HoI4 Dev Diary - LaR UK AAR

[Bandua_of_Gallaecia]

What Dan is alluding to is the fact that, actually, Germany isn't overpowered at all.

I'll simply have to disagree.
They are far stronger than France (although, France is indeed much weaker than irl, and they can rival the U.S.A and the U.S.S.R in production, which is outrageously far fetched from reality.

With the Expert AI mod which really only somewhat blunts the stupidity of the AI,

Does it really? I doubt any mod can actually change the fundamental behaviours and tactical decision making of the AI. I believe "A.I mods" are limited to changing the way the A.I picks their techs, builds their Templates, manages their production and picks their laws in order to make them follow a more optimal meta.

Barbarossa always gets stopped cold.

As it did, in fewer than 2 years mind you.

This is because Germany's production is easily matched by the Soviet Union on its own,

As it was, significantly so.

and this issue isn't so apparent right now only because the German AI is better than the Soviet AI

Is it thought? Why would that be the case?I'm not an A.I analyst expert but i i assume the A.I uses the same thought process for all nations. Sure it might suck more in one department than another, such as the navy being particularly dumb, but in the Eastern front its a fairly similar strategy for both sides. You say the Soviets are more willing to charge men into the meat grinder (which is relatively historical in the first year of Barbarossa), but as the agressor, Germany not only also does that, but they HAVE to do it, to be able to push.

If the balance of IC/production is such that the Soviets can already stop the Germans cold with ease, then how much easier will it be in 1.9 when the Germans receive a big nerf?

Can they? I am yet to see this happen. In my games the Eastern front comes to a stalemate untill i interfere on either side.

Doing balance based on AI vs AI outcomes isn't necessarily ideal because it assumes that the stupidity of each AI will cancel the stupidity of the other out

Doesn't it?

This isn't necessarily true. We know that the AI fails in different ways on different issues. For example, the AI's failure to use surface fleets is greater than its failure to use submarines well. As such, it would be a mistake to assume that submarines are necessarily overpowered when the AI can't even actually use its surface ships much at all in the first place.

Alright. Makes sense i guess. But how is the A.I different in this example in particular?

The long short of it is that the Soviet Union already in 1.8 may be overpowered based on IC relative to Germany,

Again, this is neither true, nor a problem.
In my games, Germany always outproduces the U.S.S.R. Even i as the USSR can't compete with German industry (although, granted i always play on Elite, so they do get a lot of artificial bonuses)

Even in the MP group I play with, our house mod adds some additional challenges for the Soviet Union including a very severe combat debuff for several weeks after Barbarossa begins.

Thing is, in my MP group, the German player almost always wins in a 1v1 against the Soviet player. I'm not saying my anecdotal is better than yours, all i'm saying is that it's really hard for me to understand where you are coming from.

Unmodded the scales of balance are tipped against Germany, which is perhaps only compensated for by the still iffy balance around submarines.

Wait... i thought you just previously said unmodded Germany won because it had the advantage of having a less stupid A.I and if you modded the game with expert A.I , the soviets would stop them dead cold? Or are you talking strictly about multiplayer now?

 

[imcool]

It seems like a minor changing DLC to you?!?

It's being developed for just under a year and it costs 20 bucks, which is quite a lot for a PDX DLC...

I had stayed clear of anything and all i knew was a Portugal focus tree and spying which didn't seem Like major features of course know i know about how the spying will change the game for a new flavor this was most of this was mentioned in my original post. Paradox habits are usually making a dlc in a year?. While 20 pounds is the regular for a major dlc However is it not true that eu4s cossacks was 20 pound and it brought only a single major feature?

 

[Dalym]

Nice work guys

 

[kettyo]

I refuse to believe that getting a competent AI is impossible. It's just that few people are willing to put in the time, effort, and money. If not from video game companies, then I don't know who will do it one day, but I am sure it will happen.

Theoretically a subscription based game would push developers more towards a good AI. DLC based game pushes them towards more content, no better AI and more polished game.

 

[safe-keeper]

I just realised each dev diary should have been a single-play snap. Then he could have started each with "listen carefully, I shall say this only once".

