Does the DLC Model Mean HOI4 Will Get Worse Over Time?

[safe-keeper]

Yeah. New players losing to the AI doesn't really mean that the AI is super good. It means that the players are bad. The question shifts into why new players are bad, and that is mostly because the learning 'curve' is practically a line going straight up, because there are so many complicated systems, and the tutorial is... lacking.

Also for all the innovations in UI design (just look at what a convoluted mess HoI3 was), the game is still pretty opaque. EU4 has some nice quality of life features that could be borrowed, such as warning you about the combat penalty when you are about to attack across a river. The division template designer could have some ways to gently nudge new players into creating good, balanced divisions, too, such as warning them if the division template they've designed has very low organization, for example.

Heck, you could even, and I don't know how much processing power this will take, see a "ghost" combat arrow when you select one of your divisions and hover your mouse button over an adjacent enemy division, with an estimate of how well the battle is going to go. Like, if you consider attacking a city on a hill and you see a red arrow and the number "14", you'll probably reconsider that attack. If the game could then even subtly tell you exactly why the attack is a good or bad idea (such as with small icons for things like River Crossing, Forts, Terrain, etc., or a tooltip with a bullet list of pros and cons, appearing next to the ghost combat arrow if you hover over the target for more than a second), that could be incredibly helpful to the newbies.

 

[qer]

@podcat How hardcoded are Ai features currently? Have we reach the limit onto how much is avatible to be modded? Honestly I don't mind if the base game doesn't have a super competitive Ai, and as you stated it would be bad for the general public. However, having the ability to modify it yourself would probably allow a much bigger difficulty range.

Also when will the Ai be able to use paradrops???

 

[Had a dad]

I have had such plans since 2015 but its proved pretty difficult and I wouldnt dare say anythign about it until I had it working honestly.

maybe if you ported the AI to R scripts so the statistical analyses could be done before a reply was sent back....

 

[toegut]

Yeah. New players losing to the AI doesn't really mean that the AI is super good. It means that the players are bad. The question shifts into why new players are bad, and that is mostly because the learning 'curve' is practically a line going straight up, because there are so many complicated systems, and the tutorial is... lacking.

the first time I played the tutorial (Italy) I restored the Roman Empire and conquered the USSR. I don't think the game is particularly hard (especially by the standards of gsg) and is considerably simpler than EU4 in terms of the number of systems. And again you don't actually need to master the systems to beat the braindead AI, you just wait while it suicides on your infantry entrenched behind a river and then walk through it once it's out of equipment.

 

[Kriegsspieler]

@podcat How hardcoded are Ai features currently? Have we reach the limit onto how much is avatible to be modded? Honestly I don't mind if the base game doesn't have a super competitive Ai, and as you stated it would be bad for the general public. However, having the ability to modify it yourself would probably allow a much bigger difficulty range.

Also when will the Ai be able to use paradrops???

The answer to this question is complicated. If you look at the various designated AI folders in the \common folder, you can see that a lot of the AI behavior (diplomatic actions, divisional templates, NF and research priorities, etc.) has been scripted in a way that makes it available for modding. But that's certainly not all that's in the "common" subfolder. Many game files in other folders there also pertain to AI behavior, dictating whether the AI will pursue a given diplomatic strategy, whether it chooses this espionage approach or that one, whether it seeks peace or war. THEN on top of all of that, there is the very large and very essential defines.lua file, located in the \defines subfolder, It specifies a great many in-game variables, including things the AI does, like maintain fuel stockpiles (how large should they be?), and they interact with other specified AI behaviors in complicated ways. All of this -- and probably other things I have overlooked -- is available for modding within certain parameters which are pretty broad, but also finite. And then everything else the AI does is "hard-coded." So if you want to understand what the AI does and doesn't do, and can and cannot be modded, you should devote some time to studying the installed game files. All of them can be opened simply with notepad, except the all-important defines.lua, which requires something like Notepad++.

 

[Zauberelefant]

I find that incredibly hard to imagine someone saying. That said making the AI more difficult to play against would definitely be an issue for some players. Remember that 40% of players are on easy or very easy difficulty settings. I dont remember if it was Tiger or MTG but we had quite a reaction from players who got rolled by the ai so its something I take seriously.

WtT was quite a steep increase in AI effect, I think I had to actually throw 1 or 2 Games and it took me weeks to figure out how to do an ardennes offensive.

