Adventure map, AI hero doesn't capture a castle in just 1 step

[Branikolog]

Preliminary checks

• I've checked that there aren't other open issues on the same topic.

Describe the problem requiring a solution

Based on a feedback from a player.

AI stays just in fron of the castle, has significant army and hero strength advantage, but decides not to attack and run away.

2024-09-23.15-13-46.mov

What_1.zip

Describe the possible solution

I suppose AI just decided to proceed more urgent task in this case, but I expect enemy be able to waste a few movement points for castle capturing or hero attacking before he proceeds making his current task.

Additional info

No response

[LeHerosInconnu]

Hello @Branikolog,

This is the AI's strategy for forcing the human player to waste his gold on recruiting creatures to defend his castle, even though the AI never had any intention of actually attacking this castle. :D

Edit:
Cunning and vicious, almost pervert. 😎

[Branikolog]

Hi, @LeHerosInconnu !
It's one nice strategy.
There could also be anothe strategy in this case to stand in front of the castle, so player couldn't go out. :)

[oleg-derevenetz]

In this case, yellow heroes are rushing to eliminate the red hero that is considered as threating to two yellow castles at once. However, this red hero has a relatively weak army (just a few boars), so probably it's not really so dangerous, but the current threat assessment mechanism considers any enemy hero approaching the castle as a potential threat.

[Branikolog]

Even more ridiculous case, when AI ignores empty castle and seem to rush my weak hero. He even couldn't reach the hero but didn't want my castle.

2025-01-11.12-42-01.mp4

1 step for castle.zip

[oleg-derevenetz]

Even more ridiculous case, when AI ignores empty castle and seem to rush my weak hero. He even couldn't reach the hero but didn't want my castle.

This case is quite interesting. Here, Kalindra is aware of presence of Mira at tile 1449, and AI thinks that Kalindra will not be able to protect this castle against Mira in the short term, so this castle gets a penalty. On the other hand, it is always good to eliminate an enemy hero, and that's what Kalindra is really able to do here, so AI prefers to attack your hero.

[Branikolog]

AI thinks that Kalindra will not be able to protect this castle against Mira

But there's a drastic difference in army strength.
From a flexible human perspective the best solution here is to capture castle, leave some troops (Mira has just two troops and couldn't do much harm during the siege) and rush Mira. Otherwise Kalindra could rest in the castle (+ mana) and get a movement bonus.
Even considering a generalized AI logic, he completely failed in this turn: neither castle was not captured (that could reduce enemy's income and where AI could possible hire more troops, get powerfull spells etc) nor the hero was demolished, since Kalindra didn't have enough movement points to reach it...

[oleg-derevenetz]

But there's a drastic difference in army strength.
From a flexible human perspective the best solution here is to capture castle, leave some troops (Mira has just two troops and couldn't do much harm during the siege) and rush Mira. Otherwise Kalindra could rest in the castle (+ mana) and get a movement bonus.

Mira is sitting in a boat with 12 Titans at tile 1449 and is ready to approach the castle at the next turn (you even paved a path already), and AI considers that Kalindra will not be able to hold against her, so it prefers to at least try to kill your hero.

[Branikolog]

@oleg-derevenetz
Ah, my apologies. I was confused by ?? marks in heroes names. 😅
Then the last case totally suits the initial issue: AI should have a priority boost for such close objects.

[oleg-derevenetz]

Then the last case totally suits the initial issue: AI should have a priority boost for such close objects.

Well, I don't even know, capturing the castle and even more so trying to keep it looks really useless here even from a human point of view, because it's just the guaranteed loss of a hero, but killing an enemy hero is quite a feasible thing. In general, there is a logic here, although of course a human player would have captured the castle, because he would hardly have paid attention to the distant boat with the enemy hero at all :)

[Branikolog]

@oleg-derevenetz

because he would hardly have paid attention to the distant boat with the enemy hero at all :)

Indeed. 😄

but killing an enemy hero is quite a feasible thing.

My greatest concern here is that AI couldn't reach the hero actually. So on one hand he could capture castle (reduce my income, hire troops, refresh mana, acquire some spells and get bonus while defending the castle) and on the other hand he just walked around and did nothing (the hero was not killed on that turn and I potentially won one more day for hiring troops in that castle).
If he succeeded in killing my hero... Well, that could have some benefits and I can agree with that logic more or less. 😅

because it's just the guaranteed loss of a hero

Mira couldn't reach him that turn because she need at least one more day for unboarding. So while enemy will be anyway threatened the next day by my strong hero (this could not actually happen, as something else could trigger me and I leave him and my castle for the sake of some other action) he could at least bring more inconveniences to my side.