RNG is broken

[admiralsnackbar]

RNG in Hit probabilities:

1) For testing purposes I wonder if it would be possible to export the data from battles where you could see, at least for main battery salvos, the probability of hit for each salvo and the number of hits generated. You could then compare the number of hits scored over time to the 'Expected value' generated from the sum of the probabilities in excel. 

2) There may be a human bias in looking selectively at hit chances for individual salvos and concluding that the game is biased against the player, given how probability in salvos changes drastically for each salvo, #1 would help determine whether the suspicion is warranted or not.

3) Not sure to what extent the AI constantly maneuvering renders the hit probabilities less . I've seen BB shells obliterate nearby cruisers because they were inside the cone of dispersion and got unlucky. (I'm not for or against this per se) 

RNG in Ship Design

In some of the Academy missions I think random generation goes so far as to take the skill of the player out of determining the battle. Compounding the frustration is that the very process of randomizing the fleet compositions and placements adds to the loading times which become a fixture of resetting matches that have unfavorable designs or just unlucky first salvos for the player. 

I actually am more bothered by the randomness of the allied ships you are given than the randomness of the enemy design. Whether the destroyers you spawn with have 9km range torps or 20km range torps is hugely influential in what strategies you can employ (As it was in real life). Sometimes you spawn with screens that can't keep up with your capital ships. 

The sheer number of moving parts for any given academy mission can poison people's perception of whether a mission is too hard or too easy due to odd ship design or due to imbalanced mechanics. 



 

[Stormnet]

This might be usefull in patching the accuracy RNG.

 

Maybe one way to at least partially fix RNG accuracy would be to add minimum and maximum hit thresholds based on the hit chance. In essence, these mechanisms would analyse the fighting based on the chance % and actual hits. In other words a component specially designed to intervenenin these situations.

 

If one ship keeps missing (10% constant chance of hitting, but it has fired 30 times without a single hit), or a ship keeps hitting (5% chance, but hits every 3 salvos) then the mechanism comes into action. It would initially increase/lower the hit chance in the background (tell the player his stupidly innacurate guns have 10% of hit, but secretly raise the odds to 20%). If the odds are still messed up, keep secretly raising/lowering the odds until the hits begin to match the "advertised" percentage, and then slowly reverse the changes (while constantly checking the results are stable and not getting messed up again).

 

In case the RNG is still screwing stuff up, activate 

EMERGENCY OVERRIDE PROCEDURE!!!

Which is a excessively cool name for a process which consists of lying to the player about the percentage odds, but ditching RGN all togheder in the background and making a predetermined hit patern. If the chances are 15 percent, then create a shot patern for the next 20 shots, and spread 3 succefull hits (more if the ship was constantly missing, less if it was constantly hitting to compensate the previous results and to aliviate the damage the RGN had caused) randomly throught that patern. Do this for a bit and then gradually reverse the situation.

 

The main drawbacks of this system, is that im not sure how you could code it to the dynamic odds (but I think the devs can find a way to) and a feel of rubberbanding artificial dificulty for some people (but we aint getting simulated physics anytime soon, so do you prefer these hitting/completely missing against the odds leading to these unfair defeats or undeserved victories?)

 

 

This was my crack at this problem (aka, preventing these situations). What do you think about it?

Edited April 8, 2021 by Stormnet

[disc]

Anecdotal evidence is pretty bad when examining statistical phenomena. Leads to extrapolating to false extents. 

To make conclusions about RNG, then there must be either data-mined source information or solid tabulated results (gathered under known conditions) to make anything beyond an educated guess.

This is not to say that the game's RNG is not flawed. There could easily be a horrible algorithm or a huge bug making issues. It's just very hard to conclude that right now with much confidence.

[Skeksis]

...

Edited November 9, 2021 by Skeksis

[akd]

Maybe there is nothing to fix here? Just usual human inability to comprehend probability in a rational way, especially gamers who aren’t wargamers.

[brucesim2003]

54 minutes ago, akd said:

Maybe there is nothing to fix here? Just usual human inability to comprehend probability in a rational way, especially gamers who aren’t wargamers.

Oh sure, the game is perfect. /sarc

So says a wargamer for over 30 years.

[madham82]

1 hour ago, akd said:

Maybe there is nothing to fix here? Just usual human inability to comprehend probability in a rational way, especially gamers who aren’t wargamers.

More likely the percentages being shown in game do not match up to what the game is actually doing. I can't count the number of times my salvo accuracy number is 100%. That should really be an impossibility. 