I'll get my coat.

 

[safe-keeper]

I couldn't agree more. If they pull off another MTG where the AI doesn't even put together any ships beyond 1936, then WTF should we continue to buy their DLC? At the least, hire another AI programmer and get them focused on HOI4 for the next 6 months until it is working better.

We know getting an AI right is difficult, and never really that satisfying in strategy games, and much of the challenge comes from allowing the AI to cheat to get there, but why keep on adding more and more features when the AI cannot even work with existing features?

I refuse to believe that getting a competent AI is impossible. It's just that few people are willing to put in the time, effort, and money. If not from video game companies, then I don't know who will do it one day, but I am sure it will happen.

At least there doesn't seem to be any must-have features in La Resistance hidden behind a paywall, like convoy routing in MtG. That's uplifting.

But yes, the whole point of the DLC model, in my eyes, is that it allows the devs to support their games for years and years when they would otherwise have abandoned them and moved on to sequels. But then issues actually need to be addressed when they pop up.

 

[Cybvep]

I follow the news about HOI4 semi-regularly in order to get an idea about the current state of the game. I've not played it yet and I know that HOI games require a large time investment, so I want to make sure that this game is in a really good shape before I pick it up. While the features in this expansion look quite cool, the AAR paints a rather sad picture of the overall game balance and the quality of the game's AI. In a British AAR focused around showcasing various new features, with several decisions that might seem rather questionable (like not defending France), ending the war with a decisive victory in 1942 is really not something that I would expect. The comments here are not encouraging, either.

So yeah, back to the wishlist with this one. Maybe one day...

 

[Leinad965]

I follow the news about HOI4 semi-regularly in order to get an idea about the current state of the game. I've not played it yet and I know that HOI games require a large time investment, so I want to make sure that this game is in a really good shape before I pick it up. While the features in this expansion look quite cool, the AAR paints a rather sad picture of the overall game balance and the quality of the game's AI. In a British AAR focused around showcasing various new features, with several decisions that might seem rather questionable (like not defending France), ending the war with a decisive victory in 1942 is really not something that I would expect. The comments here are not encouraging, either.

So yeah, back to the wishlist with this one. Maybe one day...

Does it mean, that AI is so bad, or that playing as Germany is hard as should be?

 

[General Von Trapp]

Theoretically a subscription based game would push developers more towards a good AI. DLC based game pushes them towards more content, no better AI and more polished game.

It COULD I suppose but I dunno about the subscription model. I would rather pay for my game straight up. I might not play the game for two months but I'm still paying for it which bothers me.

It's getting to a point where people could be paying hundreds of pounds each month through just through subscriptions. You get ones where you sign up for a year for a cheaper price BUT you cannot cancel it or you pay higher for the ability to actually cancel it.

Having said that if it did make Paradox focus on AI more I would pay subscription but there should be cancel anytime option.

 

[billcorr]

Nice work guys

Congrats on your first post!

An uplifting one, at that. You'll do well on these forums.

Yes, well done, team. Well done.

 

[Mister Analyst]

Nice work guys

Welcome to the HOI4 forum.

Also take a peek at the HOI4 wiki https://hoi4.paradoxwikis.com/Hearts_of_Iron_4_Wiki.

 

[Axe99]

Does it really? I doubt any mod can actually change the fundamental behaviours and tactical decision making of the AI. I believe "A.I mods" are limited to changing the way the A.I picks their techs, builds their Templates, manages their production and picks their laws in order to make them follow a more optimal meta.

For info, there's an awful lot of things that can be tweaked in a mod to affect the AI, from defines that impact on operational combat behaviour (when to attack, when not to, in terms of org and strength, for instance - there's more, but it's been a year or two since I looked at them closely), to AI strategies that can be set with any of the triggers available to events or decisions, and can impact on a whole bunch of things. There's a lot that can't be done in mods, of course, but there's a lot that can be done as well, beyond techs, templates, production management and laws.

