WWW - Is the AI too timid?

[Gort11]

Maybe the balancing early on for these minors is that their divisions have too little strength and start with too little equipment.

Yeah, I do wonder if one of these is the issue:

1. AI doesn't build much infantry equipment so its divisions are under-strength
2. AI doesn't research doctrines as a priority
3. AI doesn't make planning bonuses against countries that can attack them

Something seems off, since Czechoslovakia should certainly not have been the cakewalk it ended up being in the stream - even Daniel was expecting a tougher fight, which is why he saved the game before the declaration of war. In practice, the Hungarian divisions defeated the Czech divisions pretty much everywhere, which points to some kind of large-scale factor being neglected somewhere.

Here's an interesting idea - is Grand Battle Plan doctrine actually the best doctrine for blitzkrieg? Build up a nice 100% planning bonus against your enemy pre-war (thus dodging the problem of how long it takes to get such a big bonus since you're not actually at war) and then take their country before it decays. Even if your opponent makes a plan against you pre-war, you'll still be 50% ahead of him, and it seems like that bonus would be better than the bonuses any of the other doctrines might give in a short war.

And therein lies the problem.

It's good that Daniel's streams highlight these issues, though. It'd be stupid if infantry spam was the optimal tactic but all Daniel ever showed us was sweeping tank encirclements and concentrated air power, then we all get the game and find out that infantry spam is way more effective.

I do hope they make pure infantry less effective - WW2 was generally a war in which "new" (tanks, air power, trucks, submarines, carriers) weapons were proven to be more effective than "old" weapons (entrenched foot infantry, battleships, cavalry), and I think the game balance should reflect this. Spending industry on tanks should generally give you a better military than spending it on infantry, to give one example.

 

[Kazansky22]

@SteelVolt

I am wondering if you can tune the AI to a way that say. British AI tend to only move in big massive formations maintaining a solid battle line. and Maybe a more of a defense in depth by the Soviets, and more of a fast, take risks style for Germany. Or have the AI's tendencies be tied in with the type of land doctrine the player has chosen to elect.

 

[Oriflamme]

This is one of the big things I've noticed with these WWW videos. No significant use of combined arms or maneuvering, just steamrollers, all the time.

 

[Axe99]

But not heavy, and not in large numbers.

It is warfare, in every sense, since the rebels and ISIS do hold ground and administer their own lands. There are "front lines" and fortresses and all that.

That's fair - apologies for not being on top of Syria, just haven't had the capacity.

 

[AgeOfHades]

so i'm not sure if this 100% relates to the most current portion of this conversation, but a few pages ago i recall people saying daniel almost never uses tanks or anything (or something similar to this effect), whether it be in the newest streams or in the multiplayer one.

i seem to recall the front line against the Raj (ofc this is in the multiplayer WWW) stalling until he sent some Armoured divisions over, not long later we seem them absolutely slaughter the Raj.

ofc i can't say how much johan contributed to their defence (i don't recall being shown if he was ever actively defending there. that and it's a really old build now)

but daniel has used tanks and they have served him well where used.

(just wanted to get my random say after reading through all the pages)

 

[henzington]

Its funny as during the Japan WWW people were complaining about the AI being too aggressive.

 

[AgeOfHades]

Its funny as during the Japan WWW people were complaining about the AI being too aggressive.

that was complaint worthy? i thought it was great how they actually pushed through his lines, thought it didn't exactly go well for the chinese. maybe that's why they're complaining?

 

[jjjdddsss]

I think it is important to note that all the relevant Gameplay we have seen has involved the AI being significantly weaker, either in tech or numbers. It's hard to say how more even fights will play out.

 

[KalZakath]

Yeah, I do wonder if one of these is the issue:

1. AI doesn't build much infantry equipment so its divisions are under-strength
2. AI doesn't research doctrines as a priority
3. AI doesn't make planning bonuses against countries that can attack them

Something seems off, since Czechoslovakia should certainly not have been the cakewalk it ended up being in the stream - even Daniel was expecting a tougher fight, which is why he saved the game before the declaration of war. In practice, the Hungarian divisions defeated the Czech divisions pretty much everywhere, which points to some kind of large-scale factor being neglected somewhere.

