AI totally downgraded in Beta 1.1.0

Users who are viewing this thread

Dabos37

Sergeant Knight at Arms
I have been making some tests to compare the AI, and It is pretty evident how bad and downgraded the AI is in 1.1.0. In 1.0.11, the NPCs are able to block effectively most of incoming attacks, plus they are able to feint the player and fighting some elite units could be really challenging.

In beta 1.1.0, NPCs just attack, attack, and attack without any sense and rarely block incoming attacks and never try to feint. Please do not make public the new AI without fixing It because now It is totaly downgraded and It is frustrating to see how bad the AI is. I will be playing 1.0.11 these days because the 1.1.0 AI is simply not enjoyable at all if you like to play tournaments and small battles like hideouts.
 
Man, it's almost as if the beta branch was exactly for the purpose of trying things out and gathering feedback... Really surprising it would have issues, isn't it?
 
I like the new AI for the lower tier units, it makes sense. Looters shouldnt be good at combat. However t3/t4 that are professional troops need to have their improved AI skills back.
 
This is just a matter of balancing. In beta 1.1 AI fighting ability is linked to their skill in given weapon. There's only a matter of choosing the right skill lvl at which AI will start to be good. For example lords are fighting pretty well on 1.1 (both on foot and horse) but they have skills around 230 while most high tier troops are around 130.

So there's this huge gap between lvl 130 and lvl 230.
 
This is just a matter of balancing. In beta 1.1 AI fighting ability is linked to their skill in given weapon. There's only a matter of choosing the right skill lvl at which AI will start to be good. For example lords are fighting pretty well on 1.1 (both on foot and horse) but they have skills around 230 while most high tier troops are around 130.

So there's this huge gap between lvl 130 and lvl 230.

No, even with 240 weapon skill, Lords are still fighting worse than before. They are worse at blocking and never use feints to try to trick the player.

Man, it's almost as if the beta branch was exactly for the purpose of trying things out and gathering feedback... Really surprising it would have issues, isn't it?

Sure, I am just giving feedback. Do not make me wrong, I love this game and I love TW and I am not blaming them. My biggest concern is that maybe TW downgraded the AI to try to fix other things (maybe for improve performance or something like this) and It would be really bad if the decide to keep the AI as bad as It is and release this awful AI to public.
 
Last edited:
They should have several AI settings for the AI. One for small battles, like 50vs50, and one for large battles, like 400vs400 and one for sieges. I'm pretty sure, calculating all the moves and attacks and blocks for 700+ AI is very CPU heavy, especially hitboxes, grazing hits, did the sword hit or went it an inch next to the hitbox? ... and so on ... so for large battles it should be toned down. For small ones it should be full AI capabilities, since less stuff is to be calculated. And in tournaments, they should give the player hell.

I'm pretty sure Taleworlds will work something out. You simply cannot expect in large scale battles to have full usage of ALL AI capabilities, this would FRY your CPU. And as stated already, it's currently "beta", so, testing will be done.
 
They should have several AI settings for the AI. One for small battles, like 50vs50, and one for large battles, like 400vs400 and one for sieges. I'm pretty sure, calculating all the moves and attacks and blocks for 700+ AI is very CPU heavy, especially hitboxes, grazing hits, did the sword hit or went it an inch next to the hitbox? ... and so on ... so for large battles it should be toned down.

It should be significantly simplified for everything outside the player‘s viewport. The key is to have animations run smoothly from start to finish. Why would you over-simulate what’s being done in the courtyard the player can’t see so long as animations give a believable view when the player comes around the corner and then observes it? You can simplify outcomes using an algorithm beforehand and then simply prepare the animation queue beforehand—play the dance at the proper dt as needed and go from there. The player only needs to think the individual units are making independent decisions on swings and blocks and that somebody won a ‘fair“ fight. The only reality is the player‘s perception.

If you walked into a field strewn with dead and a few guys still struggling to kill each other, your brain will fill in the story. It doesn’t need to be simulated to be “real“.

Goes with that silly requirement to talk to looters (anyone) to surrender. The expected outcome should be pre-computed. If the outcome is not in question, then don’t initiate the dialogue. Why wait for the conversation menu to build the outcome model?
 
Last edited:
They should have several AI settings for the AI. One for small battles, like 50vs50, and one for large battles, like 400vs400 and one for sieges. I'm pretty sure, calculating all the moves and attacks and blocks for 700+ AI is very CPU heavy, especially hitboxes, grazing hits, did the sword hit or went it an inch next to the hitbox? ... and so on ... so for large battles it should be toned down. For small ones it should be full AI capabilities, since less stuff is to be calculated. And in tournaments, they should give the player hell.

I'm pretty sure Taleworlds will work something out. You simply cannot expect in large scale battles to have full usage of ALL AI capabilities, this would FRY your CPU. And as stated already, it's currently "beta", so, testing will be done.

I just expect that AI continue being at least as capable as It is currently in 1.0.11, I am really not expecting anything more (even better AI would be welcome though). The current 1.0.11 performance is not much worse than in 1.1.0, so if the reason to downgrade AI would be performance and to get some extra fps (we are not really sure about It), It is as simple as add a warning in Challenging difficulty saying that It could decreases performance, but challenging difficulty should be actually challenging but now It is simply too easy.
 
Man, it's almost as if the beta branch was exactly for the purpose of trying things out and gathering feedback... Really surprising it would have issues, isn't it?
Man, it’s almost as if this thread was made exactly for the purpose of giving feedback, in the hopes that it would be gathered by TaleWorlds. Really surprising that people are pointing out the issues in the beta, isn’t it?
 
This is just a matter of balancing. In beta 1.1 AI fighting ability is linked to their skill in given weapon. There's only a matter of choosing the right skill lvl at which AI will start to be good. For example lords are fighting pretty well on 1.1 (both on foot and horse) but they have skills around 230 while most high tier troops are around 130.

So there's this huge gap between lvl 130 and lvl 230.

But even lords with their massive skills fight like zombies and just spam attacks, never blocking lol
 
In the beta i have not lost a single tournament fight. The 1v1 a.i cannot block or feint even on challenge. Even high tier infantry or companions are just spam attack with zero block except for the odd one if he has a shield. Before beta i had to shield bash, feint attack because the a.i on good combat was actually decent in 1v1.

It's a complete downgrade in beta they lost the ability to block.
 
Yes, the 1.0.11 AI would wipe the floor with the 1.1 AI, but they should be aware of this so I expect a fix before the beta goes live.
 
I thing both options should be available as there is quite a few people complaining about AI being to hard to fight against, and now quite a few people are complaining about them being too easy.
 
I thing both options should be available as there is quite a few people complaining about AI being to hard to fight against, and now quite a few people are complaining about them being too easy.

Well no, in the 1.1.1 beta there is a new difficulty slider with the AI ranging from easy to hard.

The problem is all three settings are too easy, and easy at a level that I doubt anyone wants where just about any attack on the AI is guaranteed.
 
Back
Top Bottom