Users who are viewing this thread

Well if it's that easy I am sure the devs will fix it in no time! I just know, based on my own experience, that things are rarely as easy as you think when it comes to fixing complex systems. But since neither you nor I are part of the developement team we shall have to agree to disagree and see what they come up with.
Lets hope for the best. Good AI is very important for M&B, so I hope they'll pay close attention to it.
 
Lets hope for the best. Good AI is very important for M&B, so I hope they'll pay close attention to it.

No argument there! Combat is the core gameplay feature that sets this game apart from the others and AI quality will make it or break it (together with believable physics, but that is a conversation for another thread).
 
would love to see a tournament AI programmed to fight like a spastic high level player with massive amounts of feigning, chamber blocking and movement.

 
Nice to see that a lot of people are noticing this issue and creating threats about It. It is a pretty annoying issue, even more when this works perfectly in stable version. I know that I am playing a beta build, my only concern is that if this issue wont get fixed before this build will get released as stable version.

On the other hand, It is important to mention that this issue is not just happening in tournaments, I can also notice it in hideouts and in big battles (in big battles is almost noticeable though).

Please bring back the AI that we have in version 1.0.10 or make It even smarter and harder. It is great to have intense 1v1 fights.
 
What I you talking about? Most of fixes are change a couple of numbers in configs or add a couple of lines in code.. Example: AI weapon selection was bugged, and they never switched weapons. I just rewrote the whole logic from scratch, took literally 15 minutes

image.png



Of course, there are situations, when the whole system is to be refactored. In that case, you don't want to fix bugs in the current system, you want to fully focus on a new one, and it could take months. I doubt they work on completely new AI, but who knows..


While I appreciate you trying to illustrate with code, saying it is 10-20min job is not realistic. This is a quick, if-else dirty hack fix, but if one were to do this properly:
  • Some different architecture than if else if, else. Looks very rudimentary for an AI system. One can say it is not even AI.
  • Writing the unit tests or end-to-end tests for the requirement. However, maybe TW writes tests, maybe not, maybe a bit. Game developers probably don't test much, since you know, just games. If they don't they should, since it acts as a safety harness for refactoring.
  • In the case of Bannerlord, changing this one small behaviour for 1v1 could have big impacts for a 1000 troop battle. That needs to be tested.
  • Going through the CI/CD pipeline and QA team loop. Maybe where you work you can just write 2 lines and deploy. But some places have to go through other dev processes before deploy.


    All of these takes time and nothing really takes 10-20mins if you were to do it proper. But you could be somewhat right, it may be a small 1-2 hour job and TW is not prioritising it.
 
  • In the case of Bannerlord, changing this one small behaviour for 1v1 could have big impacts for a 1000 troop battle. That needs to be tested.

Sure, I agree with this. On the other hand, the AI changes in 1.1.0 are already affecting my troops performance in Hideouts. My troops used to be pretty tanky in Hideouts but now they die easier.

Maybe this new dumb AI was introduced for a mistake or maybe It was introduced to try to fix the issue where some units cannot attack because their weapon hit the team mate shield. If It is the case, TW please try to find a more elegant solution different to make the AI awful as It is now un beta 1.1.0.
 
It seems like 1.1 broke the ai, I thought they di ok before but they've completley forgotten to block.

Even more pressing though is how they still don't know how to turn. You can literally just strafe in circles around them spamming attack and they'll just sit there, neither blocking or attacking, making a futile effort to turn and face you.
 
I have not much problems with the circle abuse yo be honest. The solution is as simple as do not abuse the AI but if TW could improve It, better for us.

On the other hand, the issue about AI refusing to block is something which should be fixed ASAP. The 1v1 combat is simply not fun anymore. Plus, my troops are less tankier than they used to be in Hideouts.
 
While I appreciate you trying to illustrate with code, saying it is 10-20min job is not realistic. This is a quick, if-else dirty hack fix, but if one were to do this properly:
  • Some different architecture than if else if, else. Looks very rudimentary for an AI system. One can say it is not even AI.
  • Writing the unit tests or end-to-end tests for the requirement. However, maybe TW writes tests, maybe not, maybe a bit. Game developers probably don't test much, since you know, just games. If they don't they should, since it acts as a safety harness for refactoring.
  • In the case of Bannerlord, changing this one small behaviour for 1v1 could have big impacts for a 1000 troop battle. That needs to be tested.
  • Going through the CI/CD pipeline and QA team loop. Maybe where you work you can just write 2 lines and deploy. But some places have to go through other dev processes before deploy.


