Home Trending Latest threads New posts Developer diaries

As USA : Japan battleships and carriers : what are they doing ?

Jump to latest Follow Reply
  • We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Hearts of Iron IV 49403 Bug Reports 18246 After Action Reports (AAR)
Well, if the others’ analyses are correct, then technically one solution would be to give the AI priority to fill strike forces first and only then the patrol groups. But that could also cause the opposite problem, where they end up with no patrol groups at all, making their strike forces stay in port forever because they never find the enemy, huhauauha.


One thing I thought could help would be a new balance pass on screen ratios. I think I saw a thread about this a long time ago, but maybe reducing the requirement from 3 screens to 2 screens per capital ship could be very interesting. From what I found using ChatGPT (because I don’t have time to check proper sources right now), in real life most fleets had something like a 1:1 or 1:2 ratio of capital ships to screening ships (within the game’s classification). This would reduce the demand for screens in task forces and allow them to be distributed into other roles.


Analyzing it a bit, it's true that this could bring back the meta of light cruisers focused on tons of light battery turrets, pushing torpedo ships—like the special torpedo cruiser design—back into irrelevance. So instead of having 100% screening efficiency, the maximum value could be lowered to something around 90–80%. Many other factors would also need to be analyzed to avoid making torpedo ships become the new meta, but I think it could be a starting point.
 
  • 2Like
  • 1
Reactions:
One of the problems with AI fleets is that when damaged, they retreat to the nearest port. And those ports are usually very low-level naval bases, so repairs take a long time. I'm currently playing as the Great Britain, and I constantly have to manually move ships to larger ports for repairs, as well as replace damaged ones with healthy ones. Naturally, the AI either doesn't know how to do this, or does it poorly. For example, it sends capital ships unescorted, where we catch them.
 
  • 7
Reactions:
ljekio said:
One of the problems with AI fleets is that when damaged, they retreat to the nearest port. And those ports are usually very low-level naval bases, so repairs take a long time. I'm currently playing as the Great Britain, and I constantly have to manually move ships to larger ports for repairs, as well as replace damaged ones with healthy ones. Naturally, the AI either doesn't know how to do this, or does it poorly. For example, it sends capital ships unescorted, where we catch them.
Click to expand...

It's very bizarre to me that ships still go to random naval bases when they re-added the Home Base function. By default a TF/Fleet should go back to its Home Base. It should be a toggle button to determine where they go.
 
  • 5
Reactions:
goodcigar said:
It's very bizarre to me that ships still go to random naval bases when they re-added the Home Base function. By default a TF/Fleet should go back to its Home Base. It should be a toggle button to determine where they go.
Click to expand...
They are very vulnerable if they are attacked while damaged, especially if they have received Critical Hits. In the new system their Home Port might still be some distance away and on the other side of a sea zone swarming with enemy (e.g. German subs having to pass through the English Channel).

There should a heuristic to control this, at least a define or ideally a per-country number. If Strength is above the number the ship goes to Home Port; if Strength is below, go to the nearest port. Set it at a reasonable guess (half of the Repair setting?).

And then in an ideal world PDX run 50 campaigns to 1945 overnight, then change the number by 5 percentage points for a single, run another 50 campaigns overnight, and use the better number, repeat ad infinitum until they get a good number for each AI. I know PDX does run overnight tests but I've never heard them mention them using it to scientifically test values in this way. I guess the problem is that you can't run HoI4 without a decent iGPU and that makes it automated testing expensive.
 
  • 2Like
Reactions:
vermicious knid said:
Once I figured out what was happening I was cracking myself up imagining the glitch was AI admirals literally refusing to compromise on the number of available destroyers.
Click to expand...
Thanks a lot for the custom patch ahah.

I just have a question : do you think that the ai_will_do factor impacts it as well ?

I saw for example JPA has :

* 5 for concoy escort
* 10 for dominance
* 10 for patrol
* only one for strike forces

Maybe if we have something more balanced and more logic to me like :

* 2 for convoy escort
* 2 for dominance
* 4 for patrol
* 4 for strike forces

it can impact as well...

I will try to do some tests.
 
  • 2Like
  • 1
Reactions:
vermicious knid said:
OK, it isn't fancy but it should allow the AI to actually play the naval game until paradox can provide a more elegant solution.

steamcommunity.com

Steam Workshop::Perhaps a little compromise...

steamcommunity.com steamcommunity.com
Click to expand...
A friendly suggestion, consider inserting "navy" or "navy" "fix" (without quotes) or similar in the title so people typing to search for a mod that will make "AI's navy work better" might find it easier...

In any instance, THANKS SO MUCH for doing this
downloading and adding it !

For those that don't know

AI vs AI battles ( observer mode),
hit tilde ~ type "observe" without quotes in the fill in box that appears
then hit ENTER and watch the ant farm! See what they do with their boats, revisit every 30m or so
(FYI hit tilde ~ again to dismiss window)
 
  • 6Like
Reactions:
hnr said:
Thanks a lot for the custom patch ahah.

I just have a question : do you think that the ai_will_do factor impacts it as well ?

I saw for example JPA has :

* 5 for concoy escort
* 10 for dominance
* 10 for patrol
* only one for strike forces

Maybe if we have something more balanced and more logic to me like :

* 2 for convoy escort
* 2 for dominance
* 4 for patrol
* 4 for strike forces

it can impact as well...

I will try to do some tests.
Click to expand...
I didn't touch any of the weights. I imagine it would take a lot of trial and error to tune them.
 
  • 2
Reactions:
Thanks for the mod, I'll definitely look at the results as well.
 
