New supply system is cool, but why are the devs adding new features when the AI cant handle the current ones?

[Question]

I mean, this is how the AI handles the current supply system :

That is 150+ french divisions killing itself in Africa due to a low level port, and the AI doesnt recognize this and will not move divisions away to fix the situation. Ive had to mod naval bases to give 1k supply per level to prevent the AI from killing themselves. The same thing happens when infra gets damaged from infantry sucide attacking a province over and over, which drops infara down to 0 in a week or less.

Now imagine how the AI is going to handle the new trucks, trains and fuel mechanics.

Im getting a distinctive sense of de ja vu here. When EU4 had the new Emperor DLC come out, it introduced a whole bunch of new buildings and new mechanics, the highlights of which was the new Holy Roman Empire (HRE) and estate system. A dev confidently told me on discord that in his testing, the AI could handle playing as the emperor of the HRE now. The DLC was heavily hyped and heavily anticipated by the players who were eager to get their hands on the new mechanics.

Within 2 hours of testing the DLC, I noticed that the AI was not coded to handle many of the new features properly. Some examples :

-The AI would spam the new buildings randomly without caring about focusing on its economy. Since many of the new buildings were niche ones (e.g. coastal defences), the AI just went bankrupt faster by spamming coastal defences all over the place. No improvement was made at all to make the AI go bankrupt less, and this isnt even hard from a coding perspective (i have modded the EU4 AI myself to have very good economies simply by priotising economic buildings, its really not hard).

-Many new mission trees were not designed to be AI friendly. For example, making a mission that require Austria to own a certain province will break the AI if the mission tree does not give claims on the province, as the AI will not know that they are supposed to attack the province to proceed in their mission tree.

-No attempt was made to address hardcoded AI behaviour unique to the HRE, such as the AI absolutely refusing to attack another AI if they are both in the HRE and are not rivals, even if their mission trees give them claims on each other. For example, giving Austria claims to proceed in their mission tree doesnt work when they are hardcoded to not attack the other nation because they are both in the HRE and arent rivals. Keep in mind that this was supposed to be the DLC that reworked the HRE.

-The AI was now hardcoded to NEVER keep a large standing army in peace time. So when war broke out, you could walk next door and destroy their army before they could raise an army to fight and the war would literally last less than 30 seconds. This was immediately obvious in the first 15 minutes of the game.

-The AI had no idea how to manage the new estates and would keep triggering infinite rebellions that would destroy their small peacetime army, and then the rebels would sit there forever because they did not have enough soldiers to seige down the castle, and then the AI country would be unable to do ANYTHING till they got annexed by someone. Again, this was immediately obvious in the first 15 minutes.

-The AI STILL has no idea how to play as the emperor of the HRE, so they just let half the empire remain as the wrong faith after the religious war which prevents them from passing any laws forever and ever...which is the exact same behaviour before the DLC, with zero improvements.

(This is nowhere near a comprehensive list of issues by the way)

So now, try to imagine how the AI is going to handle the new supply system. Just imagine.

Edit : I should add that currently, the AI absolutely HATES building to max infra even if they have idle civ factories. Im not sure why or if its even moddable. But this is very obvious if you tag switch to random minors in 1941+, they might have all building slots filled, level 5 AA in every state, but will be sitting there with level 4-6 infra with idle civ factories.

So i wouldnt expect the AI to be smart enough to build supply hubs properly when they currently cant even build infra properly, let alone naval bases to bring in more supplies for amphibious landings....

 

[ruzen]

Short answer
No. Never

Long answer
HOI4 designed as Multiplayer but the majority of people play ho4 for single player. The conflict of this design isn't a concern; people are buying. So...SP people actually don't care about AI, they just want to figure out game systems and optimize it.

 

[zjmna]

People vote with their wallets, and the votes pretty clearly show that players value novel systems and new focus trees.

Keep in mind that the goal when designing AI in games isn't actually to have it play the game well, the goal is to have it give the player an enjoyable experience. I imagine most people out there don't mind when their unoptimized navy handily defeats Japan, or when they forget about Burma for six months and their front hasn't totally collapsed. People on the forum tend to be way more intense about the game than most players and probably care more about the AI, and want more of a challenge.

 

[Meglok]

Umm, you do understand they are reworking the entire supply system to solve a lot of issues reported over the years?

Since we haven't seen how the new system will work I don't see much use in just complaining about an old supply system that is being revamped or assuming the new system won't work.

 

[pro.gamer.69]

we'll see in the update. on the one hand i share your concerns, but on the other i don't think that any of the current shortcomings of the AI are easily fixable, and if any are supply is the one i'd place my bets on.

i also agree with ruzen. many of this game's AI problems are borderline unfixable without simply hardcoding the AI to play "meta," which would increase player dropoff massively. you see it somewhat on this forum but it's even more obvious on the subreddit that this is already an issue - each day there are complaints of "i watched 100 hours of tutorials, why is the game so hard" or "i think i'm good, i've put in 200 hours, now how do you edit templates" and the like. as an MP player i'm obviously biased, but I feel like if you want to LARP the AI is plenty good enough and if you want to have difficult combat MP is plenty good enough.

