Observing the AI... am I the only one cringing, laughing, and crying.

[DaleDVM]

As I sit here, playing as the Germans. I can't even get to 1939 and starting the war any more. I just get so depressed watching the AI... and then I ask myself, what am I doing wasting my time in this game? I want to love it... I repeat, I want to love it so, so badly... However, I could take as much pride in winning against this AI, as beating a squirrel in a chess match.

Today's atrocity... The invasion AI and front system. The Japanese AI (as usual) abandons the Manchurian front, to take troops down to their newly invaded beach heads. The Chinese just sit there and don't advance into the abandoned Manchurian territories. (In six months the same thing happens and the Chinese do advance. Fool me once... LOL) In the invasion zone around Shanghai the Japanese completely abandon the ports they have taken and move inland so China can counter and easily put all of their troops out of supply... further south, the Japanese AI insanely abandons an entirely successful invasion around Fuzhou. They had 13 territories and two ports in the mountains only to fully withdraw all of its divisions to reinforce a new invasion going on in the North just east of Qingdao. There they cram 20+divisions in just 3 territories and lose them all. The Japanese AI is constantly embarking and disembarking its divisions which have no organization because they just can't sit still for even a second. Often the same port is embarking divisions at the same time as new ones are landing in it.

I know none of this is new and my experience is not extraordinary. This happens every time I play, I just frequently turn my eyes away from the horror. The front line shuffling of units is absolutely insane in this game. Front lines may have solved some problems in the game. It made others that have not been addressed since they were introduced years ago. It is bad enough when the fronts are solely on land and the shuffling starts. When the front shuffling starts during an invasion it leads to a literal slaughter that the AI usually can't recover from.

Put the line shuffling, which is a big complicated problem, aside. Can't PDX at least make the AI recognize when they are outnumbered and sit defensively instead of 2 divisions trying to dislodge 16 divisions from a territory? Is that too much to ask?

I understand that making an AI perform well in a game this complex is very, very hard. However, it is also obvious that after 4 years, PDX has not been able to make a challenging AI for HOI4. They say they constantly work on the AI and I believe them. I am not certain if it is the game systems, or the complexity, or the demand it would put on our hardware, or some combination of these that is the stumbling block. No matter how much work is being done, and no matter how much better the AI is today versus on the release day 4 years ago, it would take a giant leap of AI performance to fix the crazy things I am watching the AI do now. Japan abandoning an entire front to the enemy? Letting them take the factories and resources in the entire area... How did this pass quality control in the first place?

I will give the programmers this much credit. They have balanced the game in such a way that the AIs make it look like some sort of credible WWII simulation is being run in the background. When you are playing as a country, and you have the blinders on, and really only know who is taking territory it doesn't look too bad. However, if you observe the AI at work it is just nauseating to watch. The only reason it works is because both sides are horrible and they balanced that.

I guess it is good that this game has a really steep learning curve. The complexity of the game and the lack of feedback it gives the player likely makes winning challenging for beginners. Being a veteran of all of the HOI games this version never has challenged me. Even on max settings, even with Expert AI mod, even though I play historical, with no cheesy strategies... Still no challenge. I doubt I am even a very good player. The AI is just that bad and it makes me so very very sad.

I truly want to hope for HOI4 in the future. I guess at least it was a commercial success. Not certain if that is a good thing or not? I suppose it means the game will be supported for a while. But it also means no other grand strategy game like this is going to get made. HOI4 is taking up all of the oxygen in the room.

Just my opinion and venting my frustrations.

 

[safe-keeper]

Observing the AI... am I the only one cringing, laughing, and crying.

Nope.

 

[Mousetick]

You're not the only one. But let me ask you something: why do you want to love this game so badly?

If you feel you're wasting your time and you're frustrated, but you keep going because you want to love the game, that sounds a bit unhealthy. It's pretty clear that you're not having a lot of fun and it's not going to get better any time soon (seeing as how the game has evolved since its release). Why not accept that fact and put the game aside for a while, maybe try some other games or other activities that might provide more enjoyment.

Warning: Excessive verbosity ahead!

The following view is about single-player. I don't have any experience with multi-player.

In another recent thread, bitmode asserted that a good strategy game is "easy to learn, hard to master" (Bushnell's law). I'd argue that HOI 4 is the opposite, it's hard to learn and easy to master. There is a steep learning curve and huge time investment to learn the game, from which I derived most of my enjoyment of it. But once a player has learned all the mechanics, the small details, the bugs, quirks and all - once a player "has cracked the code" so to speak, the game becomes completely transparent: one can see clearly what's going on behind the curtain of pretty graphics, that is a giant spreadsheet. Once a player has reached that "plateau", it becomes apparent that there is not much strategy involved, and no challenge offered, because everything is predictable and the AI cannot play the game properly.