Side note, I've mentioned it before. Why does the game have multiple accuracy numbers in different sections of the UI? Why do you have to hold your mouse over a target from the perspective of your ship (not just making sure your ship selected) to get the detailed accuracy breakdown? How exactly does that work when you can't actually "see" the actual enemy icon from your ship's perspective? IMO the UI's clunkiness leads to many people misinterpreting the data actually available. 

How much of that is contributing to threads like this? Who knows. 

[Cpt.Hissy]

7 minutes ago, madham82 said:

Why do you have to hold your mouse over a target from the perspective of your ship (not just making sure your ship selected) to get the detailed accuracy breakdown? How exactly does that work when you can't actually "see" the actual enemy icon from your ship's perspective?

A hint here.
If you focus the camera on the enemy and then move the camera (by WASD or mouse), this unlocks it from enemy and makes this info available from camera's position. Select your ship (but don't move camera over there) and point at enemy. May want to tilt the cam upwards, as info panel appearing above the enemy will be huge.

I do agree it's bad UI design anyway

[madham82]

Just now, Cpt.Hissy said:

A hint here.
If you focus the camera on the enemy and then move the camera (by WASD or mouse), this unlocks it from enemy and makes this info available from camera's position. Select your ship (but don't move camera over there) and point at enemy. May want to tilt the cam upwards, as info panel appearing above the enemy will be huge.

I do agree it's bad UI design anyway

Good tip I will use in the meantime. Long range engagements are a real pain otherwise. 

[akd]

4 hours ago, brucesim2003 said:

Oh sure, the game is perfect. /sarc

So says a wargamer for over 30 years.

The game has huge issues, but I have serious doubts that not messing with probability to make the player feel better is one of them.  As Madham notes, the issue is more likely communicating info better and setting reasonable expectations for a game that covers a huge time period during which there were monumental changes in gunnery.

Edited April 8, 2021 by akd

[disc]

The "100%" accuracy I suspect is rounding from values exceeding 99.9%. I'm not certain how individual gun accuracy is modeled anymore, as they seemed to make big changes a few months ago (for example, shells can now hit nearby ships they weren't shot at). I'm not sure if shells curve in flight these days.

I doubt that there is an intentional obfuscation of the numbers -- though of course I don't know that. There certainly might be bugs, though. I reckon the trouble may be the interface.

[Stormnet]

Im seing some UAD top players brawling on who's right on what, so I'm likely about to just walk into a crossfire and draw a target on myself.

I've seen fair points that the rather isolated situations and reported nature of this evidence and near complete lack of insight on how these systems really work in the background makes this an unconfirmed problem.

Maybe the RNG isn't broken by itself. The very nature of unchecked/uncontrolled RNG makes it so situations like almost never hitting/missing even with 70% chance, very unlikely, but not completely impossible. I don't have a single idea on many players UA:D has, but (pure speculation) assuming a thousand play every week, 2/3/4 small battles or 1 big one on average per session, then situations like these that were almost impossible for a single player to get become likely for at least a few players to get them once.

So, while surely unconfirmed, if the RNG really has no checks or safeguards, then these situations might be possible, although if so they apear to rare and neither a widespread nor common issue (the forum would be filled with complaints if so). Not frequent enought for it to be broken, yet not impossible for it to happen.

However, if this problem was ever confirmed, a potential solution for rare events like these could be adding checks (like those I describe above) that detect and intervene in these ocorrences.

Or maybe Im wrong and too inexperienced to know this stuff. Idk.

Edited April 9, 2021 by Stormnet

[MrStan53]

On 4/7/2021 at 1:40 PM, Stormnet said:

This might be usefull in patching the accuracy RNG.

 

Maybe one way to at least partially fix RNG accuracy would be to add minimum and maximum hit thresholds based on the hit chance. In essence, these mechanisms would analyse the fighting based on the chance % and actual hits. In other words a component specially designed to intervenenin these situations.

 

If one ship keeps missing (10% constant chance of hitting, but it has fired 30 times without a single hit), or a ship keeps hitting (5% chance, but hits every 3 salvos) then the mechanism comes into action. It would initially increase/lower the hit chance in the background (tell the player his stupidly innacurate guns have 10% of hit, but secretly raise the odds to 20%). If the odds are still messed up, keep secretly raising/lowering the odds until the hits begin to match the "advertised" percentage, and then slowly reverse the changes (while constantly checking the results are stable and not getting messed up again).