 

[ThePHD]

we'll see in the update. on the one hand i share your concerns, but on the other i don't think that any of the current shortcomings of the AI are easily fixable, and if any are supply is the one i'd place my bets on.

i also agree with ruzen. many of this game's AI problems are borderline unfixable without simply hardcoding the AI to play "meta," which would increase player dropoff massively. you see it somewhat on this forum but it's even more obvious on the subreddit that this is already an issue - each day there are complaints of "i watched 100 hours of tutorials, why is the game so hard" or "i think i'm good, i've put in 200 hours, now how do you edit templates" and the like. as an MP player i'm obviously biased, but I feel like if you want to LARP the AI is plenty good enough and if you want to have difficult combat MP is plenty good enough.

A large number of people play the game on easy according to the data the devs occasionally release. It's easy for people who have 2k hours played to say the game is too easy but they never account for ordinary players who are sufficiently challenged by the AI. Too many people don't seem to realize or care that they're designing a game for everyone and not just hardcore players. There is a strong ME ME ME mentality on this forum in general. There are way too many people who want the devs to cater to their specific tastes at the expense of everyone else. A great example of this are the historical purists who want the game to be a 100% on the rails WW2 simulator without any sort of alt history. With no consideration given to the fact that a lot of people enjoy that style of play.

 

[pro.gamer.69]

A large number of people play the game on easy according to the data the devs occasionally release. It's easy for people who have 2k hours played to say the game is too easy but they never account for ordinary players who are sufficiently challenged by the AI. Too many people don't seem to realize or care that they're designing a game for everyone and not just hardcore players. There is a strong ME ME ME mentality on this forum in general. There are way too many people who want the devs to cater to their specific tastes at the expense of everyone else. A great example of this are the historical purists who want the game to be a 100% on the rails WW2 simulator without any sort of alt history. With no consideration given to the fact that a lot of people enjoy that style of play.

sounds like you agree with me, then. my point is that the only meaningful way to improve the AI in the places it struggles most, with the exception of supply and airforce management, is to hardcode it more - otherwise it's just too hard to make one that can really "think ahead." and by doing so far fewer new players would continue playing as it would be too difficult.

also while alt history is fun i would like it if the "historical" paths were somewhat more historically accurate. that's a fair criticism, i think.

 

[ArzhurG]

Ideally for me, the AI should be improved at higher difficulty levels. I personally hate the idea of giving the AI artificial buffs. I would much prefer it if they would be replaced with an improved AI. That way those that like to play on normal can just continue, while those that want a challenge gets it with a better AI in the higher difficulties, which would feel more natural. However, I recognize that this won't happen, as it requires a lot more Dev time to make varying levels of AI, than simple static buffs.

 

[Deliberus]

If the developers know that they're going to be changing core systems of the game then why would they update the ones that are going to be replaced?

Also I doubt it stems from ignorance, but rather a priority list...

 

[.Raptor.]

I too think a lot of introduced stuff looks cool but lets all de-hype ourselves (if that's word) and see what happens in action.

 

[Deliberus]

I too think a lot of introduced stuff looks cool but lets all de-hype ourselves (if that's word) and see what happens in action.

Let's bring it all back down to reality in other words and see what happens...

 

[.Raptor.]

Let's bring it all back down to reality in other words and see what happens...

Exactly.

 

[Emren]

Ideally for me, the AI should be improved at higher difficulty levels. I personally hate the idea of giving the AI artificial buffs. I would much prefer it if they would be replaced with an improved AI. That way those that like to play on normal can just continue, while those that want a challenge gets it with a better AI in the higher difficulties, which would feel more natural. However, I recognize that this won't happen, as it requires a lot more Dev time to make varying levels of AI, than simple static buffs.

Especially as you are unlikely to notice the difference between the AI buffs and making actually play smarter in the end result. You'll still win, of course. So is it worth it to spend thousand of Dev hours so that your sense of accomplishment feels better?

Would sales increase if Paradox were able to say: "We made an AI that can beat you without buffs?". Think about it from a business perspective for a minute.

 

[Deliberus]

Especially as you are unlikely to notice the difference between the AI buffs and making actually play smarter in the end result. You'll still win, of course. So is it worth it to spend thousand of Dev hours so that your sense of accomplishment feels better?

Would sales increase if Paradox were able to say: "We made an AI that can beat you without buffs?". Think about it from a business perspective for a minute.

when modders can do better than Paradox in terms of AI then what is Paradox doing with their dev time exactly?

