Quill18 vid: "War is looking good"

[Hans_Schnitzel]

Well I disagree I always thought ordering attacks 1 hour before was very unrealistic and always bugged me and even with the plan attacks features there was very few incentives to do so.
This feature actually makes a lot of sense plus it opens for more devastating to foil or be tip off from attack attempts.

You can still give orders 1 hour ahead so that "unrealistic" (I mean hey, it's a game. You also can suddenly start a production line within one hour without any preparation, or start an air superiority mission from one day to the other etc etc) part is still there and I bet that even with the battleplan feature that players who do it themselve will probably get a big advantage in doing so. And as I said, I've never said the battleplan feature makes no sense, it makes a lot of sense. I'm also very certain that I'll enjoy using the battleplaner and probably use it a lot, too, if just for the feeling of actually planning my operations instead of just roughly figuring them out in my head.

The only thing that bothered me is that it kinda felt as if players who want to focus on the war aspect of a wargame get a disadvantage for, well, wanting to very personally control the war, but I guess as long as the bonus isn't like "+50% combat strenght" (I would imagine it to be in the 10%-20% range) it is okay in order to make up for whatever mistakes the AI can (and probably will) make.

 

[JamesJameson]

I'm sure it's one quick text file away and it's fixed, as far as the penalty goes.

However I will keep it cause I use to hate that you could just put people on a boat and ship them across the ocean like nothing in HOI3. I think that to plan a landing like that, specially a hot one, it should take a certain number of ships in the attack group and should take a long time tell it can be executed. That's just my opinion though, I think it sounds way better for simulating the logistic nightmare of WWII.

 

[Hans_Schnitzel]

I'm sure it's one quick text file away and it's fixed, as far as the penalty goes.

However I will keep it cause I use to hate that you could just put people on a boat and ship them across the ocean like nothing in HOI3. I think that to plan a landing like that, specially a hot one, it should take a certain number of ships in the attack group and should take a long time tell it can be executed. That's just my opinion though, I think it sounds way better for simulating the logistic nightmare of WWII.

But even with battleplans you can ship people across the ocean like it's nothing. This is not an arguement against the battleplan, no, I'm just trying to clear something up. I mean, you can even draw a plan from one ingame hour to the other and then execute it right away, you don't have to wait a few ingame months or weeks before you can execute the plan, we don't even know if you have to wait to get the battleplan bonus or if it comes right away. The only thing I remember from Quill's video is that he said they will gain a bonus while executing the battleplan, I'm not sure about how and when they get the "preparedness" bonus aif it will increase with the time spend between planning and executing a plan or not. I do hope they add a time to loading transport ships though, with or without plan, it makes no sense that you can just load divisions worth of units into ships from one hour to the next.

 

[safe-keeper]

But even with battleplans you can ship people across the ocean like it's nothing.

If you can, you really shouldn't be able to. It should take time to organise and execute.

 

[Evilfisher]

has the Hoi4 soundtrack been released? i love the background music in OP`s video, if thats the Hoi soundtrack then its more awesome

 

[JamesJameson]

But even with battleplans you can ship people across the ocean like it's nothing. This is not an arguement against the battleplan, no, I'm just trying to clear something up. I mean, you can even draw a plan from one ingame hour to the other and then execute it right away, you don't have to wait a few ingame months or weeks before you can execute the plan, we don't even know if you have to wait to get the battleplan bonus or if it comes right away. The only thing I remember from Quill's video is that he said they will gain a bonus while executing the battleplan, I'm not sure about how and when they get the "preparedness" bonus aif it will increase with the time spend between planning and executing a plan or not. I do hope they add a time to loading transport ships though, with or without plan, it makes no sense that you can just load divisions worth of units into ships from one hour to the next.

Like you said we don't know if this is how it works, we know very little about how it will work in the release. From watching the game play it seems you don't have to wait for the bonus to kick in, though I imagine it will go through lots of changes before release.

 

[novapaddy]

My compliments to Podcat and Team as he definitely made the interface easier to understand.

Very impressive!

One suggestion: Possible to have a border skirmish between nations who have high World Tension between them? Like the way India have with Pakistan every now and then. I mean military action short of a total war effort.