Here's an interesting idea - is Grand Battle Plan doctrine actually the best doctrine for blitzkrieg? Build up a nice 100% planning bonus against your enemy pre-war (thus dodging the problem of how long it takes to get such a big bonus since you're not actually at war) and then take their country before it decays. Even if your opponent makes a plan against you pre-war, you'll still be 50% ahead of him, and it seems like that bonus would be better than the bonuses any of the other doctrines might give in a short war.

Moreso I think is a problem that could be almost impossible to fix... The AI doesn't start gearing up for war from Day 1, while the player does. Realistically, it's Jan 1, 1936, and you as an AI minor have to be geared toward being ready for war (or about the strength that you should be) by 1939. You start with a plan that will get you there. The problem, is that your opponent is gearing for war in March of 1936.

So, do we have to add an intelligence gathering factor? Spies to be able to tell us what our neighbors are doing so we can detect it and counter it? Probably the best thing, but the espionage side of the game really isn't in yet fully (at least in my eyes).

Do we start the game with everyone gearing for a war that starts in 1936? If that's the case, get ready for some monster 'minors' by 1939.

So far all of the WWWs have been played with the national decisions made on Historical mode. Maybe if you play a minor like Hungary, the correct setting should be 'Ahistorical mode?'

SteelVolt - are the minors more likely to be prepared for war earlier on Ahistorical mode than on Historical?

thanks.

 

[fabius]

Don't look now, but you may find you've actually been here for years . Totally agree with what you say though - PI/PDS are a cut above with how they treat their players, and have a wonderful passion for making great games. There are other great strategy devs out there, but I don't know of anyone in the same class as the Paradox crew.

Oh my, is that the year... I must save-game, grab something to eat, and maybe celebrate a birthday or two.

Agreed, they are a cut above. And that's why I'll probably get Stellaris at full price even before I have a space itch.

 

[Jeremy971]

He plays on normal difficulty. HOI3 on normal difficulty is a joke as well...

 

[dburiany]

Moreso I think is a problem that could be almost impossible to fix... The AI doesn't start gearing up for war from Day 1, while the player does. Realistically, it's Jan 1, 1936, and you as an AI minor have to be geared toward being ready for war (or about the strength that you should be) by 1939. You start with a plan that will get you there. The problem, is that your opponent is gearing for war in March of 1936.

So, do we have to add an intelligence gathering factor? Spies to be able to tell us what our neighbors are doing so we can detect it and counter it? Probably the best thing, but the espionage side of the game really isn't in yet fully (at least in my eyes).

Do we start the game with everyone gearing for a war that starts in 1936? If that's the case, get ready for some monster 'minors' by 1939.

So far all of the WWWs have been played with the national decisions made on Historical mode. Maybe if you play a minor like Hungary, the correct setting should be 'Ahistorical mode?'

SteelVolt - are the minors more likely to be prepared for war earlier on Ahistorical mode than on Historical?

thanks.

I think KalZakath and Gort11 talk about the source of the issue - how the human player in 1936 knows to immediately focus on war while the AI doesn't. For example, in WWW Hungary has an army of ~30 div @ 10k for 300k in 1937 while Wikipedia says it should have been 85k. That could explain why Austria and Czechoslovakia were so unprepared - because everyone was in 1936. Plus, think of what really happened: occupation of Rhineland - no response, occupation of Austria: no response, occupation of Czechoslovakia - no response. So when the AI Yugoslavia doesn't prepare for war after those happen, well, that's at least somewhat realistic, right? So what are the solutions?

Take away the 1936 start so humans don't get an unfair head start on remilitarization? Then you can't do things like play a naval focus Germany. (Plus there are many people who want a 1933 start which could only make the problem more pronounced.)

Have the AI act as if war starts in 1936? My guess is that could have the side effect of escalating to WW2 starting in '37 every time.

Create an artificial build limit so no one can build divisions or guns or whatever before 1938? That doesn't seem right.

Maybe let the AI save-scum - force a reload of a 1936 autosave so it can then prepare for the invasion it now knows is going to happen (like we do).

In conclusion, I don't know.

 

[Gort11]

The AI can't be simply not preparing for war, since it has to be using its military factories for something. Maybe it's building things other than infantry equipment (Czechoslovakia might have been building planes, maybe? It had enough fighters that Daniel didn't even deploy his airforce) or it hasn't been building up planning bonuses against its neighbours while at peace.