That said, on the broader discussion of HoI4's AI, I personally think it's very good for a video game. It's important to keep in mind that no-one, anywhere, has produced an AI that can match human thinking, and in the 1936-1948 period at the HoI4 level there were literally thousands of individual decision-makers that contributed to the actions that were taken. It's not technically possible to model this at the moment with any computing power, let alone with that of a desktop/laptop, and the amount of work required to model it, even if it was possible, would result in a game of ridiculous cost.

In terms of a broader, abstract modelling of an operational-level WW2 game, I think HoI4 does a great job, and continues to get better. It's had its wrinkles, but many of these have been ironed out, and I'd be very surprised if more work hadn't been done for the patch that comes with La Resistance (we get an AI diary pretty soon, iirc). The land unit management, for me, is comparable in quality to Gary Grigsby's War in the East, and I honestly can't think of a WW2 grand-strategic or operational level game at the same scale (divisions) that does better, and can think of more than a few that have done worse. When playing computer games, there's likely to always be a degree of 'suspension of disbelief' required, and there's still a bit in HoI4, but in context (ie, a strategy video game), HoI4 compares very favourably with its peers, often while doing a good deal more than they are.

 

[Iskulya]

I'll simply have to disagree.
They are far stronger than France (although, France is indeed much weaker than irl, and they can rival the U.S.A and the U.S.S.R in production, which is outrageously far fetched from reality.

Bruh. Really?

You're going to do a line by line quotation?

That's lazy and poor writing. The point of writing is to make a coherent body of thoughts, communicated simply and efficiently. It's tedious, boring, and tiresome to respond in such a way, but I'll play. Once.

So? What's your point?

The above being purely rhetorical. It's just that your argument is so flawed, so self-evidently and obviously wrong that you can't actually state it outright. So I'll spell it out for the benefit of others here. It does not matter how production figures vary from historical reality in the game.

Hearts of Iron 4 is, above all, a game. The point of a video game is to entertain and provide fun. Hearts of Iron is not an educational guide. It is not a scholarly dissertation. It is not a renactment. It is not a documentary. It is a game.

The United States produced over half the world's industrial output by the end of 1945. This isn't, and shouldn't be represented in the game because it's not fun. Moreover, a videgame cannot reproduce the byzantine chain of causality and contingency that led to history unfolding the way it did, nor should it try.

If one were to take your implicit point that everything should be molded strictly around historical inputs, then the most common result is that Germany would be stopped at the French border and capitulate in 1940. If you think that sounds like it makes for a fun game, then you'd be alone in that. Fun and game balance is more important than reproducing historical inputs.

Does it really? I doubt any mod can actually change the fundamental behaviours and tactical decision making of the AI. I believe "A.I mods" are limited to changing the way the A.I picks their techs, builds their Templates, manages their production and picks their laws in order to make them follow a more optimal meta.

It's weird that you'd even try to deny something that is so easily investigated and disproven with just a few minutes of effort. There's nothing stopping you from just downloading EAI from the workshop and trying it out. Yes, it is indeed possible to make the AI less inclined to blatant suicidal attacks. There is a really strong limit to what modders can do, but it's a proven fact that you can indeed influence the AI's decisions in terms of tactics(which again is always going to be within very narrow confines, but enough to make a difference).

As it did, in fewer than 2 years mind you.

Disingenuous, and within the context, blatantly and undeniably false.

You are equivocating the Soviets losing millions of troops, losing Ukraine and getting pushed back to the gates of Moscow, and requiring years of day and night combat to turn things around to the Soviets stopping the Germans at the border.

Not only is the former not "stopping cold" the Germans. It is quite literally the exact opposite of it, in fact. There is no real equivocating between the two and it's more or less impossible to take you seriously when you say that they are the same thing.

As it was, significantly so.

Irrelevant. Game balance =/= reproducing historical production.

Is it thought? Why would that be the case?I'm not an A.I analyst expert but i i assume the A.I uses the same thought process for all nations. Sure it might suck more in one department than another, such as the navy being particularly dumb, but in the Eastern front its a fairly similar strategy for both sides. You say the Soviets are more willing to charge men into the meat grinder (which is relatively historical in the first year of Barbarossa), but as the agressor, Germany not only also does that, but they HAVE to do it, to be able to push.