 

[potski]

I mean, you can even draw a plan from one ingame hour to the other and then execute it right away, you don't have to wait a few ingame months or weeks before you can execute the plan, we don't even know if you have to wait to get the battleplan bonus or if it comes right away. The only thing I remember from Quill's video is that he said they will gain a bonus while executing the battleplan, I'm not sure about how and when they get the "preparedness" bonus aif it will increase with the time spend between planning and executing a plan or not.

One of the videos or written reports I'm sure mentioned a 15% bonus. This is not applied instantly, but builds up over time. So on the day you create the plan you are in the same situation as a micro-manager. The longer you allow your forces to prepare before the plan is executed the more bonus you get. But if there are spies, then the longer it is left the more chance it is discovered.

Plus, the enemy AI should have it's own plan. It was confirmed by the devs this will be the case - the AI countries will use the Battle Plan feature. That means the enemy are also gaining their own bonus, the longer you delay, the higher the enemy bonus.

If both sets of forces opposing each other on a front start preparing at the same time the bonus cancels out. What would be useful to know is whether you can prepare a Battle Plan when you are not at war. So, in that case, both Germany with an offensive plan and the Soviets with a defensive plan have no advantage because of the planning, at the start of Barbarossa.

The bonus is not applied on how you move your forces, but for planning. As long as you assign your Divs to a formation under a general, and create any sort of plan, then even diehard micromanagers get the bonus.

At least that is currently the case. One reviewer questioned this, whether it was entirely logical, and I guess there is a chance it might be changed. If you are preparing for an offensive and suddenly the enemy carry out a pre-emptive strike against you (so you are defending) should you still get the bonus?

But I think it could potentially have a gamey element. The micro managers get the "bonus" from being able to control the exact movements/attacks of their forces, better in some cases than the AI. But their real plan is in their head and is not discoverable by enemy intelligence.

This implies to me that the micro managers have an advantage. But only on single player. I don't see anyone successfully micro-ing all of the units of a major power on the speed in the video, and with no option to pause. And presumably, this will become the norm in MP.

 

[Moshi]

has the Hoi4 soundtrack been released? i love the background music in OP`s video, if thats the Hoi soundtrack then its more awesome

I have to agree with this. Soundtracks sounds amazing so far! More info on this if possible, please!

 

[Porkman]

What I thought would be a nice feature is that players could make Battle Plans for a country and save them to a file. Things like 'The Defense of Russia 01', 'What France Should have done!' or 'Japan's Seaborne invasion of Southern China'. These would then be loaded into the game, and they would give options for the AI. As in Soviet AI (not a player playing the Soviet Union) would check the situation (threat levels on which border, size of his army, other things) and look at it's filed Battle Plans for ones matching the situation and say it has 5 plans that fit the situation and randomly (some plans may have a higher % than others) picks one. When the objective is achieved or the situation fails apart the AI goes back and looks for the best plan(s) that match the current needs and picks one. This way you have Humans making the plans and with the random chance the player doesn't know which one. If the AI has 10 different plans for the defense of France, the player does know which one the AI picked.

As time goes on players post their plans for others to download to their game so eventually there would be many plans covering many different situations. At this point I am thinking this could be a DLC (not in the game as released). What do you all think?

This is an amazing idea but I would make it so it would be even more passive.

The game Spore would automatically upload player created creatures and then use them to populate the worlds of other players. It was a nice feature.

So here's what happens, the game is set to periodically (maybe on autosave) save the plans that a player makes, whether it's the invasion of Barabarossa or just a 3 province mop up operation in Zhejiang.

That file then gets sent online automatically to a database of plans that paradox hosts. The player could then set the game to allow for the computer to talk to the plan database periodically and pull plans from there.

This is how they programmed the original deep blue, they just fed it the results and exact moves from thousands upon thousands of chess games.

Definitely make a thread about this.

Good attributes of this idea.

1) A better AI - the AI will compare it's current situation and force composition and select a similar plan that a player has uploaded, hopefully leading to a better outcome.

