AI not aggressive enough on hard units

[Aekar]

I waited to get to Act Three before validating that fact.


In all situations, the enemy AI goes cautious around player super units (heavy tanks, Titans).
That translates into the AI not attacking them, as it can receive many damage without causing enough.

This results into a very odd behaviour.
Like Gargants never using their Deth Kannons (140 STR, 40 % Pen, 2 shots) even against Titans (which has 130 Armor) whereas it could hit it.
On the maps, this makes Baneswords, Shadowswords, Titans masters of the field of battle, but in an odd, "never attacked" way.
Such Gargants are shot from afar to their death and would never even take a single action, even when Titans are just 1 hex from them and the Gargants have lots of support units also able to hit it.
This results in very unrealistic, unbalanced situations, whereas the whole game tries to bring that theme into action.


As the battles actually involve a lot of Titantry, I strongly believe the AI needs to be patched up.
It should take into account the number of Gargants it has and decide to commit some of them to attack.
Same goes for their anti-tank infantry, or anything able to "hit a giant armor", but take hits, and not kill it instantly.

[rezaf]

Yeah, I noticed the exact same thing - it's the same problem from the PzC AI where it wouldn't dare to attack experienced Tigers.
And it's worse, because for example Titans almost cannot be attacked without risking considerable losses.
In PzC this was an issue because overwhelming qualitatively often better units with sheer force of numbers was basically the soviet doctrine ... and in Armageddon it's even worse, because - as I understand it without much background in Wh40k - orks should be all about numerically vastly superior numbers?

An obvious solution would be to make the AI suicidal (or at least have a unit behavior setting for that), attacking without regarding the odds ... I'm not sure though it it's possible to implement something like that at this stage.
_____
rezaf

[Aekar]

For some reason, on "Act 3 - avenging the Fallen" scenario, the AI fires its Gargants and tanks on Titans and Baneswords.
Previously, on "Act 3 - Delivrance of Infernus", it did so, but rarely.
The AI has better chances of doing it when the Titans have a lower morale - certainly because they take, as you say, their losses into account, and they are reduced when the unit has a low morale.


I understand the AI may be common to all units, and doesn't try to be specific.
Slightly increasing the "damage tolerance" of the AI, or more, might be very good for the player's experience.
As of now the Ai may stand idle across half the episodes of Act 2 and some of Act 3, as long as you only expose Titans and heavies to it.

I do not say it shouldn't be expected in some cases (always fun to see enemies trying to charge past the Titan without firing at it, and the Titan being overwhelmed) but an adjustment looks required.

[rezaf]

Well, since the AI basically exposed this very behavior since PzC 1.0, I'm not sure how likely this is to change anytime soon...

I feel this leads to a behavior almost opposite of what a human player does. Well, at least what I do. When confronted with such situations, I'd much rather use my most powerful units to weaken the foe and then other powerful units to hopefully destroy him or at least damage him to a considerable degree. Now, of course there are odds I'll not be willing to take, but the AI has a very low threshold when it comes to making this call.
In PzC, this leads to it strafing infantry units with it's best fighters instead of trying to gain air superiority by ganging up on player fighters or even bombers. However, the flip side of the coin is, if the AI were to do that, you'd be faced with many many complete unit losses, given the vast numerical superiority the AI usually brings to the table.
In Armageddon, the fact that taking counter damage is so common with most everyone having ranged attacks, this seems to often lead to complete AI paralysis. I don't know how it was in your second and third acts, but I know I had a lot of AI turns where not a single attack was mounted or unit moved.
_____
rezaf

[Aekar]

I had a lot of AI turns where not a single attack was mounted or unit moved.

Had that a lot. (meaning: most of the time)
Especially as long as only heavies and Titans were in sight.
Once a weaker unit got in sight, the AI sometimes called its reserves in and rushed this unit to death.
Otherwise it stayed idle.

It acted more on Act 3, though.