I don't believe the AI will ever become good at playing the game, because it's too damn complicated and there are too many variables/details on too large a scale. The AI is also hampered by conflicting goals that the game is shooting for, which seem to me are quite hard if not impossible to achieve together.

Multiple conflicting goals:

- Strategy vs. Simulation
The strategy aspect favors making interesting decisions, to which the AI can react dynamically. This is easier to achieve if the rules are few, simple and abstracted (e.g. a board game or its computerized equivalent). While the simulation aspect favors turning knobs, flipping switches and observing behaviors as realistic as possible, that the AI can follow using prescribed 'recipes'. HOI 4 is quite good at the latter (giant spreadsheet), but quite lacking in the former.

- Strategic vs. Tactical or Macro vs. Micro levels
The AI cannot keep track of long-term decisions and at the same time react to the situation on the ground in real-time. There is just too much stuff going on, on a worldwide scale. At a high level it simply follows a prescribed path without regards to anything going on around it (whether it is an enemy or an ally of the player), it just follows its path blindly. At a lower level it's basically the same thing, the AI follows battle plans. It would be unreasonable to expect that AI be good at both levels on such scale. I for one wouldn't mind if the game forego the tactical/micro level completely for a better, more aware and reactive AI at the strategic level. There are plenty other WW2 games, some reportedly quite good, that focus on the tactical aspect.

- Historical vs. Sandbox
This ties into the other points. The sandbox is not a true sandbox with unpredictable behaviors and outcomes. The AI's behavior is constrained within a framework of recipes and steps to follow, which are appropriate for a historical simulation, but their predictability by the player once the game is mastered, makes for a poor sandbox experience.

This is a really tough nut to crack. I wish the game would be more focused and less ambitious. Sometimes less is more.

 

[randomguyplays]

Just move people from sales to the development team, and hire a dedicated AI programmer if necessary

 

[El_Ploplo]

Oh boy, don't even try the pacific war if you think the AI is already incompetent on land...
A week ago, I decided to try historical USA for the first time with MTG. I never did it before because I thought the naval warfare was to damn complicated for my taste. With others countries I just spam subs and it was sufficient. This time I decided to try it once for all, with some restriction so the game cannot be to easy (cannot join war before 1941).
So I read a lot of tutorials to understand the exact mechanics of naval warfare, the importance of screen ships, of capital ships, the use of carriers and how they can change the outcome of a battle. I designed a complete fleet, I was very happy with it ! I, for the first time, invest in marines and amphibious tank and produce a ton of them. Then I go to war against Japan...
Well my fleet was completely useless, not that it was bad, no not at all but the AI doesn't even bother to defend its ports and territory, I walk into Iwo Jima then Hiroshima completely undefended, the war was over in less than a year.
I realize the AI didn't even give orders to its fleet, no wonder why it was so easy !

And now I understand why I can Sealion in 1939 as Germany, why subs have always been enough. The AI is completely garbage in sea. I will try again with expert ai, see if it is different but I think the AI is to broken to be interesting even with this mod.

 

[Shetan]

For me it would be perfectly enough if the AI learned how to hold the frontline without gaps and reshuffling. It's too easy to exploit and left me with unpleasant after-taste.

 

[CantGetNoSleep]

The AI seems to get confused and abandon a front easily. Save (if not in iron man) and reload, and it'll "wake up". Sadly, this has to be done quite regularly.

 

[Quivis]

Is AI still sending its troops by sea without any protection and losing that way literally millions of soldiers? I'm simply curious, as I haven’t touched the game for more than a year now (due to AI issues).

 

[GSP Jr]

You would not like the speed of the game, or more likely not have enough horsepower to run it, if it actually had the AI you want.

Play human opponents, and many AI "problems" can be minimized.

 

[DaleDVM]

You're not the only one. But let me ask you something: why do you want to love this game so badly?

- Historical vs. Sandbox
This ties into the other points. The sandbox is not a true sandbox with unpredictable behaviors and outcomes. The AI's behavior is constrained within a framework of recipes and steps to follow, which are appropriate for a historical simulation, but their predictability by the player once the game is mastered, makes for a poor sandbox experience.

This is a really tough nut to crack. I wish the game would be more focused and less ambitious. Sometimes less is more.

I want to love it because I love WWII simulations. The HOI series of games were the best at this over the last 18 or so years. Eventually the success of HOI 2, 3, and 4 has driven other WWII grand strategy games out of the market. I am not into tactical war games as much. Tactical games certainly have better AI. You are right my high hopes for this game are not healthy for me...

I agree 100% on your final point. Too much time has been spent on giving minor countries flavor in this game. All of those a-historical options just make it more confusing for the AI.

 

[DaleDVM]

You would not like the speed of the game, or more likely not have enough horsepower to run it, if it actually had the AI you want.

Play human opponents, and many AI "problems" can be minimized.