2) the benefits of crowd sourcing - It is impossible for the AI programmers to individually consider the innumerable game states that a game like HOI4 creates and design AI plans for them (HOI2's AI was actually coded like this, where the AI would shift to different production and province priorities based on events) . But with player uploaded plans, that's thousands of game states right there. If there is a way to record the success of the plan and weight the AI to choosing the most successful plan that's even better.

3) Expandibility?

Maybe a similar idea could be extended to build schemes and Intelligence moves... with AI's that follow the build strategies or diplomatic moves of the most successful players.

Difficulties

1) Memory and processing...

The plans would have to not only save the plan... they'd also have to save the info about forces on both sides. HOI save files are huge and for this idea to work well, the "plan files" being uploaded and downloaded would have to be considerably smaller.

2) Evaluation

Someone would have to code an AI script that tells the AI how to evaluate the plans. How much would the uploaded plan's situation have to match the AI's situation before hte AI would pick it? How much weight would the AI put on having similar air and naval assets? A similar tech level? A similar enemy tech level?

3) Consequence of success?

If this works well, it might work too well, with the player facing an AI that's as good as the top X percentage of players.... hmm...

It's an interesting programming challenge...

But Gamer_1745, you should definitely start a new thread on this.

 

[BrotherArdis]

If this works well, it might work too well, with the player facing an AI that's as good as the top X percentage of players.... hmm...

Considering how weak the AIs in most modern games are, I would be delighted to play a game where the computer opponent can actually stomp me into the ground. I don't think that simply using human-made battle plans would make the HoI AI so dangerous, especially that it would have a chance (risk) of using failed plans as well. But should this problem really arise, it can easily be solved with a simple option, "use downloaded plans: yes/no".

 

[Beagá]

I have to agree with this. Soundtracks sounds amazing so far! More info on this if possible, please!

I would also ask, right now, to allow people with HOI3 to download HOI3 soundtrack for free as DLC when you buy HOI4.

 

[LostAlone]

Considering how weak the AIs in most modern games are, I would be delighted to play a game where the computer opponent can actually stomp me into the ground. I don't think that simply using human-made battle plans would make the HoI AI so dangerous, especially that it would have a chance (risk) of using failed plans as well. But should this problem really arise, it can easily be solved with a simple option, "use downloaded plans: yes/no".

That's kinda funny.

In most games it's really not hard to make an AI that plays perfectly and can absolutely stomp essentially any human player. The devs tweak the AI down so that people can actually play the game and have fun. Even on the high difficulty levels most games typically stick with the same AI but give them other bonuses, which makes for a harder game but leaves exploitable weaknesses in the AI for players to take advantage of. Even if you look to Dark Souls where the AI is designed to kill you very many times, there are obvious flaws because you are supposed to win eventually. If they wanted to make enemies you couldn't beat they could, they just don't see why you would want that.

It is NEVER because they can't make an AI good enough to face your titanic mind, it's because the AI is supposed to be beatable even on the highest difficulty levels players have access to. If you want an AI that you can't ever beat that can be arranged, but you won't ever get better than it. Computers are simply better at handling large amounts of information than players are and they can 'see' the game in far greater depth - knowing exactly how long it'll take every unit to move to any other province, how long a smaller force will delay a larger one etc.

To put it mildly, you wouldn't want to play a game where the AI can stomp you. You might want a challenging game, even a very hard game, but if you take the stops off the AI then the AI is going to beat you.

 

[Modestus]

That's kinda funny.

In most games it's really not hard to make an AI that plays perfectly and can absolutely stomp essentially any human player. The devs tweak the AI down so that people can actually play the game and have fun. Even on the high difficulty levels most games typically stick with the same AI but give them other bonuses, which makes for a harder game but leaves exploitable weaknesses in the AI for players to take advantage of. Even if you look to Dark Souls where the AI is designed to kill you very many times, there are obvious flaws because you are supposed to win eventually. If they wanted to make enemies you couldn't beat they could, they just don't see why you would want that.

It is NEVER because they can't make an AI good enough to face your titanic mind, it's because the AI is supposed to be beatable even on the highest difficulty levels players have access to. If you want an AI that you can't ever beat that can be arranged, but you won't ever get better than it. Computers are simply better at handling large amounts of information than players are and they can 'see' the game in far greater depth - knowing exactly how long it'll take every unit to move to any other province, how long a smaller force will delay a larger one etc.