No they don't. You're just twisting the truth outright here. The German AI is markedly less prone to committing troops to an attack that is a pre-ordained failure. The Soviet AI will not only do so, but do so when division strength is even less than 50%. You only have to play one game to see this.

Also what you say is a contradiction in terms. If the German AI is making fruitless attacks, then it is doing the opposite of making a push. It is buying for time to delay an inevitable(and successful) counter-attack. But if you were honest, you'd admit that this actually isn't something that typically happens relative to the Soviets. You're just being a contrarian here, tbh.

Can they? I am yet to see this happen. In my games the Eastern front comes to a stalemate untill i interfere on either side.

You're being a contrarian once again. You're actually agreeing with me here. I said that the Soviet AI consistently stops the Germans, and by saying that it comes to a stalemate you're actually just agreeing with me. Try a little harder if you're just trying to be disagreeable for the sake of it.

Doesn't it?

You tell me. Considering you haven't made any argument here and are just being a contrarian.

Alright. Makes sense i guess. But how is the A.I different in this example in particular?

Really? Are you actually so lacking in knowledge of the game as to be seriously asking this? I think you are just being a contrarian again and stalling, but ok, I'll bite.

Using submarines successfully doesn't really require much. Just throw on the latest snorkel, put as many torpedoes as you can, put on the latest engine, and fill out the right branch of the trade interdiction doctrine and throw them on convoy raiding missions.

In comparison, stopping submarines requires much more work. To really do it most effectively, you must coordinate your airforce and your navy. Also designing and producing ASW vessels comes at the expense of vessels that are more effective in surface combat. Additionally, radar much be researched in addition to depth charges. It's a very expensive research investment compared to researching fully modern subs.

It takes a much greater investment in research, production, and effort to stop submarines than it does to use them effectively. Correspondingly, the flaws of the AI will compound themselves the more steps you add in a process. Very simple to grasp, really.

Again, this is neither true, nor a problem.
In my games, Germany always outproduces the U.S.S.R. Even i as the USSR can't compete with German industry (although, granted i always play on Elite, so they do get a lot of artificial bonuses)

Wrong on both counts. It is not hard for the Soviet Union's IC to match or even exceed Germany's by the end of 1942, even throwing into account Germany's conquest gains. It is not really helpful or accurate to speak of experiences on elite as being reflective of any kind of reality when that level of difficulty grants the AI actual cheats.

In fact, the Soviet Union gets so many factory slots it can double America's IC. Strange, then, that supposedly for being such a stickler for historical numbers that you have taken no issue with this massive discrepancy. This is because you twist the facts and outright omit realities that reveal the inconsistencies of your position.

Thing is, in my MP group, the German player almost always wins in a 1v1 against the Soviet player. I'm not saying my anecdotal is better than yours, all i'm saying is that it's really hard for me to understand where you are coming from.

I can only say the same thing to you and marvel at how bad your Soviet player is.

Wait... i thought you just previously said unmodded Germany won because it had the advantage of having a less stupid A.I and if you modded the game with expert A.I , the soviets would stop them dead cold? Or are you talking strictly about multiplayer now?

Don't fault me for your lack of reading comprehension here.

Balance is a separate issue from the inadequacies and incongruencies of the AI. In fact that's what this entire issue is about. AI vs AI performance is not a metric for balance, since balance refers to this factors that remain constant even when every country is played by a human. The entire basis for people's disagreements is that people have mistakenly gotten the idea that Germany is overpowered purely on the performance of the AI. The example of EAI was just one piece of evidence that this is untrue, but it is hardly even the most significant one by far. Multiplayer is the better metric here.

The balance situation is that the odds are already tilted against Germany in 1.8. 1.9 will nerf their factories from occupation and further tilt the balance against them. You have tried to mystify the entire situation and conflate it all by invoking historical production numbers when that really has nothing to do with the issues at stake in the game at all. Game balance has nothing to do with history. Period. If it did the game would have been designed so that it is impossible for the Axis to win. You are the one bringing in completely irrelevant factors and conflating them with the real issues at stake here, not me.

Not a very good try, though. I rate it 3 out of 10.

 

[kettyo]

There's nothing stopping you from just downloading EAI from the workshop and trying it out

That's a very good point.