I would play humans if I could get on the same schedule, and they wanted to play slowly, and they never quit mid game, so in other words, I won't be playing multiplayer. Except with a couple of friends which i sometimes can pull off... but then we all get frustrated when an important AI ally does something idiotic and ruins the war.

 

[GSP Jr]

I would play humans if I could get on the same schedule, and they wanted to play slowly, and they never quit mid game, so in other words, I won't be playing multiplayer. Except with a couple of friends which i sometimes can pull off... but then we all get frustrated when an important AI ally does something idiotic and ruins the war.

Seriously, I feel your pain, having some of those very issues myself. I have turned to mods and modding to try and improve enjoyment and scratch my dev/coder itch.

 

[grommile]

Just move people from sales to the development team, and hire a dedicated AI programmer if necessary

Ah, the "interchangeable morons" school of human resource management.

In the real world, if you want to achieve that result, you have to lay off sales folks (which, for starters, means making severance payments, since Paradox are in a country that actually believes, however imperfectly, in workers' rights), then hire devs and get them up to speed on how the project works.

Hope you didn't actually need those sales folks.

 

[Zeeth]

Then play multiplayer ffs.
AI is never going to be as intuitive as a human.
The only way to balance an AI is to let it cheat.
This is literally how AI is balanced in EVERY SINGLE GAME.

Snippit from the AI wiki:
Cheating AI

Dani Bunten was once asked how to play-balance a game. Her one word answer was "Cheat." Asked what to do if gamers complained, she said, "Lie!"
— Johnny L. Wilson of Computer Gaming World, 1994[34]

Gamers always ask if the AI cheats (presumably so they can complain if they lose)
— Terry Lee Coleman of Computer Gaming World, 1994[35]

In the context of artificial intelligence in video games, cheating refers to the programmer giving agents actions and access to information that would be unavailable to the player in the same situation.[36] Believing that the Atari 8-bit could not compete against a human player, Chris Crawford did not fix a bug in Eastern Front (1941) that benefited the computer-controlled Russian side.[37] Computer Gaming World in 1994 reported that "It is a well-known fact that many AIs 'cheat' (or, at least, 'fudge') in order to be able to keep up with human players".[34]

For example, if the agents want to know if the player is nearby they can either be given complex, human-like sensors (seeing, hearing, etc.), or they can cheat by simply asking the game engine for the player's position. Common variations include giving AIs higher speeds in racing games to catch up to the player or spawning them in advantageous positions in first-person shooters. The use of cheating in AI shows the limitations of the "intelligence" achievable artificially; generally speaking, in games where strategic creativity is important, humans could easily beat the AI after a minimum of trial and error if it were not for this advantage. Cheating is often implemented for performance reasons where in many cases it may be considered acceptable as long as the effect is not obvious to the player. While cheating refers only to privileges given specifically to the AI—it does not include the inhuman swiftness and precision natural to a computer—a player might call the computer's inherent advantages "cheating" if they result in the agent acting, unlike a human player.[36] Sid Meier stated that he omitted multiplayer alliances in Civilization because he found that the computer was almost as good as humans in using them, which caused players to think that the computer was cheating.[38] Developers say that most are honest but they dislike players erroneously complaining about "cheating" AI. In addition, humans use tactics against computers that they would not against other people.[37]

 

[SchwarzKatze]

No one is asking the AI to be Rommel or Zhukov, just having it stop doing what anyone with half a wit won't do (in this case, leave gaping holes on the main front to launch an invasion, then proceed to not guard the beachhead port of the said invasion)

 

[DaleDVM]

I would argue that the orders system for an army needs to be divided into two parts. Lets say 75% of the divisions are assigned to the front lines. 25% should be put on guard duty at ports and victory points in the area of operations. In this way reserve divisions would prevent paradrops into capital/major cities and would also protect ports after an invasion.

I am not a programmer but this seems like it might be feasible to do. If not, perhaps each large army of 24 divisions could be assigned a smaller 8 division corp/army for guarding key areas in the rear. This also might prevent fronts from collapsing as the dug in divisions in the rear would make for good spots to slow down an advance.

 

[Mousetick]

No one is asking the AI to be Rommel or Zhukov, just having it stop doing what anyone with half a wit won't do (in this case, leave gaping holes on the main front to launch an invasion, then proceed to not guard the beachhead port of the said invasion)

While I agree that such stupid behavior is detrimental to the human player's experience, I think your expectations are unreasonable and won't be fulfilled in the foreseeable future, for 2 reasons:

1. You understimate the effort required from the game developers.
Computers can make basic operations faster and in greater quantity than humans (example of basic operation: arithmetic calculation, or comparison between 2 numbers). The key word is 'basic'. Computers (supercomputers) can beat humans at chess because the rules of this game are simple. Even so, the programs that codify the simple rules of chess into basic operations are extremely complex. Launching and guarding an invasion, or manning a frontline, are far from having simple rules easily codifiable into basic operations. It's not an impossible task, but it certainly is neither easy nor quick, especially when the AI runs in realtime on an average PC.