I believe this AI behaviour may be soft-changed, by giving more penetration to some weapons.
What do you think about it?

[rezaf]

What do you think about it?

To be honest, my gut feeling is this isn't going to work. Because in PzC, the AI would even rather idle than attack if it WAS capable of causing considerable damage as long as it expected to take considerable losses in the process.
So I don't see how increasing weapon penetration could do something good.

In my opinion there's no good solution to this except recoding the AI to be willing to accept higher losses in it's attacks.
If you actually increase the power of it's units to make that happen, this is going to have side effects - for example, those units the AI IS willing to attack will be even more brutally decimated than they already are now.
And if you try to play "legit", by not investing heavily into ultra-heavy tanks and titans, this is going to make your life really miserable. Cause with everyone having ranged attacks, it's already hard enough to keep units the AI decides to go for alive.

I had mostly titans and a small number of ultra-heavies along with maybe half a dozen space marine units which I tried to keep out of harms way in the late game, and I agree that this is the main catalyst for AI paralysis.

But this is all, in my opinion, a fundamental flaw in the design for this game - one that isn't easily resolveable on a whim. Namely, I'd expect hordes of fairly weak, but numerous Ork units vs. higher quality, but far less numerous units of the Space Marines / Steel Legion in this conflict. But rather than fixing the AIs "thought patterns" early on and designing such scenarios, the old "beaten path" was chosen, by giving Orks a very large number of powerful units as well as a moderately large numbers of fodder.
In Act1, this doesn't lead to the gameplay I expected to see, but still results in a fun and for the most part challenging but not frustrating experience.
In Acts 2 and 3, though everything falls apart and the approach to designing the scenarios reminded me of the late soviet DLCs. Except, this time around you can - at your leisure - fill your core with KarlGerät mortars that hit even harder, have considerably larger range, can move further than some bombers, have more defense than a Maus and are also incredibly hardy, making them almost the only unit (along with true artillery) that can reliably amass experience in the game...
_____
rezaf

[Kerensky]

Just to chime in.

One of the features I lobbied early on for this game was for an extra AI setting to exist in this game, something along the lines of 'kamikaze' or 'berserk' along side things like 'attacker' or 'hold position active'. It was designed to be put on units, especially Ork units, to force them into the arms of death freely and with total abandon. Force them to attack units they would normally be too timid to engage. They move and attack anything they can reach, regardless of how bad their odds are. Obviously this would have been a carefully used setting, because if too many units have this it makes the AI look really stupid and that is always a concern in these kinds of games. It's purpose would be to prevent the 'roadblock strategy'.

In Panzer Corps, this mean sitting something with extremely high defense rating, like a Elephant or Tiger II in the middle of a road, and watch the AI completely flounder trying to get around it or even deal with it at all. At the same time, the Tiger II doesn't even need to expend ammo firing at targets because it does it's job of disrupting the AI advance down a choke point. It is not attacked by the AI because it's really hard to fight a Tiger II in open ground in Panzer Corps, especially given historical Allied equipment. Combat may be ranged in 40k, but roadblocking just takes a new form because there are some really durable units in this game, and generally they also end up being pretty well armed units as well. They actually aren't as well armed as some proper Assault Infantry (15 man team of 2 weapons with 2 RoF each is a lot more attack dice thrown out than a 2 tank team of 2 Shadowswords), but they are leagues more durable and don't get vaporized when caught out of cover. Nor do they have their experience levels crash because of the constant replacements they require.

As it turns out, it's easy to talk about these things and just put in requests for extra features and AI settings, but that doesn't mean they are going to be done, especially when you are creating more work for other teams. Funny story, a friend of mine recently applied to become a Game Designer himself, and he was told one of the main reasons he didn't get the job was that he 'created excessive and unnecessary workload for other departments(code/art/UI) with his designs'. Bottom line, I can't program AI, don't have the tools to do it. I know the AI well enough to understand how it thinks and can set up units accordingly, but that is a far cry from being able to create an entirely new AI setting. We made do with what we had.