 

In case the RNG is still screwing stuff up, activate 

EMERGENCY OVERRIDE PROCEDURE!!!

Which is a excessively cool name for a process which consists of lying to the player about the percentage odds, but ditching RGN all togheder in the background and making a predetermined hit patern. If the chances are 15 percent, then create a shot patern for the next 20 shots, and spread 3 succefull hits (more if the ship was constantly missing, less if it was constantly hitting to compensate the previous results and to aliviate the damage the RGN had caused) randomly throught that patern. Do this for a bit and then gradually reverse the situation.

 

The main drawbacks of this system, is that im not sure how you could code it to the dynamic odds (but I think the devs can find a way to) and a feel of rubberbanding artificial dificulty for some people (but we aint getting simulated physics anytime soon, so do you prefer these hitting/completely missing against the odds leading to these unfair defeats or undeserved victories?)

 

 

This was my crack at this problem (aka, preventing these situations). What do you think about it?

         Personally, as someone ironically actually coming on here to complain about the failings of RNGesus and to shriek XCOMmery, I think personally this is a bit too complicated of a solution. I'm actually gonna be the first to say RNG definitely needs to still exist (in fact I'm one of the few people who plays TF2, as a backburner pyro nonetheless, and wants random crits and spread). With this solution, it doesn't necessarily make the game fair to the AI either, which could eventually lead to player boredom or disinterest. It's a bit overdesigned as well and as the game can already run slowly at times it'd result in even laggier times. The issue, I think, is that RNG appears to be incredibly flat and uses only one roll to determine. I can't verify this for sure but playing the game it feels like hits and damage percentages vary far too much, and combining this with what feels to be an unfair advantage towards the AI in many cases. 

         To transition to a totally different game entirely, let's turn to the classic Dungeons and Dragons from the Satanic Panic era and before. Oftentimes with many dicerolls during the game (most notably during your character creation) many that could be done with D12s or D20s were actually substituted with 2D6 and 3D6 rolls. Though this was also to make picking up the game easier, it added a unique function. Most rolls would result in a bell curve formation, as a single roll of a 1 could be made averaged out by two 6 rolls. RNG was a factor for sure of course, however it was far less likely to do spectacularly or abysmally, making it easier to have more balanced characters and actions. For example, my most recent quest as a thieving halfling resulted in a very averaged character that was more resilient to sub par rolls (granted with physical 6 sided dice you can't have 0s, but as this is programming it's different).

         My proposal for a change would be first and foremost to make the RNG roll the dice more times and come up with the mean average result. It is downright unrealistic beyond extraordinary circumstances at the moment as the OP said that a crippled BB that was objectively inferior won against all odds. In an RNG system with multiple rolls and averaging, this simply would be an extraordinary situation, and that, (combined with the fact that the AI can track you out of sight range and perfectly track torpedoes) is probably the biggest issue the game has at the moment beyond regular unfinished stuff. The RNG system should ideally produce a bell curve, but as it currently stands that bell curve is either too flat or outright nonexistent. I imagine the game rolls a 1D500 with no secondary diceroll for something like a 10" gun against 5" of armor at 5K, whereas it should be rolling a 3D500 with after all results are added, the game divides this by 3. Granted, my biggest concern is how would the current engine (which to a point is laggy as is even on my current medium-high end machine) handle these new calculations?

Of course feel free to call me a dope or whatever I'm just some weirdo on the internet with an unusual obsession with floating blocks of steel for whatever reason, but that's my observation 🙂.

[Stormnet]

20 hours ago, MrStan53 said:

         Personally, as someone ironically actually coming on here to complain about the failings of RNGesus and to shriek XCOMmery, I think personally this is a bit too complicated of a solution. I'm actually gonna be the first to say RNG definitely needs to still exist (in fact I'm one of the few people who plays TF2, as a backburner pyro nonetheless, and wants random crits and spread). With this solution, it doesn't necessarily make the game fair to the AI either, which could eventually lead to player boredom or disinterest. It's a bit overdesigned as well and as the game can already run slowly at times it'd result in even laggier times. The issue, I think, is that RNG appears to be incredibly flat and uses only one roll to determine. I can't verify this for sure but playing the game it feels like hits and damage percentages vary far too much, and combining this with what feels to be an unfair advantage towards the AI in many cases. 

