Stormnet Posted September 25, 2021 Share Posted September 25, 2021 (edited) #stop_turret_rape Look at those stats, and tell me if they are acceptable for a 66k ton ship in 1940. Edited September 25, 2021 by Stormnet Link to comment Share on other sites More sharing options...
Stormnet Posted September 26, 2021 Share Posted September 26, 2021 (edited) Going back to the barbettes, why the F*** is this still an issue? (image by RedWing) Is it that hard to make a "tag" for "accepted callibers" so the AI doesnt F this up? I get the AI making glass cannons. I get them messing the designs on the "Idea" level. But I dont get them pulling this kind of stuff off because it simply is not possible for us. One of the most common feedback of this game is the AI. It makes impossible OP ships, it makes brain-dead ships, and it makes ships that overlap somewhat. And the most common solution is to make a list of presets that the AI can partially modify, so they look diferent, but the f*** up rate is drastically reduced. Yes, the AI has gotten much better. But it is still flawed and doing stupid s***. The very good ships it ocasionally makes are likely derived from sheer luck (it just happened to pick values that actually make sense). At this point I suspect it is just a Sunken Cost Falacy. We, the players, are more than willing to supply working designs for the devs to use. Give up wasting time patching this AI that will never properly work. Edited September 26, 2021 by Stormnet 3 2 Link to comment Share on other sites More sharing options...
Arrigo74 Posted September 26, 2021 Share Posted September 26, 2021 Well, no screen here... but my AI had built a dreadnought with at least 5 different calibers... (Naval academy Power of the Dreadnought). I distinctly remember Andrew Lambert, Norman Friedman, Howard Fuller and myself talking about dreadnought having basically a single caliber main gun... also twin centerline turrets and triple side ones... In y dream land the AI would pick from historical designs rather than attempt its worst in post-modern warship design... That would solve a lot of issue on Naval Academy, possibly even in custom battle. Wondering about the campaign... but again using historical ships as templates... 1 Link to comment Share on other sites More sharing options...
arkhangelsk Posted September 26, 2021 Share Posted September 26, 2021 On 6/27/2021 at 8:12 PM, Stormnet said: Theres a secondary on a tall barbette, that blocks a superfiring main from firing front, defeating the while point of superfiring those turrets. They put the mains on barbettes so they could fire throught the secondaries gore in the deck. Its a secondary calliber collector ship. Who knows about that weight balance. Chogei, what a name. In the plus side, it sure packs quite a punch with those 9 20" for a ship that size Oh, this is an old design, but I actually find it interesting - not that I think a real navy will end up choosing this one, but it does have its merits. Little guns that almost don't weigh anything aside, its batteries are centerline 20" and 8" guns. A conventional design will put the secondaries to both sides, but then half of them would be unusable in the Main Combat Mode of a battleship (broadside firing). Centerline placement allows a heavier secondary battery to be deployed against enemies while in combat. The only shame here is that it is two twin 8-inch turrets - if they can buff it to two triple 8-inch guns, the number of 8-inch guns will increase to six which would be much better in terms of fire control. The superfiring value is reduced due to the 8-inch B-turret, but it is not completely gone. If C turret is on the deck, it would be completely unable to fire over the 8" turret. In its current form, it's conceivable that at maximum superelevation it would just be able to do so. I don't think I want to be in B turret when that happens due to blast but at least C-turret won't shoot through B-turret. So the emphasis is on Broadside firing. If forced to chase a target this design can engage it with 6 20 inch and 4 8-inch guns by using only slight weave. In a desperate pinch where the target must be slowed no matter what and has gone beyond the maximum range of the 8 inch, the forward 20-inch batteries can be employed without completely destroying the 8-inch turret. And 長鯨 or 超鯨 (long whale or super whale) aren't the most hopeless names ever made. Link to comment Share on other sites More sharing options...
Skeksis Posted September 27, 2021 Share Posted September 27, 2021 (edited) ... Edited November 9, 2021 by Skeksis 1 Link to comment Share on other sites More sharing options...
SonicB Posted September 27, 2021 Author Share Posted September 27, 2021 Near Jutland mission, two restarts because of crappy AI design. https://imgur.com/g3QMAdWhttps://imgur.com/g3QMAdW tl;dr: 4 Link to comment Share on other sites More sharing options...