Secondly, it's about unit representation. If we make a game of 300+ units, but more than half of the Ork units are so rare they show up a grand total of once or sometimes not at all (Where is Matilda II in PzC 1.0 campaign!?!!?), there's a problem. It's an even bigger problem when these are the 'cool' units who take many times more effort to create and configure than more basic and simplistic looking infantry. By a pure relative comparison there are about 15 Ork infantry unit types and something like 100 Ork vehicles. Their ID numbers go from 330 up to 468, but there are gaps in the numbers between unit types. I would estimate at least 50% of the Ork presence on any given scenario is infantry forces. 15 units have 50% force representation while 100 units share the other 50%? That is already pretty lopsided, and any further tipping would lead to unacceptable rarity levels for many of the most awesome and powerful Ork units.

We're talking big and awesome tanks with multiple weapon systems and weapon effects and damage effects flying left and right in a wonderful display of sound and color. Yes the Orks are very well equipped with some pretty heavy duty equipment, but so is the Imperium of man, especially the lore behind the Steel Legion of Armageddon. We cannot have a lore backed faction notorious for it's over-abundance of heavy weaponry and equipment fight battles against poorly equipped hordes of fodder. That's not even mentioning the step and above quality of units the Space Marines bring to the battlefield with their heavy infantry who are basically walking tanks.


All that said, we are always listening to feedback. While we can't please everyone all the time, we can certainly take the lessons we learned with this brand new game and apply them forward to both further refine the base game and to propel future content to even more spectacular heights!

[JimmyC]

Regarding AI behavior, I thought it would be a somewhat simple algorithm based on expected damage caused and expected damage received. I have 0 modding experience and am just a casual gamer, but wouldn’t it just be a matter of changing the formula slightly so that the AI was willing to accept slightly worse attack odds?

I’m still in act 1 and am enjoying the game immensely. I’m going to try to go for a mixed force in acts 2 and 3 and hopefully that will keep the enjoyment up.

[Razz1]

The best way to implement it is to assign a trait like 'kamikaze' or 'berserk'. Then when a unit has that trait it uses a different formula script for combat.

So we have the method but no code or too much code to do it.

[rezaf]

Just to chime in.

Thanks for the recap Kerensky - interesting read.

One of the features I lobbied early on for this game was for an extra AI setting to exist in this game, something along the lines of 'kamikaze' or 'berserk' along side things like 'attacker' or 'hold position active'.(...)

You obviously didn't lobby hard enough.
Seriously, I'm not sure if this is the job of a map/scenario designer (I assume that was your main role?) to lobby for such changes. The person responsible for the core game design ought to realize stuff like that himself.
Now, I can't repeat often enough how highly I respect Rudankort - it takes a lot of effort and dedication to single handedly complete games, and he's done it multiple times now, and was basically the (almost) first and only one to try and recapture the Panzer General model and revive it in new commercial projects ... that's no small feat. That said, as early as in Panzer General Forever it was pretty evident that having an eye for small details such as this one wasn't his strong suit.
IIRC it was back then that the "improved" approach of the AI having to pay in prestige for it's repairs as well was implemented - which sounds like a perfectly reasonable idea on paper, but in it's implementation just lead to the AI spamming hordes of cheap defense units, which made many scenarios considerably harder to play (harder not because it was actually hard to dispose of them, but this took time and thus it became harder to complete the mission in the small Major Victory timeframe). And, I'm sure you will agree, this problem was never really solved in PzC, merely worked around by content designers such as you, mostly using scripting.