 

[George Parr]

A large number of people play the game on easy according to the data the devs occasionally release. It's easy for people who have 2k hours played to say the game is too easy but they never account for ordinary players who are sufficiently challenged by the AI. Too many people don't seem to realize or care that they're designing a game for everyone and not just hardcore players. There is a strong ME ME ME mentality on this forum in general. There are way too many people who want the devs to cater to their specific tastes at the expense of everyone else. A great example of this are the historical purists who want the game to be a 100% on the rails WW2 simulator without any sort of alt history. With no consideration given to the fact that a lot of people enjoy that style of play.

What a weird post. You start of stating many correct things, before suddenly moving on the one of the worst strawman you can find. There is absolutely no one who demanded a "100% on ther rails WW2 simulator without any sort of alt history". It would be blatantly false to make such a claim. In fact, virtually everyone of who you call "historical purists" has argued for historical plausibility, with alternative routes being closely connected to the time-period and actual possibilities, instead of completely fictional ideas of resurrection empires that vanished centuries ago.

 

[ETAIPOS]

It is not super powerful AI that people in this forum want, I believe. Rather, they want AI that:

1. Can operate on limited supply. Currently AI allies will flood the frontline with troops until no attacks are possible as units are constantly in red supply. Without fixing this issue, all the bells and whistles they are planning to add, would only make things worse.

2. Can launch naval invasion that will not consistently collapse due to above issue. I'm pretty sure that with the new railroads and invasion massively increasing need for the trucks, AI invasions will fail even worse (always?)

3. Knows the difference between infantry division, tank division, marines division and mountain division. Currently it sends tanks into mountains, mountaineers and marines into plains and ordinary inf into contested landings. If it can't even take into account static terrain and infrastructure how on earth is it supposed to handle dynamic system like weather with changing infra, railroads, need to build rr depots and constantly (and widely!) fluctuating number of trucks needed to supply army in the field?

Plus as long as you can't affect how AI allies operate their infrastructure, this all would make playing as Allies impossible - how am I (as UK or US) keep my troops supplied in liberated France if I can't build infra there?
And now I will not be able to build railroads and would need to send them trains and truck hoping they would not waste them somehow?

 

[DocDesastro]

I would give the new system a try before singing swan songs of the AI not being able to use it. Let them roll out the stuff, give the players time for testing and feedback (maybe the spiffing brit will find nice exploits) and then patch, polish, rinse and repeat. Game is growing and with trains and supply hubs it is a good way it is moving towards.

 

[ep1c_fail]

It is not super powerful AI that people in this forum want, I believe. Rather, they want AI that:

1. Can operate on limited supply. Currently AI allies will flood the frontline with troops until no attacks are possible as units are constantly in red supply. Without fixing this issue, all the bells and whistles they are planning to add, would only make things worse.

2. Can launch naval invasion that will not consistently collapse due to above issue. I'm pretty sure that with the new railroads and invasion massively increasing need for the trucks, AI invasions will fail even worse (always?)

3. Knows the difference between infantry division, tank division, marines division and mountain division. Currently it sends tanks into mountains, mountaineers and marines into plains and ordinary inf into contested landings. If it can't do even take into account static terrain and infrastructure how on earth is it supposed to handle dynamic system like weather with changing infra, railroads, need to build rr depots and constantly (and widely!) fluctuating number of trucks needed to supply army in the field?

Plus as long as you can't affect how AI allies operate their infrastructure, this all would make playing as Allies impossible - how am I (as UK or US) keep my troops supplied in liberated France if I can't build infra there?
And now I will not be able to build railroads and would need to send them trains and truck hoping they would not waste them somehow?

I completely agree that the player needs to be exert more control over its AI partners, especially regarding when and where it launches invasions and construction projects in its territory.

 

[Question]

Umm, you do understand they are reworking the entire supply system to solve a lot of issues reported over the years?

Since we haven't seen how the new system will work I don't see much use in just complaining about an old supply system that is being revamped or assuming the new system won't work.

People said the exact same thing with the EU4 Emperor DLC and oh...every paradox DLC ever...how many of those actually fixed issues instead of making things worse? Those who fail to learn from history are condemned to repeat it.

EU4's "colonial nation rework" DLC also broke the colonial nation AI which still remains unfixed more than 5 years later. Paradox's "reworks" are well known to make things worse.

Right here, in this very forum, you can find countless posts about how MTG has made the AI worse in handling the naval aspect of the game. There is plenty of precedent for this.

 

[Question]

If the developers know that they're going to be changing core systems of the game then why would they update the ones that are going to be replaced?

Also I doubt it stems from ignorance, but rather a priority list...

The devs have never bothered to make the AI build infra properly since the game was released.

Now you expect them to make the AI build supply hubs, trucks AND trains properly?

That is just wishful thinking. If your employee cant even show up on time for work a single day in the last 5 years, do you trust him to manage a large corporate event? You sure as heck would not.

Mark my words, when the DLC is released, the forum and subreddit will be flooded by people going "the AI is killing itself with the new supply system" posts followed by people going "lol what did you expect, hoi4 AI has always done that" and "were you seriously expecting good AI from paradox?".