 

[KalZakath]

Yeah Gort - but, how is it preparing for war? It's not at war, and there's little in the way of world tension to make it worry to much? Its conscription laws are probably as lax as they go (remember how Da9l only had like 25k manpower until he upped his?), its production is probably heavy into making more consumer factories rather than producing max weapons.

Maybe there needs to be another stat I'll call Regional Tension? Yes, the majors might not be horribly worried about anything, and the US might not care at all, but if Hungary goes after Austria, the regional states (Czechoslovakia, Romania, Yugoslavia, Italy, and maybe even Bulgaria and/or Greece would definitely take notice of it and gear up as if it were '39 instead of '36, upping conscription laws, mobilizing reserve divisions, at least taking some steps to protect against the already shown to be aggressive neighbor.

Not sure, though - that would be able to probably be gamed too by someone to make powerful neighbors to buffer against an antagonist, lol.

 

[Gort11]

its production is probably heavy into making more consumer factories rather than producing max weapons.

This isn't something you can do in HoI4. Military factories can't build civilian factories. They have to be making military stuff. The question is, what military stuff?

I still think it's more likely to be a planning bonuses thing - the Czechoslovakian divisions didn't look low on strength.

 

[KalZakath]

Hmm - thought I remembered something from the WWW about the trees that Da9l was talking about that had him choose between upping the civilian factories - can't find it now so I definitely could have been mistaken. Was thinking that they might have chosen that over something that would help them quickly build more arms.

 

[AgeOfHades]

Hmm - thought I remembered something from the WWW about the trees that Da9l was talking about that had him choose between upping the civilian factories - can't find it now so I definitely could have been mistaken. Was thinking that they might have chosen that over something that would help them quickly build more arms.

maybe you're thinking of when the differences between concentrated and dispersed industry was much bigger? (with concentrated being more or less a no brainer at that stage)

 

[Betawerewolf]

Anybody here who has played XCOM 2 will remember that there were a lot of issues with the AI in that game too. I remember when one guy, Beaglerush, found a gamey tactic in the game which caused Firaxis to fix the AI to counter the "Beagle Manuever". AI issues will always exist but can be fixed by modders or hotfixed by Paradox.

As for the WWWednsday vids

The multiplayer session saw the Soviet Union ROFLstomped because of an issue in that build which screwed over the Soviets. It's probably the only reason why the Axis won the war as we could see that Daniel's forces were starting to become overextended (Johan's sustained invasion of Italy and Daniel's failure to invade North Africa).

The Sino Japanese war was won with infantry that is true but remember that the Chinese were ill-equipped and Daniel did use cavalry, navy, and motorized divisions to crush their nation.

The Austro-Hungarian war I agree was completely unrealistic but that is because nobody joined in the war to fight the Hungarians. If we look at Darkest Hour for example in the 1936 scenario, you can see that the Austrian's independence is guaranteed by the United Kingdom, Italy, and Hungary. Now I'm not saying we just copy and paste Darkest Hour but I do think that the League of Nations or Italy would've intervened if there was a war that close to France.

 

[Gort11]

The Sino Japanese war was won with infantry that is true but remember that the Chinese were ill-equipped and Daniel did use cavalry, navy, and motorized divisions to crush their nation.

His navy never fought and he had one motorised division in total.

What I'd hoped to see was Daniel grouping his more mobile units under a general with relevant traits and then giving them specific orders to advance in terrain that favoured them. In the Japan video Daniel just chucked them all under a field marshal and treated them like infantry. In the Hungary video Daniel straight-up converted all his non-infantry units into infantry units.

If your best player's strategy is to spam a single type of unit, you've got problems.

 

[AgeOfHades]

His navy never fought and he had one motorised division in total.

What I'd hoped to see was Daniel grouping his more mobile units under a general with relevant traits and then giving them specific orders to advance in terrain that favoured them. In the Japan video Daniel just chucked them all under a field marshal and treated them like infantry. In the Hungary video Daniel straight-up converted all his non-infantry units into infantry units.

If your best player's strategy is to spam a single type of unit, you've got problems.

to be fair for the Hungary one, they were only horrible cavalry divisions with 3 brigades in them.

though i will say that even tho his infantry ended up with 10 brigades, i recall they either only had 1 or no support companies, so he probably should have had a far, far harder time against the czech.