2. You assume this is important for the developers.
Given that the effort required is quite large relative to the overall scope of the game and the many goals it's trying to achieve, it is possible, even likely, that this not a high priority for the developers. Otherwise, the issue might have been addressed earlier. It's also possible that the developers are putting it off for later as part of an even larger effort, such as an extensive redesign of existing parts of the tactical AI - which may never happen.

Consider there are whole computer wargames solely dedicated to a single theater of operations, or even a single battle. By trying to satisfy too many tastes, and to have too many features, the game and its developers cannot hope to perform well in all of them. At best the game is good in some areas, while it is only average, or downright mediocre, in others. However it is unique in what it does, which makes its experience at least interesting, if not always pleasant or rewarding.

This kind of recurring discussion about the game's shortcomings, specially the AI, and how to solve them, are always fruitless and can only lead to frustration, or even anger for some. It would be more sensible IMHO to simply accept that the game is bad in some aspects, that it will remain so for a long, undefined period. Enjoy the better aspects, or move on.

 

[SchwarzKatze]

1. You understimate the effort required from the game developers.

2. You assume this is important for the developers.

accept that the game is bad in some aspects, that it will remain so for a long, undefined period. Enjoy the better aspects, or move on.

If your whole argument is "this is too hard for the devs" and "it's not important enough for the devs", you should provide a source on the claim.

And "take it or leave it" isn't an argument, more so when you're not even on the development team.

 

[Fulmen]

I can't even get to 1939 and starting the war any more.

I've never been able to enjoy singleplayer in HoI4, and much like you, always find my attempts at playing it ending in quitting before or by early war. The AI and the lack of a "WW2 feel" due to gross lack of realism and historical accuracy is just too damn disappointing.

What's the point of playing the UK for instance, if I can just capitulate Germany in 1939? I want to go through the strategic challenges the Brits faced in 1940-41: I want to fight the Battle of Britain, the years-long naval campaign in the Atlantic against the U-Boat menace, the slogs that were the campaigns in Norway, Greece, Libya and Egypt, and maybe even be defeated here and there. None of this should involve giving the AI magic boosts, and heck, it doesn't: all those boosts do is make the game grindier, not more engaging.

MP is the only HoI4 scene where I've ever been able to find some kind of "WW2 fix", but those days are long gone now, and all that's left of MP are semi-private groups that play mods more concerned with sandbox balance than realism, or tiny private groups that lack the players to even man all the majors, let alone the important minors, and take weeks to finish a game because they play on speed 1-2 with frequent pauses.

 

[Metz]

The last roadmap said that they wanted to work on the AI having guided long term goals and strategies as opposed to just having historical or random paths.

This could work by having "National Strategies"

For Japan, for instance, one would be to secure naval presence near the Mariana Islands. Adding a fleet of X size set to patrol the sea tiles would accomplish national strategy and then the reward would be a small amount of naval experience. For Britain it could be to have an X amount of planes intercepting over South England which would represent the Battle of Britain and maintaining X amount of air supremacy for Y amount of days would give a certain bonus to air experience and political power (to represent victory and safety in the air). For Germany you may have one that deals with submarine warfare in the Atlantic, sink X amount of convoys for a reward of naval experience and as well as a drop in stability for the UK.

The developers also said they wanted to add more detailed effects to weather and a Barbarossa update is coming. This could mean that the German player or AI may have to stop their attack during winter instead of just steamrolling the USSR 24/7. This would make it so that the war in the east lasts longer and it would give the USSR breathing space to reinforce.

Regarding an easy Operation Sealion, perhaps they should separate convoys from transport ships so that Germany actually has to build transport ships with its few shipyards. Transport ships should cost more than convoys since they tend to be rather big and carry thousands of troops. An amphibious invasion of Britain was possible, the issue was keeping up the supplies once the troops land. This is why they should implement supply depots as buildables in provinces. Divisions then tap into these based on which one their army is assigned to. You should then be able to stockpile materiel in supply depots for future use. Having supplies come in from your capital on a daily basis is a bit unrealistic, especially if your divisions are far away. Supplies realistically should come in large batches at a time from nearby places. Trucks should be implemented to show movement of supplies like a land-based version of the convoy.

Regarding movement and infrastructure, there should be railroads that connect the minor cities to each other and to the capital. That way movement can be faster if the railroads are used. Not all would exist so they would be something you build during the game on a need by need basis. If a country uses a different gauge then it needs to be changed so you can use it which was the case during Germany's invasion of the USSR. The rails would be visible on the map and to use them you'd have to click on the railroad option in the army/division interface.