To put it mildly, you wouldn't want to play a game where the AI can stomp you. You might want a challenging game, even a very hard game, but if you take the stops off the AI then the AI is going to beat you.

What exactly are you talking about here the ability of the AI to out think a player or bonuses that allow the AI to survive against a player long enough to win?

 

[Porkman]

That's kinda funny.

In most games it's really not hard to make an AI that plays perfectly and can absolutely stomp essentially any human player. The devs tweak the AI down so that people can actually play the game and have fun. Even on the high difficulty levels most games typically stick with the same AI but give them other bonuses, which makes for a harder game but leaves exploitable weaknesses in the AI for players to take advantage of. Even if you look to Dark Souls where the AI is designed to kill you very many times, there are obvious flaws because you are supposed to win eventually. If they wanted to make enemies you couldn't beat they could, they just don't see why you would want that.

It is NEVER because they can't make an AI good enough to face your titanic mind, it's because the AI is supposed to be beatable even on the highest difficulty levels players have access to. If you want an AI that you can't ever beat that can be arranged, but you won't ever get better than it. Computers are simply better at handling large amounts of information than players are and they can 'see' the game in far greater depth - knowing exactly how long it'll take every unit to move to any other province, how long a smaller force will delay a larger one etc.

To put it mildly, you wouldn't want to play a game where the AI can stomp you. You might want a challenging game, even a very hard game, but if you take the stops off the AI then the AI is going to beat you.

Someone who knows more about AI theory can probably explain better why you're really, really wrong but here goes my attempt.

HOI4 is an infinitely more complicated game than Dark Souls or Chess for example.

Human brains are the most efficient and effective pattern recognition and rule making objects on the planet.

A human brain can figure out to navigate and survive in the infinite complexity of the real world.

Humans are very good at prioritizing and making heuristic rules. Give a computer a huge ton of data and they might make a good evaluation but generally they are bad at evaluating out irrelevancies.

The AI in HOI or strategy games has to be given artificial "environmental" buffs because the programmers don't know how to program a perfect AI.

Yes, the AI has theoretical access to perfect information. It is omniscient but not smart. Sure it knows that the line is thin in province X, so it should attack... but it's actually a plot! You placed strong backup forces three provinces back to aid in the creation of the pocket.

So the programmers tell the AI to not attack if there are backup forces... problem solved!

Except, maybe I like putting in weak backup forces and instead preload some units a province or two away along the line from the breakthrough province and then move them in quickly to cut off the spearhead from the sides.

So the programmers tell the AI not to attack if there are possible backup forces near the weakly defended province... problem solved.

Except now I'm leaving three provinces undefended because I'm going to do a massive encirclement Incheon style with prepositioned marines around the back of the enemy once the front line moves forward...

There is too much information for the programmers to make an AI that can effectively evaluate, prioritize, and react to it as it comes. Humans are far better at quick prioritization, evaluations, and reaction. Computers can do these faster, but we haven't figured out how to make them do it well.

Feeding the AI tens of thousands of battle plans might cause the AI to play well by having several good paths to mimic.

In Chess, with it's limited amount of game states and simple rules, they have made unbeatable AI's. Chess has 64 squares and 6 different kinds of pieces. There are still more possible chess games than there are atoms in the universe.. Still, since Chess is a pretty limited game in complexity, we've figured out how to program computers to use their strengths on that limited complexity to beat human every time.

. Now compare that to HOI4 with thousands of provinces and hundreds of divisions. You can make a computer see it... but short of having it autoroute naval forces to supposedly undetected taskforces, there's not much the computer can do to effectively use that information.

 

[BrotherArdis]

That's kinda funny

(...)

I didn't think I'd actually have to explain this, but very well...