SodaBit Posted September 27, 2021 Share Posted September 27, 2021 This one's old, but the AI really needs to learn how to prioritize larger secondary guns. Or at least de-prioritize peashooter calibers... Link to comment Share on other sites More sharing options...
Arrigo74 Posted September 27, 2021 Share Posted September 27, 2021 Here are my last adversaries in Power of the Dreadnought... the secondary (shielded) turret on a barbette is a must... as the line of secondaries on the side... basically can the AI design a dreadnought? Link to comment Share on other sites More sharing options...
Stormnet Posted September 30, 2021 Share Posted September 30, 2021 On 9/27/2021 at 5:20 PM, Arrigo74 said: Here are my last adversaries in Power of the Dreadnought... the secondary (shielded) turret on a barbette is a must... as the line of secondaries on the side... basically can the AI design a dreadnought? It can design good ships, but it does have a very bad f***-up rate of shitty designs. 1 Link to comment Share on other sites More sharing options...
RedParadize Posted September 30, 2021 Share Posted September 30, 2021 (edited) AI version of a "All or Nothing" armour scheme... mostly on the nothing side... in a game where AoN do not work. I think it would have been better with mega barbette for its 13" guns. Its all good trough, it has 13.9" of armour on the 8" casemates guns! Edited September 30, 2021 by RedParadize 1 Link to comment Share on other sites More sharing options...
RedParadize Posted October 1, 2021 Share Posted October 1, 2021 Doesn't clearly belong in the clown car tread, as there is merit to paper tiger design, I just trough it was worth mentioning: guns are 15"er. 2 Link to comment Share on other sites More sharing options...
Stormnet Posted October 2, 2021 Share Posted October 2, 2021 13 hours ago, RedParadize said: Doesn't clearly belong in the clown car tread, as there is merit to paper tiger design, I just trough it was worth mentioning: guns are 15"er. It does belong here. The nonsensical barbette placement adding weight and blocking turrets, the sluggish speed. 1 Link to comment Share on other sites More sharing options...
DougToss Posted October 2, 2021 Share Posted October 2, 2021 It was painful to read through this thread. There are more than enough people here who have read up on Naval Architecture, or failing that, can read and understand ship plans, to make solid and historically informed templates for the AI. RTW does a great job scaling designs by AI nation, and technology. We could very easily do the same here, and the game would be much better for it. 3 Link to comment Share on other sites More sharing options...
SonicB Posted October 2, 2021 Author Share Posted October 2, 2021 On 9/27/2021 at 2:22 PM, SodaBit said: This one's old, but the AI really needs to learn how to prioritize larger secondary guns. Or at least de-prioritize peashooter calibers... Thanks, but just a reminder that Nick threatened to shut this thread down if older versions were posted. This is a thread for the lols but at the same time it does hopefully help the devs, so let's stick to clowncars from the current version please. 1 Link to comment Share on other sites More sharing options...
neph Posted October 2, 2021 Share Posted October 2, 2021 9 hours ago, Stormnet said: It does belong here. The nonsensical barbette placement adding weight and blocking turrets, the sluggish speed. Not sure that that's "nonsensical barbette placement", the C barbette is built-in to the hull. Also, an A/B/C/D/E turret configuration... thinking about the nomenclature cracked me up a few times writing this haha Link to comment Share on other sites More sharing options...
Stormnet Posted October 3, 2021 Share Posted October 3, 2021 (edited) 19 hours ago, neph said: Not sure that that's "nonsensical barbette placement", the C barbette is built-in to the hull. Also, an A/B/C/D/E turret configuration... thinking about the nomenclature cracked me up a few times writing this haha You are right, looking more closely I see a barbette, so it may not be as braindead as I thought. But it still seems weird to have a massive barbette at the B turret for no reason. What is its porpose? To fire forwards? Then remove that and let the C turret do it, as that barbette blocks C from doing so. Same firepower for less weight. Is it to fire be able to aim further back and bypass the superstructure? But then, with that superimposed C turret, the benefit is neglectible. I think I know why the AI actually did this. A third look reveals a few secondary guns next to the B turret. The AI often priorises deck placement of secs over proper weight and barbette management. If you check my retarded meme-list of AI fails, number 16 "Big turrets are afraid of piranha secs roaming the decks, so they hide on top of barbettes.". This is one of the flaws of the AI. Smaller mistakes that look inocent at first but proove idiotic upon a second thought. Now, if a player wants to do this, then he should be allowed. We actually (or at least usually) know what we are doing, and we may have an idea in my or be just messing around. The porpose of the AI is to be a challenge, not to mess around. Thinking twice, I look a bit weird responding in this big passive-agressive wall of text... heh. Edited October 3, 2021 by Stormnet 1 Link to comment Share on other sites More sharing options...