Now, I don't remember enough about Panzer General Forever to say if the AI aggression problem at hand goes so far back, but it definately was in PzC 1.0 and thus it's disappointing it was never adressed. However, it was a "running system", and maybe Rudankort was just worried such change might break something else - fair enough, I can accept that.
It is, however, much more disappointing that, given the chance to basically start from scratch an reimplement the engine from ground up, he chose to ignore the problem once more so it's likely it'll plague Armageddon and possible subsequent games for many more years to come.

Funny story, a friend of mine recently applied to become a Game Designer himself, and he was told one of the main reasons he didn't get the job was that he 'created excessive and unnecessary workload for other departments(code/art/UI) with his designs'.

Well, one could also say the moral of this story is the game designer himself has to make these realizations, what is strictly neccessary and what is just icing on the cake. In this regard, I think a wrong call was made.

Secondly, it's about unit representation. If we make a game of 300+ units, but more than half of the Ork units are so rare they show up a grand total of once or sometimes not at all (Where is Matilda II in PzC 1.0 campaign!?!!?), there's a problem. (...) That is already pretty lopsided, and any further tipping would lead to unacceptable rarity levels for many of the most awesome and powerful Ork units.

You are kidding me, right? That's your reasoning? Really?
I mean, are we even talking about the same game? Because, the one I'm playing has dozens and dozens of totally superficial units on the player side. I'm sure it's great for a Wh40k affictionaldo to have all these different variations of units, but while I'm sure they have a place in the tabletop, due to more intricate combat rules or special systems and so on, but in Armageddon, these units serve no purpose and just make life complicated for the player - because they make the haystack bigger in which to find the good-unit-needle. I never used 98% of the stuff you get access to when the Space Marines enter the game. All these new Steel Legion tanks, why should I buy them when I can get a titan instead? And even if I could not, I'd maybe pick one or two, at the very most three vehicles and stick with those. And don't get me started on the recolored SM squads or all their pointless vehicles, or the flyers that seem utterly pointless in this game ... and so on and so forth.
So, in a game that has so many stuff most players will likely NEVER use, it's a concern not to see some of the rare ork vehicles frequently used?

We cannot have a lore backed faction notorious for it's over-abundance of heavy weaponry and equipment fight battles against poorly equipped hordes of fodder. That's not even mentioning the step and above quality of units the Space Marines bring to the battlefield with their heavy infantry who are basically walking tanks.

But why not? That's the nature of the conflict you're depicting. Think about it like a game about the british fighting the Zulu, would you give the africans equivalents to machine guns and cavalry and stuff just so they are on a more equal footing with the british and the game can thus proceed in a more conventional manner?

All that said, we are always listening to feedback. While we can't please everyone all the time, we can certainly take the lessons we learned with this brand new game and apply them forward to both further refine the base game and to propel future content to even more spectacular heights!

Well, I've been there since PzC was just released, and thus I know the listening to feedback is mostly handwaving - a few examples were outlined in this very post/thread.
Now, I'm certain there will be plenty post release support and some breaking changes, but I know from PzC that many changes are just out of the picture because they would break existing content.

But who knows, maybe I'll be proven wrong in my concerns. I'd be very happy if that were the case.
_____
rezaf

[orci]

Just read this thread yesterday night.
After thinking about it I realized I have the same problem. I am currently in Act 2.

For me the main problem is that if you move a weaker unit to front it will be shredded into pieces. Even tough Space Marine infantry in cover will die because the whole enemy army shots at one of my units.
Had major Steel Legion infantry losses in last scenario while my titans and super heavies remain nearly unharmed.
I also realized that I had adopted my play style ("Hey Reaver just go ahead") due to this.
Compared to Act 1 this is really different and honestly I enjoyed playing without the super heavies and titans much more due to the diversity of combat.

I have a programming background and my opinion is that it is possible to add some new traits like Kerensky proposed in a way they don't interfere with the rest of the AI. It could be done.
So I would really like to have an adjustment here. Especially Act 1 is very very good to play.

