Quick Battle OP?

[loganjamesalex]

I have been finding that in the late game when I have 5 or more fully upgraded stacks that quick battle gets a much better result than I am able to, even when loading a save to replay the battle. Such as the fight at the end of Stoutheart 3, quick battle only loses a small handful of troops. It doesn't seem to do this in the early game and I can usually get better results then.

[Ail]

I'd put it like that:

In wielder-battles auto-battle (and thus also quick-battle) does a fine job, whereas in battles against neutrals it looses troops unnecessarily.

According to the AI-dev the AI has 5 different levels of aggression that it situationally pick.

From my observation this depends on an evaluation of which side is better at magic and the army-composition of ranged vs. melee-troops of both sides.

If the quick-battle result is better than your manual-result, I recommend choosing manual but then letting it auto-play in order to learn how the AI plays differently. As of 1.1.2 the AI has a very good understanding of what spells are good in what situation. I'd suspect that it simply uses its essence more efficiently than you do. Maybe because you have some misconceptions about what certain spells do and how useful they are.
It is also good at picking what kind of approach it should use.
It is also really good at locking down ranged troops with melee and deciding whether it's worth being melee-hit to move to a certain tile or not.
So outdoing it in these kinds of battles requires a lot of experience and knowledge about all the spells.

In battles against neutrals, the presence of ranged-troops on the enemy side triggers it to be more aggressive than it should be. I've seen it expose it's own ranged troops to those of the enemies for a -50% damage shot, to then get dealt 100% by the enemy. This was completely unnecessary and they could have forced the neutrals to move in instead, simply threatening winning by default due to being a wielder who gathers essence and could theoretically win anyways with spells. No need to put it's ranged-troops out of position to take that -50% shot.

[Wirtzak]

Originally posted by Ail:

I'd put it like that:

In wielder-battles auto-battle (and thus also quick-battle) does a fine job, whereas in battles against neutrals it looses troops unnecessarily.

According to the AI-dev the AI has 5 different levels of aggression that it situationally pick.

From my observation this depends on an evaluation of which side is better at magic and the army-composition of ranged vs. melee-troops of both sides.

If the quick-battle result is better than your manual-result, I recommend choosing manual but then letting it auto-play in order to learn how the AI plays differently. As of 1.1.2 the AI has a very good understanding of what spells are good in what situation. I'd suspect that it simply uses its essence more efficiently than you do. Maybe because you have some misconceptions about what certain spells do and how useful they are.
It is also good at picking what kind of approach it should use.
It is also really good at locking down ranged troops with melee and deciding whether it's worth being melee-hit to move to a certain tile or not.
So outdoing it in these kinds of battles requires a lot of experience and knowledge about all the spells.

In battles against neutrals, the presence of ranged-troops on the enemy side triggers it to be more aggressive than it should be. I've seen it expose it's own ranged troops to those of the enemies for a -50% damage shot, to then get dealt 100% by the enemy. This was completely unnecessary and they could have forced the neutrals to move in instead, simply threatening winning by default due to being a wielder who gathers essence and could theoretically win anyways with spells. No need to put it's ranged-troops out of position to take that -50% shot.

You sure have studied the AI and been keeping yourself well informed with the AI updates. Thank you for informing others in these forums. Your replies are really well nuanced and helpful to the community :D

[Ail]

Originally posted by Albin (Lavapotion):

You sure have studied the AI and been keeping yourself well informed with the AI updates. Thank you for informing others in these forums. Your replies are really well nuanced and helpful to the community :D

I wish I could study the AI even better. Game-AI is kinda my hobby and I've worked on the AI for several open-source games.
It's always kinda sad to me when a great commercial game comes out and I feel so much temptation to work on it's AI but can't.
So if at any point in the future you'd be fine with me having a branch of your repo to experiment on, I'd love to do that and provide you with whatever results I'm achieving.

For reference: Here is my github and the project I'm currently most active there is the OpenXCom one.

https://github.com/Xilmi

[LexGeneralis]

But...still to answer the OP: if AI does - even with you having full stacks - a better job in battle, there's a big skill gap for you.
An expert player gets out with zero casulties, while AI would often loose a few units even if it is considered to be a simple threat.

[loganjamesalex]

Originally posted by LexGeneralis:

But...still to answer the OP: if AI does - even with you having full stacks - a better job in battle, there's a big skill gap for you.
An expert player gets out with zero casulties, while AI would often loose a few units even if it is considered to be a simple threat.

I don't really see how any player could get out of the final battles in some of these campaigns with 0 casualties, I would like to see that. Also I forgot to mention I am playing on 'Worthy' difficulty so far. I didn't know I could give control to the AI during a battle and watch, I guess I will have to do that.

[LexGeneralis]

Not against other wielders, that's true!
Although often there are opportunities to deal at least more casulties to the enemy and sometimes still win a fight if auto battle can't see a way.

[loganjamesalex]

OK I watched some videos and I can see that there is (or at least used to be) a lot of exploit-tier things you can do with stacking spells, so I can see there is a lot more depth you can learn