DougToss Posted October 3, 2021 Share Posted October 3, 2021 Everything you said makes sense to me. I think you identified the challenge of teaching the AI naval architecture - a lot of this stuff is not intuitive. Link to comment Share on other sites More sharing options...
neph Posted October 4, 2021 Share Posted October 4, 2021 9 hours ago, Stormnet said: You are right, looking more closely I see a barbette, so it may not be as braindead as I thought. But it still seems weird to have a massive barbette at the B turret for no reason. What is its porpose? To fire forwards? Then remove that and let the C turret do it, as that barbette blocks C from doing so. Same firepower for less weight. Is it to fire be able to aim further back and bypass the superstructure? But then, with that superimposed C turret, the benefit is neglectible. I think I know why the AI actually did this. A third look reveals a few secondary guns next to the B turret. The AI often priorises deck placement of secs over proper weight and barbette management. If you check my retarded meme-list of AI fails, number 16 "Big turrets are afraid of piranha secs roaming the decks, so they hide on top of barbettes.". This is one of the flaws of the AI. Smaller mistakes that look inocent at first but proove idiotic upon a second thought. Now, if a player wants to do this, then he should be allowed. We actually (or at least usually) know what we are doing, and we may have an idea in my or be just messing around. The porpose of the AI is to be a challenge, not to mess around. Thinking twice, I look a bit weird responding in this big passive-agressive wall of text... heh. I think fundamentally it's that the AI prioritizes # of guns over anything, particularly # of big guns. Simplistic but making AI isn't trivial Link to comment Share on other sites More sharing options...
DougToss Posted October 4, 2021 Share Posted October 4, 2021 1 hour ago, neph said: I think fundamentally it's that the AI prioritizes # of guns over anything, particularly # of big guns. Simplistic but making AI isn't trivial But making templates is, or at least is substantially easier. It would take care of those problems, because a template would set the appropriate number and calibre of guns for a given hull/year/nation, and that could be made into a dynamic range a la RTW2. Link to comment Share on other sites More sharing options...
Skeksis Posted October 4, 2021 Share Posted October 4, 2021 (edited) ... Edited November 9, 2021 by Skeksis 1 Link to comment Share on other sites More sharing options...
SonicB Posted October 4, 2021 Author Share Posted October 4, 2021 7 hours ago, Skeksis said: The fact is, from a player opinion, generally, the AI is producing great designs, not all the time but it is a WIP. We don't actually need templates. From this player's opinion, I disagree (and I've played through all of the Academy missions since the core patch wiped their progress.) Roughly 60% of the time, every AI design in a scenario is acceptable, albeit flawed. 30% of the time, I have to restart the mission because either an allied or enemy AI design is bad enough to unbalance the scenario, or just aesthetically awful enough to break immersion (this usually follows a materially bad design, because poor placement and bad turret choices look bad and work terribly.) Only 10% of AI designs are what I would consider completely realistic in terms of performance and aesthetics. Link to comment Share on other sites More sharing options...
