Invade and Defend Decisions

[z3n]

Defend Decisions

defend_minimal - do not distribute troops in that direction
defend_normal - equal distribution to frontlines and towns
defend_raid - equivalent of invade_raids in defense scenario (run around aimlessly in small stacks)
defend_frontline - defend outside on the frontline (see my main tutorial for info on frontline map command)
defend_fortified - defend in forts or on the borders
defend_deep - defend in or around towns

Invade Decisions

invade_immediate - invade ASAP (preferably at 1:1 strength ratio) from the frontline into enemy lands
invade_buildup - slower invade strategy than immediate, prefers to build up on the frontline to specified
invade_raids - small patrols invade enemy lands cause devastation but is too advanced a strategy for the high level AI
invade_opportunistic - invade targets that are or become weaker than ours even if it is not planned
invade_start - plan a future invasion
invade_none - don't plan invasions but can force invade via the exe unless otherwise specified

Notes:
Also note that invade_priority plays an important role in which forces invade where and when.
Cavalrycmdr has said he thinks that the invade_priority is offset by distance from a target, which can make or break naval invasions. And personally I agree as it seems like a pretty good explanation for some of the things that happen in the LTGD logs.

Invade_immediate also may affect naval invasions, although invade_buildup works too.

1) can_force_invade , turns on or off invasions ordered by the exe and overriding the decisions unless otherwise specified
2) See here for further reading

[z3n]

One thing I forgot.

There's absolutely no way I could find to circumvent the hardcoded defense plan. The AI will not leave a region to release a siege of another close by, no matter how close. It will only do that with separate armies classified ad free strength I believe.

[Titus le Chmakus]

I have this error and I cannot get where it comes from ?

[data.xml.tag] [error] File:[mods/Faerun/data/descr_campaign_db.xml]Error reading Attributes.(78:32)

[z3n]

The error is located on line 78 and in column 32. Press ctrl+g then type in 78 in your text editor (I assume it is notepad++).

[Titus le Chmakus]

ok thanks, normally the log says it more clearly !

[Piterx93]

You're wrong. Invade_raids is the best strategy. Attacks faster than invade_buildup and recruits a better army than invade_immediate. It attacks with large forces.

[bitterhowl]

And what about mysterious and mighty invade_opportunistic recruit profile?

[bitterhowl]

I want to share some old gems of "pre-reversal era" wisdom that I find.

https://forums.totalwar.org/vb/showt...omacy-AI/page2
Excellent research! CavalryCmdr and Nikolai1962 are really "AI prophets" from early modding era. Just read about defend_decisions comparement and unexpected effects of invade_start!

[bitterhowl]

An interesting tale told by LTGD log. "at_war" condition changes apply only at defend_decisions phase.

I'd like to reveal from engine reverse what's the meaning ofat_war entry.

[Dominick]

I want to share some old gems of "pre-reversal era" wisdom that I find.

https://forums.totalwar.org/vb/showt...omacy-AI/page2
Excellent research! CavalryCmdr and Nikolai1962 are really "AI prophets" from early modding era. Just read about defend_decisions comparement and unexpected effects of invade_start!

Thank you for this link, I read most of the thread and it gives a rather good image of how the campaign AI actually functions. Considering how much time has passed, it felt like going through a historical source. This game really is an unpolished gem, I wonder how much is still possible, that we aren't currently aware of.

[Macaras]

In the vanilla campaign ai I found an entry as following:

<decision_entry>
<!--
If we have a cease hostilities mission against the target, 50% chance to use defaults
(this will reduce chance of factions attacking some targets if it will cause excommunication)
-->
<min_entry has_ceasehostilities="true"/>
<max_entry rand="0.5"/>
</decision_entry>

Do you know if it is working correctly?

Anyway, sorry for double posting, I received an answer here:

https://www.twcenter.net/forums/show...1#post16076803

[bitterhowl]

I want to share some old gems of "pre-reversal era" wisdom that I find.

https://forums.totalwar.org/vb/showt...omacy-AI/page2
Excellent research! CavalryCmdr and Nikolai1962 are really "AI prophets" from early modding era. Just read about defend_decisions comparement and unexpected effects of invade_start!

During ongoing testing of my upcoming Westeros release I revealed that defend blocks are more important than invade. Had my Ironborn just sitting around their isles. Changed some blocks in defend part - now they conquer the entire map at 250 turn.

My hint is that defend_frontline has not defend purpose, but invade one. It helps to accumulate invading force on border increasing frontline_balance to necessary value.

Old article says that only defend_normal has good performance at defending actions, because frontline and fortified plans are ineffective for sieging enemy and defend_deep is too passive.

[z3n]

Here's some other tidbits from CavalryCmdr about it, I've found it to be true as well, so my favoured defense decision in any war is defend_frontline

Quite useful, can be used with “invade_opportunistic”, “invade_immediate” and “invade_buildup” with great effect. As it implies the AI will attempt to match up against the target faction on the frontline. Army size determined by invade type and nearby enemies. This decision type uses “invade_priority” to determine whose frontline to line up on. Note: One weakness is it will not “recall” these troops to intercept/lift siege unless they can arrive in one turn. DO NOT use against non-neighbouring factions.

https://medieval2.heavengames.com/m2...ials/ai_essay/

[Macaras]

Thank you for sharing this guys, one can learn something. I think its simple but very important..

[bitterhowl]

Some updates from CavalrtCmdr in another thread here, that invade_opportunistic with can_force_invade=false can make great naval invasions. Haven't tested it myself, but included it into new CAI for mod that porting Rome 1 to Medieval 2, for Romans to invade Carthage, and mod author says it's going good. (He released the mod just some hours ago).

Another thing I have to say, that military_balance_plus_enemies is very interesting parameter for the sake of making AI behaviour more realistic and smooth. I'm using it to make additional reasons for AI to invade Neutral faction at the most preferable moment.

This parameter is sensitive to strategical interactions and seems it's only one which can have fast change - if faction is ok and hadn't crushing defeat in battle then military_balance between Neutral neighbouring factions wouldn't have big changes. But, when your neighbour starting war against third faction, then military_balance_plus_enemies may have dramatic change. So it maybe a moment to sudden attack, if you prepared it previously and gathered forces via defend_frontline, but waited for a good chance. Military_balance_plus_enemies changes when faction goes to war could mark such a moment to attack. I think it may be useful to script AI behaviour against human player, to prevent unprepared attacking of any neighbour.

[CavalryCmdr]

To specify, invade_opportunistic, can_force_invade=false
The AI can do a quasi naval invasion, that is if they have another reason to go to a region (by ship) they can then attack that region even without a bordering region or if that region is an island. It is very rare as there is little other reason for the AI to move into enemy regions except to attack. The reason I've seen them do it only with invade_opportunistic is because the AI is more likely to hire mercenaries, and will travel around to do so, the AI army was in that region to hire mercs and decided he had gathered enough of a force to attack, after he was already there, so it's technically not a naval invasion.

When I say very rare, we're talking a handful of times running 20 or 40 one hundred round AI only campaigns (I was testing different personalities at the time.)

Sorry if I wasn't clear in the other post.

[bitterhowl]

Got some strange AI behaviour during siege on player - lifting it after 1-2 turns even with good odds.