 

[Voigt]

AI modding is mostly giving the AI goals, but it is hardcoded how they reach it. But sometimes you can make the goals specific enough so there is only one way to reach it.
For example with ai_strategy_plans you can set a list of foci the AI should go down (so what the AI does in historical mode for example). You can also give a huge reason to befriend/ally or antagonize/conquer a nation. You can also set the weight for picking certain "ideas" which includes laws and advisors to 0 and other ten time the weight than other ideas, so you can create a sequence of ideas the ai should pick. You can force the AI to research certain techs ahead of time, all those things behind condtions you can set, so for example research fighters 3 years ahead of time, if the AI has done the focus that gives it a research bonus for fighters.

But there are things where the AI is to flexible overall, like the division templates (especially with a restricted template field in mods, so instead of 5x5 beeing 4x3) and mil assignment for production. (The AI only produces equipment it has a template which uses that equipment, the AI isn't that smart gaining ArmyXP to get all the templates it needs, and early enough so it can start producing necessary equipment). Also Espionage system, naval system, naval production is overall to flexible but also to longterm for the AI to really handle well.

For those restrictions there are ways around it, as far as I know the mod Total War uses a completely separate scripted division template system, where they wrote their own scripted AI for.

But overall, don't forget Paradox can modify "hardcoded" stuff, since they can change the code. We modders are a bit more restricted, but with complicated scripts using lots of fail-safe conditions for our scripts we can do quite alot.

 

[FlinzP]

Ai questions aren't easiest in games. How I see hoi4 ai currently is that it currently needs some improved guidelines rather than being on human levels with tactics etc. It doesn't need to be smarter but it needs to do things right to feel fair deal. For example some more or less simple things would make huge difference:

1. Option to set specific custom rules for production/tech roles for ai nations. For example to set German ai focus more on tank research and production resulting more up to date and numerous tank divisions. Otherwise it would keep doing things normally. Also you could set Italy on navy and keep Japan on balanced current if you wished so.

2. AI trade priority system changes as it seems rather weird. Often ai do not buy resources from their allies in faction but keep trading from same long distanced nation. It's also weird that puppets sell their resources to neutral countries and their masters are left with nothing.

3. Give ai some basic understanding with navy (and possibly air force). What kind of fleets and how much ships needed. 1. Would also help with research priorities. This alone would not make ai too good with ships but atleast make them look like they are doing something useful with their ships.

4. Having continental priority and faction wide sense. Okey I admit this goes slightly over basic guideline level but anyway. Now ai seem to be slightly lost with where to do battles or idle troops. Hungary can sit it's troops in cores while war with USSR is turning on it's allies. US rather battle over africa than Asia or Europe. Vichy (just example nation) can sen their troops all they way to eastern front just to pull them back immediately. This would help in situations when for example D-day happens so like Spain in axis would react to it sending forces instead shipping them to the Africa or just keep doing nothing.

These alone would make game much more enjoyable to me without resulting stomper ai. Yet allies would feel useful and enemies require some work from player. It would allow playing with few friends funny multiplayer like games without massive multiplayer setup.

 

[L'Iroquois]

Sometimes the AI is indeed smart, but other times it gets 15 divisions encircled in a single tile and does nothing to stop it. Don't even get me started how many times it repeatedly sent unescorted troop convoys straight into submarine infested waters. Th AI is better than it was 5 years ago, but that's not much of an achievement.
I even managed to encircle AI troops that I'm not even at war withView attachment 715437

Ah, yes. The famous member of the Greater East-Asian Co-Prosperity Sphere, Nationalist Spain.

 

[Alpha2518]

I think reducing stupidity is the main thing needed for the AI, there is no need to make it brilliant. Just make it understand the things that are already in the game, this is a larger issue with DLC's as it tends to not understand the extra features like intrigue, ship editor, new naval combat etc. The main problem I can see is that it would perhaps anoy some if such fixes are part of a DLC, but I can say directly that I am fine with it as it's far more important that it's done then that it's passed over due to not generating any money. I guess that also solves the issue for those that don't want the AI to get stronger as they can pass on the DLC.

It would be helpful if there was a code that fed the AI what the players are doing thats effective in games where players are winning from Division builds that are fighting in certain terrain types, to even economy management.

But as for laws, just give the AI set conditions to where it MUST change it's laws. Manpower below a certain threshold? Well time to up the manpower laws. Don't got enough factories for your war economy based on he size of opposing economies being faced? Time to change the economic laws. And these should take first priority over any PP purchasing the AI does to include designers, high command, and ministers with exception to focus tree requirements.