Jump to content
Sign in to follow this  
You need to play a total of 5 battles to post in this section.
Tpaktop2_1

Video - The Price of Randomness - Balancing RNG

15 comments in this topic

Recommended Posts

357
[TFK]
[TFK]
Alpha Tester
1,076 posts
12,879 battles

Here is a video by the Extra Credits people who have all sorts of videos on game designs. It is interesting when they bring up Portnow's Postulate.

 

 

So what do you guys think of this video? In RNGJesus's name we praise :Smile_honoring:

Edited by Tpaktop2_1
  • Cool 2

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles

While an interesting mental exercise, I don't see that it has much to do with WoWS.  Random events, e.g. shell hits, are too small to be visible as determinate of a loss.  Thus while there may be player frustration, it is not attributable to a random event.  Like poker, the game should be viewed as one of skill rather than one of chance.

 

Share this post


Link to post
Share on other sites
388
[C-CA]
[C-CA]
Beta Testers, In AlfaTesters
1,423 posts
3,367 battles
2 hours ago, iDuckman said:

While an interesting mental exercise, I don't see that it has much to do with WoWS.  Random events, e.g. shell hits, are too small to be visible as determinate of a loss.  Thus while there may be player frustration, it is not attributable to a random event.  Like poker, the game should be viewed as one of skill rather than one of chance.

iDunno, a salvo that misses or rolls at the absolute bottom for damage losses the player a pretty definite and visible amount of time, that being the time it takes to reload. It's not a huge amount, sure, but say you shoot five salvos in a row, all land centered on your target BUT RNG deigns that your dispersion for each shot will bracket the ship, either falling short or long but not hitting. That can and has happened to players in some ships *cough*Queen Elizabeth*coughcough*, and has nothing to do with "skill".

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles

A "RNG" shell miss out of what, a hundred?  two? is not only individually meaningless, but irrelevant as RNG variances disappear as the population increases. 

Of course, a player will always attempt to find an excuse outside of himself.

 

Share this post


Link to post
Share on other sites
388
[C-CA]
[C-CA]
Beta Testers, In AlfaTesters
1,423 posts
3,367 battles

You are aware how atrocious the dispersion is on some of these guns, right? This is hardly a "one in a hundred" thing, it's a "multiple times per salvo" thing unless you're at point-blank range blasting right into the belt. And in principle I'm not saying that every shell miss is necessarily a bad thing, the game would be less interesting and many ships too overpowered if every shot hit every time. But when every shell for five shots either lands short or long, with both misses in the same salvo and no hits, then something probably isn't right.

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles

That's a characteristic of the gun, not of RNG.  A gun has an intrinsic (in)accuracy quite independent of individual gun events.  A true RNG event is one like drawing a card in Blackjack: you either get a good card or you don't.  All the card decks are the same; there are no more or less favorable decks.  With something like guns, there are.  RNG does not determine the overall performance of that gun.

 

 

Edited by iDuckman

Share this post


Link to post
Share on other sites
388
[C-CA]
[C-CA]
Beta Testers, In AlfaTesters
1,423 posts
3,367 battles

But... The guns aren't real, they're data, a digital recreation of a gun which uses random number generation. If this were IRL then yes, it would be something completely uncontrollable. But it's not. It's a videogame. Literally the ONLY thing that determines how close your shot lands to where you aimed is RNG and other stats that the devs can control. Which is the point.

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles
23 hours ago, Landsraad said:

Literally the ONLY thing that determines how close your shot lands to where you aimed is RNG

NO.  You're missing the point entirely.  When you invoke a PRNG (Pseudo-Random Number Generator) you give it parameters that directly affect the result.  For a linear PRNG that would be the range, e.g. minimum 0.0, maximum 1.0.  The production of a value of 1.1 is impossible.  The allowed range can be tweaked, e.g. assume a hit is >=0.85:

  • accurate gun - (0.3, 1.0)
  • okay gun - (0.2, 0.95)
  • bad gun - (0.1, 0.9)

A bad gun will produce a hit far less often than an accurate gun.  (Of course, in WoWS it's much more complicated since the generator is non-linear and there are more parameters, but the principle is the same.)

Now contrast that with a pure RNG event, e.g. a random card from a deck. 

  • deck 1 - (0.0, 1.0)
  • deck 2 - (0.0, 1.0)
  • deck 3 - (0.0, 1.0)
  • etc..

On a fresh deck the chance of a given card is always the same: 1 in 52. 

The two cases are essentially different.

 

Edited by iDuckman

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles

Here, for instance, is a normal distribution such as it used in modelling shell dispersion. There are five different functions (lines) that vary only by σ (the Greek letter sigma, representing a standard deviation)**.  Let the x value represent the distance from the target (at x=0).  The y value at any point along the curve is the likelihood of the corresponding x value being produced by the PRNG. 

Note that ALL values generated by the PRNG fall under programmed curve and are limited in this case by x={-4,4}. (I.e. x=5 is impossible.)   Clearly, with the PRNG programmed to σ=0.5 (the red curve) an x value near zero [a hit] is far more likely than an x value of 2.  Most shells will hit, even though the result is actually random. 

With the PRNG programmed to σ=1.5 (the purple curve) an x value near zero is much less likely.  PRNG is producing a random value along the x axis, but the likelihood of a hit is much different just by varying the sigma parameter.

