Clown Car Thread

[neph]

All these philosophical comparisons to real life are absurd. It's a game; this is an optimization problem. Why would you compare the AI's struggles to historical approaches to ship design? The AI knows nothing of historical design and never will. Criticizing it will neither hurt its feeling or help the developers. It does not identify issues with the design model as they actually exist; it's just pretentious. Congrats, you cleverly & devastatingly insulted an algorithm. It'll never know or care. Unhelpful.

What's the actual issue? It's just an optimization problem. Right now, quantity of guns is prioritized over more practicable concerns. If the AI started off with "I need 3-5 primary turrets (of unified size)" and fooled around (probably using some kind of Monte Carlo method; I'd spring for basically simulated annealing) until it could fit the heaviest broadside possible (with armor defending against the same).

Then (only after main guns have been placed) would it decide upon what secondary armament it wants & how it's going to be placed, with the important restriction that it does not block main firing arcs. If you want to permit it some degree of "hey, can I slide around the main turrets & superstructure a bit" so you could, eg, fit a superduperfiring yamato-style secondary armament behind the main guns, that'd probably be near-trivial.

However, right now, it's all done at the same time, producing the smorgasbord you'd expect to see.

Templates dictating placement etc are far too heavy-handed of a solution. You don't need to do that--an automated solution is absolutely within grasp. That's not to say there isn't a role for some kind of template, however. What the AI will probably never be able to do without an MA model is look at the scenario & decide what type of ships it needs to bring.

Let me clarify: a BB is a BB, but there's a big difference between a BB designed to chew up escorts & one designed to go toe-to-toe with other BBs. So if there are templates, they might look like this:

• Template 1 (Cruiser-killer)
  • Optimize the main battery for time-averaged broadside @ 10 km (meaning the total weight thrown downrange over the course of a minute to 10 km range). This will produce solutions favoring rapidly-firing, shorter range, lower-caliber main batteries with a higher # of barrels.
  • Require some % of displacement remain after placing main battery & main belt+turret+deck armor (say 10% displacement)
  • Optimize the secondary battery for the same as the main battery. Since there's a healthy displacement margin, it ought to be able to plop in some 6/7/8" guns in there. Require that it provide an extended armoring against the same caliber as the secondaries.
  • If there's any margin left fill it up w/ tertiaries.
• Template 2 (Line-of-battle-ship):
  • Optimize the main battery for single-salvo broadside @ 18 km (adjust range requirement by era).
  • No % of displacement is required to be left over.
  • Fit secondaries in remaining space.

Of course, this is a very reductive model--there's still components & speed to consider. But frankly, the AI doesn't seem to have that hard of a time putting that stuff together--it's in battery design where its failings are so obvious. I guess call this a jab at the low-hanging fruit.

Edited October 4, 2021 by neph

[DougToss]

The problem is without templates the built ships won’t be “optimized”, because as you said - we can’t teach the computer naval architecture. That just means we fill in the blanks with a template, the computer can optimize and vary from that, but there needs to be a baseline or ships will continue to be wonky.

 

That you think this is insulting and pretentious reflects on you. 

[Stormnet]

27 minutes ago, neph said:

All these philosophical comparisons to real life are absurd. It's a game; this is an optimization problem. Why would you compare the AI's struggles to historical approaches to ship design? The AI knows nothing of historical design and never will. Criticizing it will neither hurt its feeling or help the developers. It does not identify issues with the design model as they actually exist; it's just pretentious. Congrats, you cleverly & devastatingly insulted an algorithm. It'll never know or care. Unhelpful.

The point of that was to show (some) people that the AI is not magically working now like some say. In its current state, its still not working. Its because that, despite several (and I mean several) attempts at balancing the AI, it is still broken as of now, that I leads me and some to conclude that likely the system is flawed at its core. You can patch a good if rough system. One that apears to never work no matter how much efford is put into seems like it simply is flawed to the core, and all the fixing in the world may make it work better, but never properly.

A lot of it can be optimized. The choosing of tech, systems, and callibers may be improved.
But it still requires templates like you say.

And even if the AI stops going with stupid decisions, without disposition templates, will this fix the big issues with turret disposition? Some of those flawed ships could work by moving the pieces. You proposed a doctrine type of template (which is still a template), that could fix the weird choice of callibers, armor, and speed (maybe even subsystems), but the AI may still mess stuff up with the right pieces and ideas by just putting the pieces in the wrong place. You tell the AI to use 8 15" turrets for this hull and doctrine, what will stop it front cramming those at the front in barbettes for no reason? IOr putting them in quad turrets at the back? Will this allow force it to place the stuff where it needs to be placed at? 

 

Edited October 4, 2021 by Stormnet

[Skeksis]

...

