Tighten spread, reduce damage

(Kendle) #341

[QUOTE=its al bout security;348003]i hope you arnt applying this to assault rifles?

im not sure of your context but i am sure this isnt correct for ARs[/QUOTE]

When firing from the hip I think it is, ARs go from min to max spread almost immediately and take noticeably longer to return back to min, in my experience. Of course I could be wrong, but I think the ARs are only really viable when ADS.

IMO burst firing isn’t really an option, because the damage you can do full-auto (even tho’ most of your bullets miss) out-weighs the accuracy you gain by bursting.

(Ghostdog) #342

Trying to rationalise the idea of random spread, I thought of this.

(suho) #343

You are right. If both players were aiming for the head the one with the better aim would probably win the fight more often if repeated. But it is still pretty damn random.

The problem with huge spread is however that aiming at the head is not a good idea because a lot of bullets will miss, basically half the circle isn’t on target. Players will alway aim and shoot the way they get the maximum damage output. In this case that means aiming at the torso because then you have a much better alignment of the circle and the target and the head is still within the circle for both players. Since the head is not in the center anymore, less bullets will result in headshots but the overall damage is higher because more bullets are going to hit the target. That way the game decides who is going to get the headshots and who is not.

It takes control away from players. The game decides about my fate not ME! If I loose a fight I should blame myself and not the game for giving me bad luck! How am I going to improove at all in this scenario? I have no inclination to do so when I know it depends on luck rather than skill. If it were skill dependend and I got owned by someone very good, next time I see this guy ingame I would try and concentrate even harder on my aiming because the game forces me to do that I if I wanted to win that fight.

(shirosae) #344

Yeah, this was dealt with yesterday.

But y’know, thanks for playing. And also for the demonstration that you aren’t reading the thread before spewing the nonsense; I was wondering, but the confirmation is nice.

Anyway, I was thinking a bit more about the DoD even-spread stuff Kendle was talking about.

If you take the totally circular spread pattern, and assign the maximum radius R, and scale it so R can run from 0 to 1 (so a bullet can land anywhere in the circle), then the area of the circle is pi.

The total probability of a bullet landing inside the circle is 1, so you have an easy conversion from area size into probability.

The inner ten percent of radius (R = 0 --> 0.1) generates an area of 0.01 pi --> 1% chance of bullet hit.

The outer ten percent of radius (R = 0.9 -> 1.0) generates an area of 0.19 pi --> 19% chance of a bullet hit. (EDIT: Fix’d)

(Originally read: The outer ten percent of radius (R = 0.1 -> 1.0) generates an area of 0.19 pi --> 19% chance of a bullet hit.) D’oh. The outer 10% isn’t R = 0.1 -> 1.0.

This is trivially obvious, since the outer 10% radius region of the ring has 19 times the area of the inner 10% radius region.

That is, the cumulative probability of finding a bullet rises as r^2. For weapon effectiveness, an ‘even’ spread would be a gun where the cumulative probability rises linearly, so there’s as much chance of a bullet being at max radius as there is of it being dead on.

Trying to balance things in terms of areas just breaks everything.

It’s obvious as soon as you focus on the system properly, but even so, wow. This is kinda deceptive at first glance. Even spread doesn’t help bad aim guy in the diagrams I posted because the areas are equal, so by definition both probabilities will be the same, but yeah, wow. Oh, also, I intentionally set the ‘lottery spread’ pictures up with roughly the same area of covered chest/shoulders between the good/bad aim, to make the other odds even.

(Spendlove) #345

Boring poll is boring.

Someone edited a post to make a comedic dig that Brink has only a thousand players left on all three platforms. Even then that still says that a whopping 90% of current comedy active players don’t give a flying fark about this poll.

Move on whiners, move on.

Can’t one of you find anything positive to say? A dev posted somewhere on this forum that if they gave you a wheelbarrow full of gold you would complain it was too heavy.

Fail community really. Such a shame. Still I and apparently the hundreds of players in my region with less than 100 ping enjoyed the game last night? We had fun. Perhaps you guys should play the game or gtfo and plat etqw which your also moaning haz no playaz. Go play then and perhaps it will.