To further complicate matters, the limits on the range of the x value are the horizontal and vertical dispersion parameters of the gun.  That doesn't appear in this graph where the limits are presumed to be ±4.

As you can see, the "RNG" output can be -- and is -- sculpted by the programmed parameters in WoWS. 

Capture.JPG.8060f3781fea9f76204aaaad7619

** Clearly, the sigma values used by WG are an inverse of the sigma values in this graph.  A WG sigma of 2.5 would probably correspond to the σ=0.75 curve here.

Share this post


Link to post
Share on other sites
388
[C-CA]
[C-CA]
Beta Testers, In AlfaTesters
1,423 posts
3,367 battles

I think you're missing my point. You can't count cards in World of Warships, at least not where damage and dispersion variation is concerned. Yes, dispersion is not truly, technically, 100% "random". Yes, you do have control over aim and where the center of your dispersion circle lies. Yes, different guns do have different stats that you should familiarize yourself with.

BUT

Once you pull the trigger, skill has no effect on where the shells land or how much damage they do. This isn't like poker where only one event is random and the rest get progressively less so as a result of physical limitations; you can't draw four of the same card, but you can hit the same area of dispersion four times in a row (though it is very improbable). The better analogy than poker would be a tabletop RPG. You can add whatever skills and modifiers to a roll that you want, but a critical failure is still a critical failure, it has to be for game balance. You can get three ones in a row, just like you can have three salvos that go high and overpen in a row. THAT is my point. RNG, PRNG, Weighted RNG, whatever, which one it is doesn't matter. It is an event with a randomized outcome that, though not outside of the player's INFLUENCE, is outside of their total CONTROL. A random event.

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles

No, it's not a random event.  It's a series of events, each of which is influenced by randomness.  Both poker and WoWS are games of skill.  You might say that part of the skill involves managing random events.  There are players that excel in WoWS even though their shells sometimes miss.  There are players that excel in poker even though they are sometimes dealt losing hands.  And conversely there are potatoes.

Skill doesn't affect a random event.  How you throw a 20-sided die does not (or should not) affect the result. 

What are we arguing about discussing again?

 

Share this post


Link to post
Share on other sites
388
[C-CA]
[C-CA]
Beta Testers, In AlfaTesters
1,423 posts
3,367 battles

Well, I was pointing out that things like shell dispersion and damage rolls are inherently random and should be treated as random events, you appeared to be saying that there is no randomness and it's all skill. At least, that's my perspective on it.

To go to your example in the first paragraph, yes, skilled players can make the most of a bad situation, but that does not mean that their success is inherently unaffected by random chance. Potatoes aren't the only ones the get streaks of low damage shots, low accuracy salvos, or detonated; and unicums aren't the only ones that get streaks of high rolls and random tight groupings that all citadel and delete a ship despite the sigma value of the guns they're using. Skill has no affect on the RNG factor beyond a certain point. A good player in WoWs is someone who knows how to work with the mechanics, to optimize their build, and manage their movements in such a way as to do maximum damage (or teamplay) for minimum risk. It's the same as a game like D&D. You can think of player skill, captain skills, ship upgrades, etc. as the modifiers a character has to their damage or hit chance. You can have a super awesome Rogue who's practically Spider-man with +8 dexterity or what have you, but it still won't guarantee a success on a maneuver that needs a skill-check of 15 to succeed, a 1 to 6 on that d20 will still fail. Poker meanwhile, which I really wish people would acknowledge isn't just about skill, but whatever, seems like a much MUCH more simplified version of this dilemma to the point that yes, in many cases skill DOES all but completely cancel out randomness. Because poker has far fewer variables to take into account. What do you have to worry about in Poker? X number of players, bets, and 52 non-repeating cards. What do you have to worry about in an RPG or WoWs? All the other players, random events that ARE repeatable, numerous stat-blocks that may or may not even be used in that particular scenario, terrain placement, trigonometry and ballistic trajectories (in boats' case), etc. etc. etc.

Okay wow that came out more wall-o-text-y than I intended. But, uh, anyway, that clear up my point?

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles
9 hours ago, Landsraad said:

Skill has no affect on the RNG factor beyond a certain point.

I say the reverse is also true.

Re: poker.  It depends on how you define the game.  If the object is to win hands, then yes, it's primarily RNG-driven.  But that's not the object for stakes players; it is to take everyone else's money.  A very good player can do that with the worst cards in the deck.  Skill, not RNG.**

I still maintain that while there is a randomness factor in WoWS, it is rarely determinate of the outcome. 

Maybe we should just call an impasse.

 

** A skill that escapes me.  But at Spades, watch your wallet.

 

Share this post


Link to post
Share on other sites
388
[C-CA]
[C-CA]
Beta Testers, In AlfaTesters
1,423 posts
3,367 battles

I'm not saying that skill is irrelevant in determining outcome either, I'm just saying that RNG is a factor which needs to be accounted for when balancing the game and which can determine the outcome of match. Not that it does ALL the time, but that it can enough that it shouldn't be dismissed out of hand.

So basically what I'm getting is that we were trying to convince each other of roughly the same thing from two different directions with two different wordings.

Share this post


Link to post
Share on other sites
1,249
[WOLF3]
Members
6,497 posts
2,439 battles

That's fair. 

Glad someone cares enough to be analytical.  :cap_book:

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×