[JimmyC]

Fingers crossed you are right and they can make it a bit more interesting by making some of the orcs actually attack Titans

[Skanvak]

Ah yes, that was something I was going to post too now that I have finish the campaign as designed. My problem with the "berserk/suicidal" unit setting is that it is wrong in most scenario. I mean if you attack a very strong unit and suicide lots of units doing so, it will only makke the game easier for the attacker. The point is that the AI should take into acocunt other variable like :
_ its own level of HP as a Gargant/Titan will not lose power by losing some hp so the -4/-3 exchange is not that bad.
_ the danger level / high level of hp the ennemy unit. Killing the enemy Titans can worth some lose as the titan is very dangerous and wearing it out is a valid strategy.

The best way should be to make the AI thiink like in chess in term of excahnge and final result of the exchange. May be a hit on a titan should value more than a loss on a cheaper unit. For the Gargant I have some difficulty as I do refrain from attacking Gargant unless I can deal at least half it pv or my hp is high too.

Another thing about the AI is that I would like to have several level of different AI. Ie this AI consider and easy and a better AI consider are difficult. Because asking for a better AI may make the game too difficult for some people.

One last thing : make some move random so that the AI cannot be predicted. Her I see the AI stun or doing only the garantee kill. This is kind of boring. I would like some surprise from time to time and that can only be done by doing strategy mix (random choice of tactics in the same situation).

[Aekar]

I have a game programming background as well and I approve your points.
These elements are in the right direction.
I still hope the devs would be able to look into it for the third patch due for the end of this month ...


@ rezaf :
Your last post was a bit intense
I understand the good point in it, is that it gives to the devs a better/larger spectrum in terms of feedack.
And you are a Lt Col (forum-wise) so an old timer!
I hope they did not mind the intensity of it too stongly, because there surely are good points there too.

[rezaf]

Your last post was a bit intense

Maybe you're right. I'm a man of clear words - even in RL - and in forums this may mean that I sometimes dance one the border to trolling like the Gleiwitz incident that started WW2 was dancing on the border of war reason fabrication.
To my credit, I TRIED to balance things a bit by including some praise for Rudankort, but, yeah, maybe it was a bit overboard. English isn't my first language, so I hope at least there's no namecalling?

If any portions of the post are particularly offending, point them out to me and I'll remove them.

My apologies. At least, if I didn't care about Armageddon (and Panzer Corps), I wouldn't even waste time to write such posts, but I guess that's not very comforting.
_____
rezaf

[Kerensky]

I personally took no offense, all I see is passionate debate.

[Aekar]

I think you were right to "shake the dust off the carpet", as english people may say.
I would not know how they say it, as english is a foreign tongue to me as well.

I believe the devs need to know that act 2-3, hence half of the whole title, might have suffered a bad gameplay because of "a few missing lines of AI code".(there may be more to that of course, and we are ill placed to know; but it also possibly is that). Considering it, there is always room for passionate reactions

What is serious is that a lot of work is put on the whole title, and a few things like this can but the situation and the balance off.
What is clear, though, is that this is what happens in video games all the time - this is a world where tiny things can make the whole lot tremble, and a world were most devs are also oblivious to the problems.
Fortunately not, I would hope, in this case.

The only good reasoning behind this is that problems need to get fixed, for all the rest of the artistic and technical work to shine as they should shine.
Right now you could have two dozen maps, three dozen enemy units on each, but if AI units stay idle 90 % of the time because they just "fear the Titan in sight", there is an obvious problem that needs to get addressed.
This is an Orc Horde, that I shall fear as a player, not a piece of cake situation. (EDIT: and I just cannot imagine massive ork armies and vehicles sitting around as they get shot the whole day)

I suppose patch 1.02 will balance things better though - the Scout Titan is much weaker, so Act 2 gameplay might just rock in 1.02.

[Falke_MatrixForum]

In my current game the AI is a lot more aggressive, but this will be due to me restricting my force to Blood Angels Units only (From Act2 onwards)