Edited November 9, 2021 by Skeksis

[RedParadize]

39 minutes ago, neph said:

All these philosophical comparisons to real life are absurd. It's a game; this is an optimization problem. Why would you compare the AI's struggles to historical approaches to ship design? The AI knows nothing of historical design and never will. Criticizing it will neither hurt its feeling or help the developers. It does not identify issues with the design model as they actually exist; it's just pretentious. Congrats, you cleverly & devastatingly insulted an algorithm. It'll never know or care. Unhelpful.

My background is the movie industry. For "historical" movie, the approach taken is to stay within the limit of "Suspension of disbelief" of the audience. Take Band of brothers for example, one of the best WW2 show ever made. They did take many liberty regarding characters, company, and event. Purist may complain about the historical inaccuracy of having the 506th liberating a concentration camp, but without that we would not have gotten to see one in the show. You can't really talk about WW2 without covering that subject. So, in a sense, Band of brothers is a condensed WW2 reality, greater in its depiction than reality itself.

It is important to note that even the most purist audience would not like the most accurate depiction of the life of a WW2 soldier... As it would be extremely boring.

This is how I think a "historic" game should be made. Now, if we go back to the AI design, I do not expect them to be historical at all, but they need to be believable and competitive vs player design. Atm they clearly are not.

[SonicB]

2 hours ago, RedParadize said:

This is how I think a "historic" game should be made. Now, if we go back to the AI design, I do not expect them to be historical at all, but they need to be believable and competitive vs player design. Atm they clearly are not.

All very fair points. In warship design, form generally follows function. Warships don't look a certain way for aesthetic reasons... (side-eyes interwar Italians and whoever designed the Type 21) but because physics and the technology of the period dictated that was the optimum shape for a fighting vessel with those specifications.

Therefore, if the game engine is working anywhere near as intended, an authentic-looking ship should also be a competitive, well-designed ship. I don't want actual historical designs, but I do want historically plausible ones.

[neph]

It's perfectly acceptable, if rather unkind to the devs, to beat the dead "ai bad" horse for pages and pages. I'm not saying it's wrong, I'm saying its unhelpful. We know the AI designs can be horrible. We're on page 8 of this thread. You don't have to spend a paragraph describing how horribly it's failing and how incompetent it would be if somehow it were hired by an interwar naval bureau. I mean, if it makes you feel good, I can't stop you, but I can say it's a touch condescending.

 

Posting examples of the designer's failing & identifying why they're broken--not in "what a failure of naval architecture, has the AI really never studied naval warfare before" terms but in "there's too much/not enough XYZ here; it has ABC without DEF etc etc" terms--that's (possibly) helpful.

Edited October 5, 2021 by neph

[DougToss]

41 minutes ago, neph said:

"there's too much/not enough XYZ here; it has ABC without DEF etc etc" terms--that's (possibly) helpful.

 

[AurumCorvus]

A brilliant design choice by the AI. And I do mean truly brilliant... at least from all the ammo detonations going off around my fleet!

(Also, take a quick look at the hit to damage ratio! 400 hits for 33k damage! And I'm not even done yet.)

 

Edited October 5, 2021 by AurumCorvus

[neph]

4 hours ago, AurumCorvus said:

A brilliant design choice by the AI. And I do mean truly brilliant... at least from all the ammo detonations going off around my fleet!

(Also, take a quick look at the hit to damage ratio! 400 hits for 33k damage! And I'm not even done yet.)

 

everybody knows you don't need armor if you've got 15" guns; you just line the shells along the hull--that's 15" of protection right there!

EDIT: and it's still only going 21 kt?!?!

Edited October 5, 2021 by neph

[SpardaSon21]

20 hours ago, Stormnet said:

This one is not a complete blunder, and is a better CA than the pseudo CL-CA above, but it still mixes secondary callibers for no reason

 

That's because you can't put 5" guns in those top casemates, IIRC.  Only 4" guns, and so the AI maximizes its firepower.  Not like there's no reason to do that since there's no such thing as splash confusion, unlike IRL.

20 hours ago, Stormnet said:

This one has a pretty weak firepower of 4x 2 12"s, and mixes callibers, but its not that stupid at least...

 

Errmm....  Granted its far, far heavier than the South Carolinas themselves were, but its also far faster and has better armament.  Keep in mind that in 1911 only the USN had switched over to an all-centerline, superfiring armament for their BB's. The German Helgoland had four wing turrets (two per side) and only two centerlines, and even the British St. Vincent class had a non-superfiring X and a set of wing turrets, no B mount.  It wasn't until HMS Neptune that the Brits finally figured out how to do a superfiring X, but even then it still had wing mounts, with just three centerline guns.  Of course, the USA had been rocking a five-turret centerline ever since the Delawares that had been built concurrently with the South Carolinas, and again with the succeeding Florida-class battleships.  It wasn't until the Orions that commissioned starting in 1912 that the Brits finally went all-centerline, and 1914 for the Germans with the Koenig-class.  Of course, by that point the Wyoming-class Standards and their six centerline turrets were commissioning so... oops, try again next time, Europe. IMO only the USA should get to put barbettes on Dreadnought 1 hulls for that very reason, and all other nations should be restricted to just three centerline turrets for Dreadnought1 and 2 hulls.