Yes, I'm aware that AI is nowadays executed by way of "artificial supidity" with gazillion bonuses. I really have noticed that. But it doesn't make for a challenging OR fun game, contrary to what you're saying. Once you find the intentional weaknesses, any challenge goes out the window. Same for fun. There may be frustration, when the bonuses make the game borderline unbeatable, but nothing else. And no, I don't want to be roflstomped every time, all the time, either. I'm not a masochist, I don't take pleasure from being hurt. But I'm tired of being treated like a retard. I want an AI that at least tries to think and adapt. I want an AI that can beat me because it outsmarted me (even if it makes me look dumber than an amoeba), not because it had bonuses to production, research and unit strength.

I remember FPS games in the early 2000s that were making actual progress in this department. Half Life or FEAR didn't have to resort to perfect reflexes, 100% accuracy, millions HP or quad damage. These games created a challenge by presenting you with opponents that BEHAVED in a pretty smart and reasonable way. So it can be done. I know it's difficult, and AIs in strategy games have to deal with more complex issues, but this really shouldn't stop developers from trying (which is not an accusation towards Paradox, mind you).

It is NEVER because they can't make an AI good enough to face your titanic mind, it's because the AI is supposed to be beatable even on the highest difficulty levels players have access to.

When I watch the so called "AI" in Civilization 5 shuffle its units for 10 or 20 turns and still not come up with any sort of formation, then lose all those units because it can't notice 2 or 3 ranged units of my own near the target city, or abort its plans altogether because it can't deal with terrain, I'm pretty sure it's not out of care for my ability to beat the game. And I said nothing about the quality of my brains, so that sarcasm is unnecessary. As for difficulty levels, I really see no reason why all should be guaranteed to be beatable. When it says "impossible", it should be reasonable to assume it probably is nigh on impossible.

 

[LostAlone]

I'm always glad to see that I get taken out of context three times in three posts, especially by the person that I originally answered.

My point is that modern AIs can play better than most players and are better at dealing with complex systems then human players. No, I don't think AIs are somehow magic, that they are intelligent in the same sense as people, but their ability to handle large amounts of data and find the most optimal moves is radically better than humans. In fairly simple systems, AIs are about equal to the best humans to ever play the game (Chess) In more complex systems, AIs have a major leg up.

And all I meant by saying that was simply that you really don't want a game where the AI can just stomp you without cheating. You don't. Because you won't beat it. There might be ways to exploit it's behavior to win, but you won't out-play it, you'll just make it beat itself. If you want a hard game, ask for a hard game, that's fine. But you don't want a game that turns you into a fine paste with no possibility of victory over and over. I mean, why would you want that?

 

[Porkman]

I'm always glad to see that I get taken out of context three times in three posts, especially by the person that I originally answered.

My point is that modern AIs can play better than most players and are better at dealing with complex systems then human players. No, I don't think AIs are somehow magic, that they are intelligent in the same sense as people, but their ability to handle large amounts of data and find the most optimal moves is radically better than humans. In fairly simple systems, AIs are about equal to the best humans to ever play the game (Chess) In more complex systems, AIs have a major leg up.

And all I meant by saying that was simply that you really don't want a game where the AI can just stomp you without cheating. You don't. Because you won't beat it. There might be ways to exploit it's behavior to win, but you won't out-play it, you'll just make it beat itself. If you want a hard game, ask for a hard game, that's fine. But you don't want a game that turns you into a fine paste with no possibility of victory over and over. I mean, why would you want that?

I also want an AI that's not unbeatable... but you're missing my point. Computers aren't magical. HOI4 is very complex with thousands of different tasks and thousands of possible win conditions that can be achieved in one thousand different ways.

I was objecting to your blythe assertion that they could invent the "winningnest AI that doesn't cheat" and just refuse for game play reasons. They can't, they really can't.

Human brains take a lot of information in, subconciously select the relevant bits, and then send it up to our conscious brains for further action.

This is crippling for computers.

Think about the physics involved in throwing a ball 30 feet into a basket. You stand there, shoot, and make a basket... simple. Except it's not.

Your eyes guage the relative position of you and the basket, this is a complex operation that mathematically would take some nice trigonometry to express to a computer. Your nerves in your arm guage the weight of the ball and then they go back up to the brain which subconciously does it's own set neural paths that take into account height, relative position, weight, force needed, the drop in the parabolic arc of the ball, the possibilty of the bounce off the basket, the amount of signal to send to each muscle in your arms for the proper force, the path your arms need to take...