(suho) #346

ehm? 1% + 19% == 20% != 100%

P = (Area(R=1)-Area(R=0.1)) / TotalArea

The inner ten percent of radius (R = 0 --> 0.1) generates an area of 0.01 pi --> 1% chance of bullet hit.

The outer ten percent of radius (R = 0.1 -> 1.0) generates an area of 0.99 pi --> 99% chance of a bullet hit.

(shirosae) #347

Oops! This is supposed to read (R = 0.9 -> 1.0).

I’ll amend the original post, thanks for the catch.

(suho) #348

ahh now I understand your point :slight_smile:

(suho) #349

Actually there is a slight problem if you think about the implementation. Since it is a 2D problem two random numbers need to be determined. I guess it being done in the way that two angles are determined vertical/horizontal spread form the tip your your gun. That determines the trajectory of the bullet. So in order for the bullet to be close to the edge of the circle two events need happen that is both random numbers (angle) need to be very high.

EDIT: Scratch that it bull****, you just need one to be high but there is still a relationship bwtween the two.

(shirosae) #350

Just saw your edit: Yep.

You could do it a load of different ways. You could do Cartesian co-ordinates, but you have the problem of confining the spread to a circle without compression screwing up your evenness.

If you just RNG your x co-ord, then RNG your y-co-ord and scale it to fit the vertical scale if your circle at ‘x’, then you get compression as you go to either side (you have a square matrix, compressed into a circle shape).

You could RNG both co-ords, then check to see if they fit inside your circle and re-roll if not, but that’s silly.

Polar co-ordinates makes more sense (since it’s easier to give individual radius and angle weighting), but I wonder if it’d be even easier to just precalculate a bunch of quantised angles and give them all names, then assign your randomness via weighted RNG.

(suho) #351

I guess that is more efficent. Im not really a programmer but I don’t see how you would calculate a ‘radius’ because it depends on how far away the target is. Or do u mean determining the trajecrtory by assuming a target wall one meter away from your gun? I could see that work.

(Kendle) #352

The point of the poll, and the discussion in general, is to petiton SD to make changes so that instead of you and 100’s of others enjoying the game, you and 1000’s or 10’s of 1000’s of others can enjoy the game.

(tokamak) #353

For the direction in which the bullet spreads you can easily use fixed numbers, sure. But using it for the amount it spreads is redundant as the amount is already dynamic.

(Shinigami) #354

before i make an opinion on this topic, i would like to ask for some knowledge. The knowledge I have on programming is very basic, although I have very good knowledge of algorithms and mathematics.

So, how exactly is it that this spread is randomly generated? is there an algorithm that causes the bullets to randomly generate a direction?
e.g. if i were asked to do a program for something like this, the approach i would take in this scenario would be to convert that spread circle into polar coordinates, and set the algorithm up so that the actual probabilities are higher for lower values of the radius, thus higher values of the radius (exterior of circle) would have less probability. the angular variable would not be important in this case.

although i must say, i have never really understood how it is possible for algorithms to select a random choice. only way i see that possible is with the variable of time, but even that would require some sense of randomness.

please bare with me in this post. like i said, i have knowledge of math and algorithms, but no so much for programming.

edit: oh i didnt notice people were already talking about this kind of thing until after i made my post! hah im a dork i enjoy conversations like this

#355

[QUOTE=shirosae;348046]Yeah, this was dealt with yesterday.

But y’know, thanks for playing. And also for the demonstration that you aren’t reading the thread before spewing the nonsense; I was wondering, but the confirmation is nice.

Anyway, I was thinking a bit more about the DoD even-spread stuff Kendle was talking about.

If you take the totally circular spread pattern, and assign the maximum radius R, and scale it so R can run from 0 to 1 (so a bullet can land anywhere in the circle), then the area of the circle is pi.

The total probability of a bullet landing inside the circle is 1, so you have an easy conversion from area size into probability.

The inner ten percent of radius (R = 0 --> 0.1) generates an area of 0.01 pi --> 1% chance of bullet hit.