    All of these takes time and nothing really takes 10-20mins if you were to do it proper. But you could be somewhat right, it may be a small 1-2 hour job and TW is not prioritising it.
unit tests? CI/CD? QA Team? I don't think you're paying attention bud, Bannerlord uses write and pipe methodology! They're coding up what is getting pushed as a hotfix today and we're the testers.
 
unit tests? CI/CD? QA Team? I don't think you're paying attention bud, Bannerlord uses write and pipe methodology! They're coding up what is getting pushed as a hotfix today and we're the testers.


I am honestly happy they use us as testers, so they can focus on features and fixes. But I really find it annoying after such a long time the AI is still broken as hell. I don't understand it. Maybe it is hard job, maybe not, maybe it is not their current priority, maybe its Maybeline.
 
To be honest, the army AI in open battlefields is pretty decent and I have had really hard battles and even lose sometimes against similar forces. Sieges AI needs more work and 1v1 AI (individual NPCs AI) has been really enjoyable in 1.0.X, my duels in 1.0.X have been great and if we do not take into account the circle abuse, It is probably the best AI I have seen in duels in all the games I have played.

I am really concerned on this because while I think that TW is probably working on improving the AI, I am afraid if even after fixing It, the combat wont be as challenging as It is currently in 1.0.X (I am jugding by looking how NPCs with high amount of weapon skill are not as good as It should). The challenging AI difficulty should be at least as good as It is currently in 1.0.X. More difficult is welcome, but not easier please!!!
 
1v1 AI (individual NPCs AI) has been really enjoyable in 1.0.X, my duels in 1.0.X have been great and if we do not

The 1.0.X bots were also way too easy. Have a look at this warband video of AI bot duelling, this is what I'm talking about for the hardest combat difficulty. Notice how the bot parries probably 99% of the player's hits? If Warband can make a bot like this, Bannerlord can too!

 
The 1.0.X bots were also way too easy. Have a look at this warband video of AI bot duelling, this is what I'm talking about for the hardest combat difficulty. Notice how the bot parries probably 99% of the player's hits? If Warband can make a bot like this, Bannerlord can too!


Dude it was mostly something in the newest patch i think they did something that broke it, it wasnt too bad last patch, they will surely come around to fix it, but its good you gave it attention on the forum
 
Dude it was mostly something in the newest patch i think they did something that broke it, it wasnt too bad last patch, they will surely come around to fix it, but its good you gave it attention on the forum
If I remember correctly, they implemented a new interface for selecting AI combat difficulty. I guess the interface is there, but AI always stays at easiest despite selected option. The current state of QA testing seems to be - if I can select this option and the game doesn't crash, ship it. Or may be there are auto-tests only. Nobody checks that feature actually works. Look at perks for example, half of them are doing nothing.

Yesterday I tested tournament combat without circling abuse, definitely too easy. You can just block and immediately attack, they will eat it almost every time.
 
Last edited:
If I remember correctly, they implemented a new interface for selecting AI combat difficulty. I guess the interface is there, but AI always stays at easiest despite selected option. The current state of QA testing seems to be - if I can select this option and the game doesn't crash, ship it. Or may be there are auto-tests only. Nobody checks that feature actually works. Look at perks for example, half of them are doing nothing.

Yesterday I tested tournament combat without circling abuse, definitely too easy. You can just block and immediately attack, they will eat it almost every time.


I really hope they add automated tests sometime, I can understand not now since it is launch and super busy. If they keep breaking things when changing and adding new things, well that's going to cost you more time over the long run.
 
It is good that it is getting addressed in the forums and hopefully someone has written a step by step bug report.

Now here is a simple workaround for everyone:
Stop cheesing the AI, by running counter-clockwise all the time :smile:
 
I really hope they add automated tests sometime, I can understand not now since it is launch and super busy. If they keep breaking things when changing and adding new things, well that's going to cost you more time over the long run.
For stability - sure, but writing auto-tests for gameplay is a complex task in itself, better leave it to QA and us. But please, make sure that the feature works before you ship it. Why make thousands of people grind a perk for hours, just to find that it literally does nothing? I can't agree with this approach. They could do super easy thing - say "NOT IMPLEMENTED" in text description, at least.
 
Last edited:
Back
Top Bottom