Break it down to throwing the ball and you can see that your brain is doing a very complex operation and set of calculations. But only a little of this is conscious, and it would take a lot of programming to get a computer to do the same thing, since it would have to be explicitly told what to do.

When we play HOI, we don't count our force ratios and calculate the optimal amount really... we have a feeling and then we weigh what we want to accomplish with what we can build and weigh force composition... there are many, many variables that go into. Again, many of them unconscious.

Another thing that we are good at, which computers are bad at, is ignoring information when it's not relevant. At a certain level of complexity computers start being at a disadvantage again since they can't prioritize as well as we can.

A few years ago, DARPA the American military research agency, held a great robot race. The goal was for several teams to compete and produce a self driving vehicle that could make it through 100 miles of NEvada desert.

Teams from the best universities and companies designed vehicles, giving them cameras, infrared sensors, radar, sonar... laser finders... basically, these machines had all of the survellance technology available. They also had the routes waypoints pre programmed. The course was long, but not hazardous.

The race started and none of the cars finished. The best went for 8 hours and then just stopped.

What caused the failure in most cases was the surveillance systems. Essentially, one of these innumerable systems would point out a rock that wasn't there, and then the care would refuse to drive forward.

They held the race again a year later. This time 5 cars finished the track. One of the programmers of the winning car (standford) said that the key was teaching the computer how and when to ignore all of this surveillance data. If the sonar saw a rock but the radar didn't who was right, what if the laser finder also saw the rock?

This is how we have self driving cars now.

If you threw the same amount of money to developing an optimal HOI AI that the people have thrown at developing a self driving car, you could do it. But like self driving cars, it would take a decade and more money than Paradox will make in a 100 years.

Paradox does not have the resources to develop an amazing precoded AI. They will be working their ass off to build an AI that's mildly competent and that is going to be super, super difficult.

HOI is not a game where a programmer could build the optimal AI. It would have to be taught and it would have to be taught via brute force.

 

[Gamer_1745]

This is an amazing idea but I would make it so it would be even more passive.

Good attributes of this idea.

..Snip..

If this works well, it might work too well, with the player facing an AI that's as good as the top X percentage of players.... hmm...

It's an interesting programming challenge...

But Gamer_1745, you should definitely start a new thread on this.

Thanks for your support and I like your additional ideas very much! I don't even know if my ideas are feasible with the current state of the HoI engine and I don't even know how well the AI understands battle plans. Some of your ideas are very good, but they seems to me to be a bit ambitious. Evaluation of plans is very good, just not sure how feasible it would be, that is why I think a human evaluator at one stage or another in the process that could give each set of plans a 1-4 rating would give the better one greater wieght. I think even just picking a set of plans that fit current conditions at random would be better that what I have seen so far from HoI AI (HoI3 playing & HOI IV videos)!

 

[BrotherArdis]

And all I meant by saying that was simply that you really don't want a game where the AI can just stomp you without cheating.

As a matter of fact, I do. I do want an AI that can kick my ass with a very heavy boot. And when it does, and I can find no way of beating it, I'll lower the difficulty setting. I'll finally have a game that is actually challenging.

Because right now, there's no challenge. AIs in most of current games are hopeless on equal terms, and the bonuses they receive on higher difficulties aren't really doing their job. Either you can compensate for them without effort, once you learn a few simple tricks, or they are so great that you just can't make up for them at all.

[for example, the AI in Command & Conquer 3 is made exactly the way I like it. At brutal, it just plays way better than me, without any artificial bonuses. So I switch it a notch lower. But C&C isn't a grand strategy game dealing with WW2, sadly.]

I'm always glad to see that I get taken out of context three times in three posts, especially by the person that I originally answered.

Not sure where you see "three times", I was the only one to quote a specific sentence from your post. But anyway, where did I put your words out of context? You said "it's never because they can't make an AI good enough" and I objected to that, plus I commented on the rest of the sentence. That I addressed it separately from the rest of your statement didn't twist its meaning.