  • 2Like
Reactions:
hnr said:
Thanks a lot for the custom patch ahah.

I just have a question : do you think that the ai_will_do factor impacts it as well ?

I saw for example JPA has :

* 5 for concoy escort
* 10 for dominance
* 10 for patrol
* only one for strike forces

Maybe if we have something more balanced and more logic to me like :

* 2 for convoy escort
* 2 for dominance
* 4 for patrol
* 4 for strike forces

it can impact as well...

I will try to do some tests.
Click to expand...
I would discourage messing with these. In my test games, the AI is creating Strike Forces and assigning them to sensible places. They're just not coming out to fight. I would want to see whether the changes in this mod are sufficient to fix that issue, and whether that reveals problems elsewhere, before changing anything else.
 
  • 2
Reactions:
Had a short test game with the mod on (the usual setup - played Haiti with max intel on all the major powers and fog of war off) and this is what I saw:

UK Navy:

12 March 1940: 3 BC, 12 BB, 7 CV, 176 DD, 17 CA, 48 CL, 33 SS
12 March 1941: 3 BC, 12 BB, 7 CV, 222 DD, 17 CA, 50 CL, 33 SS
12 March 1942: 3 BC, 12 BB, 8 CV, 230 DD, 17 CA, 51 CL, 36 SS
12 March 1943: 3 BC, 15 BB, 8 CV, 233 DD, 18 CA, 49 CL, 42 SS

Italian Navy:

12 March 1940: 7 BB, 99 DD, 13 CA, 14 CL, 54 SS
12 March 1941: 8 BB, 21 DD, 4 CA, 4 CL, 10 SS
12 March 1942: 9 BB, 23 DD, 3 CA, 4 CL, 2 SS

German Navy:

12 March 1940: 2 BB, 24 DD, 4 CA, 12 CL, 26 SS
12 March 1941: 2 BC, 2 BB, 28 DD, 3 CA, 15 CL, 38 SS
12 March 1942: 2 BC, 2 BB, 25 DD, 3 CA, 15 CL, 47 SS
12 March 1943: 2 BC, 1 BB, 24 DD, 2 CA, 14 CL, 11 SS

US Navy:

16 August 1941: 15 BB, 7 CV, 192 DD, 20 CA, 35 CL, 84 SS
16 August 1942: 15 BB, 8 CV, 165 DD, 17 CA, 34 CL, 86 SS

Japanese Navy:

16 August 1941: 3 BC, 6 BB, 7 CV, 136 DD, 14 CA, 30 CL, 61 SS
16 August 1942: 4 BC, 6 BB, 8 CV, 92 DD, 8 CA, 24 CA, 52 SS

Tagged around a bit and the carriers and battleships are definitely getting into fights - it's just that countries' heavy ships never fight each other. Every ship I saw sunk by a battleship or carrier has been heavy cruiser sized or smaller. That said, I didn't track down what happened to that one German battleship they lost.

It seems better with the mod on. I'll run the game to completion when I get time and see where it ends up. It's not a magic bullet that makes the AI play the naval game well, unfortunately, but I think only Paradox employees have the access required to do that.
 
  • 2Like
  • 2
Reactions:
I suggest you use the observer mode, there really is no need (or a difference) to 'playing' a minor. +your game won't get paused by events and bombarded by alerts (and it's less typing into console).

Then just CTRL+click on any country you want to observe.

To topic:
The game needs to run longer, 1942 is about just getting started. However I don't see much difference in the data. It still looks like just the screens and subs get massacred, while BBs either hide in a port or chicken out of every combat immediately. 2 of Germany's BBs are pre-dreadnoughts, a heavy cruiser could probably sink that. In 1.16 by this time the Italian navy was basically wiped, and japanese/german already showing visible losses.
 
  • 2
  • 1Like
Reactions:
Yeah, I verified that they are forming the strike forces. What actually happens once the combat starts is another matter altogether. At least actually creating the strike force is a start?
 
  • 3Like
  • 1
Reactions:
OK, getting the AI to form the strike forces is definitely only a partial solution. I just spent a chunk of my life dangling the German fleet in front of the British home fleet after confirming that the English AI had active patrols and a strike force ready to go.

Couldn't get a solid bite.

So...they managed to break the Naval AI in multiple new ways simultaneously. That is actually pretty impressive.

It occurs to me that port strike is probably massively powerful right now.
 
  • 4
  • 1Like
  • 1
Reactions:
vermicious knid said:
OK, getting the AI to form the strike forces is definitely only a partial solution. I just spent a chunk of my life dangling the German fleet in front of the British home fleet after confirming that the English AI had active patrols and a strike force ready to go.

Couldn't get a solid bite.

So...they managed to break the Naval AI in multiple new ways simultaneously. That is actually pretty impressive.

It occurs to me that port strike is probably massively powerful right now.
Click to expand...
So basically, for some reason, fleets are staying in ports after the new patch?
 
Kamis51 said:
So basically, for some reason, fleets are staying in ports after the new patch?
Click to expand...
Well, they are leaving to go on patrol and convoy raid. I think they are leaving to go try and engage as strike forces, at least briefly.

I wonder if the new ability to select which ship types a task force will engage/not engage is causing problems, but I don't see any quick way to check AI usage of that feature.

In a perfect world I would have loved them to roll out these changes to naval warfare one community beta at a time, so we could see that thing 1 was working before thing 2 was thrown into the mix. Reworking the esoteric details of naval warfare wasn't really the content they were using to market the DLC so I don't think it would have hurt their release buzz.
 
  • 2
  • 1Like
  • 1
Reactions:
You must log in or register to reply here.
Top Bottom
Back