Stormnet Posted October 4, 2021 Share Posted October 4, 2021 (edited) 8 hours ago, Skeksis said: But adaptive AI can/is/will be doing the same thing. People in general, well free will internet people, will complain first because... well that's what internet people do, well kinda IRL too, there's little comeback on internet too. When you said, and I quote, "It was painful to read through this thread", you must have thought these were all the facts, but they are not, here in this thread they are the just the complaints and/or alpha testing reports, just a cross-section of AI designs. It's not the true representation of what the AI is actually accomplishing. Nobody posts good AI designs, well not generally anyway. So by not playing the game you're not getting the full representation of what the AI is doing. Though I can understand your conclusion based only on these reports. I think you don't have the facts of good AI designs so far. The fact is, from a player opinion, generally, the AI is producing great designs, not all the time but it is a WIP. We don't actually need templates. PS, though it would be nice to have some 'named' ships e.g. Bismarck etc. from time to time. Adaptive AI, great designs? All claims that I decided to test, not with arguing, but with experimenting. 4 consecutive battles featuring most classes on each side, all with AI generated designs. So lets jump into it. 1911 Austria-Hungary VS United States So, first design ever with Auto-Generate going into the battle and its a blunder. 1 Mixing lots of callibers 2 Non-uniform main-gun armaments. This is not 1905. We are well in the dreadnought era. 3 Giving the only 20 cm turret (only one of that calliber) a large barbette it doesnt need. You could use something smaller 4 The wing turrets cannot fire backwards, they are blocked by secondaries. 5 That sec behind the 20 cm has a barbette... for reasons (I suspect it is to fire over the single 5.1s next to it... great fix I know) I clicked the button another 3/4 times before having to settle on this Again, similar blunders 1 Mixing lots of callibers 2 Non-uniform main-gun armaments. 3 Giving the that 15 cm turret an excessively big barbette. But ofc, the AI also likes to inovate in failing 4 Those side 15 cm could be pulled back to allow the forward turret to fire back wards, but nope. 5 There is a funnel right behing the secondary tower... wanna smoke sausages? But lets sail gloriously into battle... S***. Alright, this one may be somewhat acceptable. That B turret probably doesnt need a barbette that big and tall, and you likely dont need those 4 single 7.6cm, but its ok-ish as a destroyer. Yet again, the AI fails in the classic ways while also innovating in how to middle-finger naval design. 1 Excessively large barbettes for the B and "C" (not a main turret, but another lone high-calliber secondary). 2 Single side turrets... why? 3 The funnel, that could be perfectly on the superstructure, is outside, because reasons. 4 Another "Collect the callibers" ship. At least it is very fast... Not that its gonna matter with the pathetic firepower of 9"s mains. Why is this CA exactly like a pre-dreadnought? Heh, doesnt matter for the design. This thing is a Light-Heavy-Cruiser. Its very light, and it only has a pair of double 15.2, 4 casemated 7.6, and 2 casemated 5.1 as it's firepower. Enought for a 1911 CA? Doubt it. That ship must be sinking anyways, the water level is that tall. Im pretty sure that is a early 1900s CA hull, but anyways, there is some mixing of sec calliber, and those single 17.8 may be slightly overkill in calliber but not much in rate-of-fire/accuracy ratio. Also, I just noticed that the Light Cruiser is bigger in tonnage than the CA above... LOL. 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 Not a complete blunder (it actually picks the right sized barbettes, HOORAY), and it does have a very good speed for the time while still keeping some acceptable armor (still would fit better being a BC), but still some doubtful design decisions, including: 1 Why is there a barbette at the back (spoiler, to fire above the secs). 2 Some mixing of calllibers. 3 It has a ABC-Z (X? I never know) layout, but it doesnt pull the superstructure back to lay the turrets. This leads to me to question the weight balance on them. One must suspect is a large front weight offset. This one has a pretty weak firepower of 4x 2 12"s, and mixes callibers, but its not that stupid at least... 