         To transition to a totally different game entirely, let's turn to the classic Dungeons and Dragons from the Satanic Panic era and before. Oftentimes with many dicerolls during the game (most notably during your character creation) many that could be done with D12s or D20s were actually substituted with 2D6 and 3D6 rolls. Though this was also to make picking up the game easier, it added a unique function. Most rolls would result in a bell curve formation, as a single roll of a 1 could be made averaged out by two 6 rolls. RNG was a factor for sure of course, however it was far less likely to do spectacularly or abysmally, making it easier to have more balanced characters and actions. For example, my most recent quest as a thieving halfling resulted in a very averaged character that was more resilient to sub par rolls (granted with physical 6 sided dice you can't have 0s, but as this is programming it's different).

         My proposal for a change would be first and foremost to make the RNG roll the dice more times and come up with the mean average result. It is downright unrealistic beyond extraordinary circumstances at the moment as the OP said that a crippled BB that was objectively inferior won against all odds. In an RNG system with multiple rolls and averaging, this simply would be an extraordinary situation, and that, (combined with the fact that the AI can track you out of sight range and perfectly track torpedoes) is probably the biggest issue the game has at the moment beyond regular unfinished stuff. The RNG system should ideally produce a bell curve, but as it currently stands that bell curve is either too flat or outright nonexistent. I imagine the game rolls a 1D500 with no secondary diceroll for something like a 10" gun against 5" of armor at 5K, whereas it should be rolling a 3D500 with after all results are added, the game divides this by 3. Granted, my biggest concern is how would the current engine (which to a point is laggy as is even on my current medium-high end machine) handle these new calculations?

Of course feel free to call me a dope or whatever I'm just some weirdo on the internet with an unusual obsession with floating blocks of steel for whatever reason, but that's my observation 🙂.

Four things:

1 Thanks for aknowledging that post exists.

2 Well... I do think your system is simpler and more resource efficient...

3 Thought regarding being fair or not to the AI, I originally thought applying this to all ships, so not even player ships got better treatment.

4 Despite not completely eliminating this issue, as there are no thresholds, it should drastically reduce single shot randomness, and reduce these possible yet rare situations to next to impossible.

 

 

Edited April 18, 2021 by Stormnet

[HailCOBRALA]

Yup. RNG needs to be adjusted or at least the percentages that apply TO the RNG. They say you can't fix RNG but you CAN change percentages to AFFECT RNG. I had an AI destroyer that was flooding, had 4% flood value left and miraculously recovered back up to 30. Hit him AGAIN, back down to another 4-5%. Back up to 30. AGAIN I hit him, back down to 4, back up to 30. It's ridiculous. That destroyer should have been dead after the first or at LEAST the second flooding, ESPECIALLY after his superstructure was down to 30%. Most of the crew would be busy fighting fires, shoring up compartments and.. well most of the crew would be DEAD. You wouldn't HAVE the crew to do that sort of counter flooding. This is what the game does not take into account. The amount of actual crew required to do damage repairs in WW1-2 and how useless it usually was in the end. 

[disc]

6 hours ago, HailCOBRALA said:

I had an AI destroyer that was flooding, had 4% flood value left and miraculously recovered back up to 30. Hit him AGAIN, back down to another 4-5%. Back up to 30. AGAIN I hit him, back down to 4, back up to 30. It's ridiculous. That destroyer should have been dead after the first or at LEAST the second flooding, ESPECIALLY after his superstructure was down to 30%.

It appears that a partially flooded compartment always will be pumped out eventually.

However, if the water reaches the top of the compartment, the space is permanently flooded and can never be emptied.

I believe this is why ships can recover from extremely extensive flooding. I am not sure if structural integrity (grey, green, yellow, red) is a factor.

Weird mechanic.

[Stormnet]

7 hours ago, disc said:

It appears that a partially flooded compartment always will be pumped out eventually.

However, if the water reaches the top of the compartment, the space is permanently flooded and can never be emptied.

I believe this is why ships can recover from extremely extensive flooding. I am not sure if structural integrity (grey, green, yellow, red) is a factor.

Weird mechanic.

I thought the "pumping" was related to the integrity of the compartment. Slightly damaged compartments should be able to pump out the water, while red compartments... that are no longer even compartments, shouldnt. 

This explains why sometimes I'm hit by a torpedo, and if it fills to the top, that vessel pretty much becomes so slow and useless in battle, while others just recover completely like nothing ever hit them.

[madham82]

And flood recovery also seems tied to the bulkhead setting. More bulkheads = less chance of a compartment being completely flooded, hence the magical recovery.