Edited October 5, 2021 by SpardaSon21

[DougToss]

RTW follows that development, with wings, centreline non super firing, en echelon with firing across the deck, aft super firing, bow super firing.

I think it varies by nation. As the Royal Navy I think I nearly always have the ability to have a super firing XY pair before the AB pair.

[SonicB]

11 hours ago, neph said:

It's perfectly acceptable, if rather unkind to the devs, to beat the dead "ai bad" horse for pages and pages. I'm not saying it's wrong, I'm saying its unhelpful. We know the AI designs can be horrible. We're on page 8 of this thread. You don't have to spend a paragraph describing how horribly it's failing and how incompetent it would be if somehow it were hired by an interwar naval bureau. I mean, if it makes you feel good, I can't stop you, but I can say it's a touch condescending.

 

Posting examples of the designer's failing & identifying why they're broken--not in "what a failure of naval architecture, has the AI really never studied naval warfare before" terms but in "there's too much/not enough XYZ here; it has ABC without DEF etc etc" terms--that's (possibly) helpful.

Might I remind you this originally started as a joke thread to entertain people in the long wait between Alphas? Labelling it as 'condescending' is in itself pretty condescending.

Besides, if a design looks bad, you can pretty much guarantee it is bad. I'm not going to bother repeating my previous post.

[SonicB]

1 hour ago, DougToss said:

RTW follows that development, with wings, centreline non super firing, en echelon with firing across the deck, aft super firing, bow super firing.

I think it varies by nation. As the Royal Navy I think I nearly always have the ability to have a super firing XY pair before the AB pair.

If I recall, RN superfiring turrets right up to the Iron Duke class couldn't actually fire down the axis because of Fisher's insistence on top-mounted sighting hoods. I'm curious - did this get modelled in RTW?

[DougToss]

It’s a technology, and I think it happens later for the UK player than most, based on a modifier, but I don’t think it is specifically stated to be due to sight hoods. 

[Arrigo74]

Quote

(side-eyes interwar Italians and whoever designed the Type 21)

I think, but not 100% sure that the latter (I assume you are talking of the Type 21 RN frigate), involved the dad of one of my classmates in PhD school... 😅

 

Said that, I disagree with Skesis about the AI being rough but good... I still have to see a competent design after the pre-dreadnought period. Templates are becoming a necessity (and to be quite honest they are also a logical step, after all templates is how naval architecture works in broad sense). 

 

As far historical movies are concerned...

19 hours ago, RedParadize said:

I do not expect them to be historical at all, but they need to be believable and competitive vs player design. Atm they clearly are not.

Both agree and disagree. I agree with the latter statement, disagree with the former. I do not want carbon copies of all historical ships (but that will not hurt either) but I want design that follow roughly the same principle rather than Frankenstein monsters... There are reasons behind every real design, not least the fact that the design had to work and be 'competitive'.  Suspension of disbelief is not the only element in an historical game design (and no, Hollywood should not be our benchmark...)

[Speglord]

Doing "Defend Your Convoy" and came across this monstrosity...

[AurumCorvus]

34 minutes ago, Speglord said:

Doing "Defend Your Convoy" and came across this monstrosity...

Love the double barbette right in the middle, as well as the two 6/7in guns by the X turret, probably unable to track a full arc. Also, all that wasted space...

[RedParadize]

That double barbette, by sheer luck, is back to back. One of the few instance where that barbette do not look out of place. Its heavier than the secondary tower trough!

[arkhangelsk]

Game, I don't think you can ever go wrong if

a) You just lock the bulkheads to maximum.

b) The fore / aft guns are always at least as big as the side guns.

c) This is a Custom Battle. We have infinite money, so start with Quality 100 crews.

BTW, I'll believe a 44803HP engine can push a 20000 ton thing at 28 knots when I see it.

[neph]

The Secondary Turret of Damocles

[AurumCorvus]

5 minutes ago, neph said:

The Secondary Turret of Damocles

It makes the firing of an incompetent Captain/Admiral much more... literal.

[Steeltrap]

^

Looks like a bad case of warts.

[Steeltrap]

On 10/6/2021 at 4:59 AM, Speglord said:

Doing "Defend Your Convoy" and came across this monstrosity...

My mind immediately went to....

https://www.youtube.com/watch?v=aVA-cqASg8k

 

[SpardaSon21]

What the hell sort of 1930 design is this?12" guns, 8" secondary turrets, 7" casemates, a lone 4" turret in the middle on a barbette, and 2" guns shoved in tight as can be next to the barbettes.