1931/1932 British Empire VS Russian Empire This one is pretty sluggist. I think 1930s ships could be faster than 22.6 knots. The main firepower is very good for a ship its size, and it has thick armor too, but it only has 1 tripple 15.2 and 4 7.6 cm as seccondary batteries. It likely sacrifices on speed to get this armor and firepower. I has a mediocre speed and secondary armaments, but I think it earns the "Good design" award based on the firepower and protection. Lots and lots of single, double, and tripple 5.1s and 7.6s. Coupled with torp launchers blocking the rear turret. Again, not a complete failure, but could use several layout optimizations. Also the name "Porpoise". Say that to a friend and see if you 2 can keep a straight face. This DD aint getting the benefit of the doubt. The confusing and senseless use of barbettes at the front, the 5.1s being scattered around the ship. At least it has good torp armaments. This CA has a weaker firepower than Porpoise, with secs and torps blocking the front turrets, turrets in weird layouts. Its just... confusing... At least it beats Porpoise in speed... wait, isn't it supposed to be the CL beating the CA in speed and vice versa for firepower? Going back to the classics, the AI forgets correctly-sized barbettes exists, that you need more secondary armaments than a single double 20 cm and several 5.1s and 7.6s scattered across the deck blocking the main turrets fire arc, and again, funnel behind sec tower. I forgot to take another pic of this one when my ships identified it. While I cant see its stats nor its overall layout (and thus cant say if it is completely bonkers), you can see those secs are located on the very edge of the deck. Their ring must be drilled on the belt armor. AGAIN, more mixing of callibers, using a large high calliber secondary alone (though this time they actually bothered to add more secs than that), and mediocre firepower (15 13" cannons do not make up the armament of a good 60.000 ton ship in 1932). First of all, this is clearly a recycled destroyer hull. Second, this thing has several (and I mean several) 12 cm double turrets because... bigger tripples bad? Third, that secondary tower is being harrased by both a non-existant rear view and smoke. This one gets a pass. While the layout is a bit weird, I dont see much pea-brain logic here. 1940 Germany VS Italy (I was getting a bit tired at this point so I shortened the battles) We start off by rather weak main batteriy for something this size, even more mixing of callibers and more single turret batteries. At least it has good armor. This one is alright. Some mixing of sec callibers, but picked appropriate secondary armaments, and has a decent main battery. Im getting real tired now, so lets wrap this up with one final battle. 1940 Japan VS France Back in force,the AI reveals to us nonsensical use of barbettes (you dont need a barbette that tall buddy [and that thing doesnt look wide enought anyways] and you wouldnt need a barbette at the back if you didnt cram similar calliber secondaries there), more weak firepower, and mixing callibers... AGAIN. I have a few questions. 1 Why is your 14k ton CA almost as strong as your 50k ton battlecruiser in terms of armor (and a bit of firepower). And why is it slower than an average battleship? And if you like calliber mixing so much, why also mix the barrel amount of the 5.1s? If these questions dont matter, then I think this ship is alright too. Sigh, it doesnt end, does it? MORE calliber mixing. MORE funnels harrasing rear towers. MORE anorexic barbettes. Mix that calliber, Put a barbetted sec (belonging to a 1 turret battery) turret in front of the rear X gun, And more mediocre firepower. It is fast and has a reasonable armount of armor... at least. But its sinking pretty much. And that marks the end of my experiment. The results? 1 or 2 actually good ships A couple of alright ships Several flawed ones A bunch of complete garbage designs. Does this sound like great designs and great "adaptive" AI? Dont think so. Say it's great now. Edited October 4, 2021 by Stormnet 3 Link to comment Share on other sites More sharing options...
DougToss Posted October 4, 2021 Share Posted October 4, 2021 10 minutes ago, Stormnet said: Does this sound like great designs and great "adaptive" AI? Dont think so. I’m confused about that too. I’m struggling to find by what metric those are sound ships. At the most basic principles, there’s no consistent naval architecture present in the design. Even the most simple things like not mixing calibre are absent. These are Vasa and Mary Rose levels of ineptness, in the period immediately after (1850-1880) scientific naval architecture was beginning to be established. These designs are less excusable than the many Victorian ships that capsized or floundered, because the key lessons, if not tested in battle, then at least in seaworthiness, had been learned by 1900. 1 Link to comment Share on other sites More sharing options...
Cpt.Hissy Posted October 4, 2021 Share Posted October 4, 2021 it doesn't end, heh. Don't you think that big part of AI being so helpless still lies in the flawed ship constructor design? In the use of huge predefined parts, namely. Many of those stupid situations happen because it's just impossible to arrange those parts in any other configuration because they don't fit. 3 Link to comment Share on other sites More sharing options...
Skeksis Posted October 4, 2021 Share Posted October 4, 2021 (edited) ... Edited November 9, 2021 by Skeksis 1 Link to comment Share on other sites More sharing options...
Recommended Posts