to the offsets and also to distance at least.

<faction_ai_label name="testing">
	
<defend_decisions>

<decision_entry>
<min_entry	target_faction="slave" is_neighbour="true"/>
<max_entry	target_faction="slave"/>
<faction_attitude	at_war="true" defense="defend_minimal"/>
</decision_entry>

<decision_entry>
<min_entry	target_faction="slave"/>
<max_entry	target_faction="slave"/>
<faction_attitude	at_war="true" defense="defend_minimal"/>
</decision_entry>

<decision_entry>
<min_entry	is_neighbour="true"/>
<faction_attitude	 defense="defend_deep"/>
</decision_entry>

<decision_entry>
<faction_attitude	defense="defend_minimal"/>
</decision_entry>

</defend_decisions>


<invasion_decisions>

<decision_entry>
<min_entry	target_faction="slave" is_neighbour="true"/>
<max_entry	target_faction="slave"/>
<faction_attitude	invade="invade_buildup" invade_priority="820"/>
</decision_entry>

<decision_entry>
<min_entry	target_faction="slave"/>
<max_entry	target_faction="slave"/>
<faction_attitude	invade="invade_buildup" invade_priority="800"/>
</decision_entry>


<decision_entry>
<min_entry	is_neighbour="true"/>
<faction_attitude	invade="invade_none" can_force_invade="false"/>
</decision_entry>
		
<decision_entry>
<faction_attitude	invade="invade_none" can_force_invade="false"/>
</decision_entry>
			
</invasion_decisions>
		
		
</faction_ai_label>

I've never tried has_cease_hostilities, it may work. Honestly I've not specifically tested has_ceasehostilities, but I have done numerous math tests on invade priority (which is affected by has_ceasehostilities) and they always work out numerically according to the LTGD log's invade priority.
I will specifically test both shortly.

I'll have to re-test it. Quite possible I've made a mistake.

Speaking of which one of my previous "seems to be" theories wasn't right. Having all the defend settings the same e.g defend_deep doesn't make them merge more. Just a side-effect of me mainly watching one faction at a time when testing. Factions with roughly circular terriotories centred on their capital can be made to merge almost 100% but it's more a function of the basic geometry it seems.

This all fits in with my experiences, on #4, the AI will definately attack multiple targets if the invade_priority for each is within a cirtain range, that range depending on the precived strength of the considering faction.

Diferent regions priorities with the base being the invade priority for the target faction is easily seen in the LTGD with naval invasions your given the final invade priority for the individual target settlements. I have a list of what province is what number for the vanilla map somewhere, I'll see if I can find it if anyone wants it.

Glad you mentioned that. I noticed those priorities being adjusted when i skimmed the log but i hadn't thought about using them to give me a baseline for the sort of number ranges the game is using. I could use a set of them to help figure it out. I'd definitely find a list of region id's useful if it's easy to paste here.

Currently it's looking like the ranges are quite small, (need to test a lot more specifically to be sure), but drops in 10s or 20s seem enough to significantly alter the distance the AI will mount naval landings.

Apart from anything else it does I'm thinking the invade priority acts as a kind of "desire filter" so, if the numbers being used by the game to decide the final invade priority are quite small, then too wide a variation in the set of invade_priorities in the AI file may end up having too exaggerated an effect i.e too aggressive or too passive.

Will see. I'm going to pick some test cities and use my simplified profile and try and nudge the invade priority up and down until i get roughly the number that is the minimum for a faction to attack then try and change settings to see if i can get some rough numbers. Will probably try a naval landing one as i can use the priority offsets in the log file as clues.