In my experiences in vanilla current version Germany defeats everyone, in EAI normal mode Germany fails quite hard, and in EAI challenging mode it is quite balanced.

So AI scripts have a very big effect on AI performance obviously.

 

[Kadanz]

Bruh. Really?

You're going to do a line by line quotation?

That's lazy and poor writing. The point of writing is to make a coherent body of thoughts, communicated simply and efficiently. It's tedious, boring, and tiresome to respond in such a way, but I'll play. Once.

He adressed your arguments point by point by quoting specific lines. Considering this is a text based discussion, I'd say this is the best possible way of adressing specific without confusion.

Is it always necessary for you to be condecending in your responses to other people or are you just that insecure when you get challenged with other points of view?

 

[Had a dad]

Play nice...

 

[CountCristo]

Am I missing something @YaBoy_Bobby - how did the decryption bonus help you hold a defensive line? +15% defense is meaningless given that defense will always exceed the AIs attack (unless you made them a WHOLE LOT better) and therefore do nothing and all the other bonuses are offensive.

 

[Leinad965]

I think what Dan is talking about is the effects of the new occupation system on Germany's total IC. We already know that this is going to be a strong nerf to Germany.

What Dan is alluding to is the fact that, actually, Germany isn't overpowered at all. With the Expert AI mod which really only somewhat blunts the stupidity of the AI, Barbarossa always gets stopped cold. This is because Germany's production is easily matched by the Soviet Union on its own, and this issue isn't so apparent right now only because the German AI is better than the Soviet AI(which has a tendency to repeatedly suicide troops much more frequently than the Germans do).

If the balance of IC/production is such that the Soviets can already stop the Germans cold with ease, then how much easier will it be in 1.9 when the Germans receive a big nerf? That is the central issue here. Doing balance based on AI vs AI outcomes isn't necessarily ideal because it assumes that the stupidity of each AI will cancel the stupidity of the other out, and that therefore the result is indicative of industrial capacity, national spirits, and other factors of balance.

This isn't necessarily true. We know that the AI fails in different ways on different issues. For example, the AI's failure to use surface fleets is greater than its failure to use submarines well. As such, it would be a mistake to assume that submarines are necessarily overpowered when the AI can't even actually use its surface ships much at all in the first place. Similarly it is a mistake to assume that Germany's IC situation in 1.9 isn't unbalanced just because it performs decently against AI soviets.

The long short of it is that the Soviet Union already in 1.8 may be overpowered based on IC relative to Germany, but this issue will just become greater in 1.9. Even in the MP group I play with, our house mod adds some additional challenges for the Soviet Union including a very severe combat debuff for several weeks after Barbarossa begins. Unmodded the scales of balance are tipped against Germany, which is perhaps only compensated for by the still iffy balance around submarines.



Pretty much, but at least it looks fun.

From everything we've seen there's no reason not to think that the best thing to do as the player is just go all out on using your spies offensively. Although in multiplayer, there's potentially a careful balance because your agents can be captured and leak a lot of intel without the enemy having had to venture for it at all. Against the AI, this doesn't factor in at all because the AI is probably mostly, or entirely, incapable of adjusting its plans and strategies in reaction to the player's production strategies, troop movements, fleet deployment, etc. It won't help the AI much, if at all, for it to have 100% knowledge of everything you do. On the other hand, the player DOES benefit from using spies to discern the AI's production, fleet deployments, and so on.

In Expert AI you get from harsh occupation goverment only 50% of factories, while in vanilla 1.8 90% factories. It looks like, that you will get about 50% in new 1.9 version, so no new nerf from Expert AI point of view, only from current vanilla version perspective.

 

[BlueTico]

This DD tells us zero.

 

[currylambchop]

Reason why Paradox doesn't update AI is that most players who play singleplayer are so bad that the AI is hard for them.

Obviously people on this forum would be good enough to shred the AI as they're interested, and so play more/understand more mechanics. But many people find Germany in vanilla hard lol. Just go on Steam community forums and you'll see.

So Paradox doesn't have a real incentive to update AI as those bad players will keep playing and not even notice the problems.