The outer ten percent of radius (R = 0.9 -> 1.0) generates an area of 0.19 pi --> 19% chance of a bullet hit. (EDIT: Fix’d)

(Originally read: The outer ten percent of radius (R = 0.1 -> 1.0) generates an area of 0.19 pi --> 19% chance of a bullet hit.) D’oh. The outer 10% isn’t R = 0.1 -> 1.0.

This is trivially obvious, since the outer 10% radius region of the ring has 19 times the area of the inner 10% radius region.

That is, the cumulative probability of finding a bullet rises as r^2. For weapon effectiveness, an ‘even’ spread would be a gun where the cumulative probability rises linearly, so there’s as much chance of a bullet being at max radius as there is of it being dead on.

Trying to balance things in terms of areas just breaks everything.

It’s obvious as soon as you focus on the system properly, but even so, wow. This is kinda deceptive at first glance. Even spread doesn’t help bad aim guy in the diagrams I posted because the areas are equal, so by definition both probabilities will be the same, but yeah, wow. Oh, also, I intentionally set the ‘lottery spread’ pictures up with roughly the same area of covered chest/shoulders between the good/bad aim, to make the other odds even.[/QUOTE]

Actually, my flame spewing community member, I must apologise, my browser had not refreshed the following page of comments even though I had pressed “F5” this morning.

So yes, it was said before, I just had not seen it…yet. So apologies. But thanks for testing my flame retardant jacket! :cool:

(suho) #356

[QUOTE=Shinigami;348080]before i make an opinion on this topic, i would like to ask for some knowledge. The knowledge I have on programming is very basic, although I have very good knowledge of algorithms and mathematics.

So, how exactly is it that this spread is randomly generated? is there an algorithm that causes the bullets to randomly generate a direction?
e.g. if i were asked to do a program for something like this, the approach i would take in this scenario would be to convert that spread circle into polar coordinates, and set the algorithm up so that the actual probabilities are higher for lower values of the radius, thus higher values of the radius (exterior of circle) would have less probability. the angular variable would not be important in this case.

although i must say, i have never really understood how it is possible for algorithms to select a random choice. only way i see that possible is with the variable of time, but even that would require some sense of randomness.

please bare with me in this post. like i said, i have knowledge of math and algorithms, but no so much for programming.[/QUOTE]

Random number generator give you the same sequence of numbers if the starting conditions are the same in that sense they are not truely random. Except some CPUs have an internal generators which depends on thermal noise and is used for generating numbers for encryption.

(suho) #357

The only way for to implement such random spread would be to calculate a radius and an angle in terms of polar coordinates on an artificall wall that is 1m away form your gun. That would also be an easy way to change the probabilties we were talking about earlier since you only have to change the probility density function for the radius.

@tokamak: You can easily just multiply the predetermined numbers for the radius by a dynamic number which increases as you hold down the fire button.

(Shinigami) #358

oohh thats cool. i wasnt really familiar with thermal noise. but leaving ‘decisions’ to natural occurrences is as close as we can get to generating acutually generating randomness

(Apples) #359

One distribution of probability in x, one other in y, you have your joint distribution, tadaa, you got your “random” into your “circle”, yes its not really a circle and yes it’s not random (well it can be random if your two distributions are uniform laws for exemple, if they are normal laws then its weighted toward whatever you want), just generate some variance around your distribution, adding some noise, and you’ll have some sense of randomness (which doent exist by the way). to include the distance you just have to move the minimum and maximum bound (in case of uniform law) of the distribution AND the variance too (in case of Normal laws)

Mmm I did something similar in a bayesian approach with markov chains (Monte carlo simulation) with somthing about pi and a circle, I will look thru my papers see if I get this back.

Peace

(suho) #360

[QUOTE=Apples;348094]One distribution of probability in x, one other in y, you have your joint distribution, tadaa, you got your “random” into your “circle”, yes its not really a circle and yes it’s not random (well it can be random if your two distributions are uniform laws for exemple, if they are normal laws then its weighted toward whatever you want), just generate some variance around your distribution, adding some noise, and you’ll have some sense of randomness (which doent exist by the way). to include the distance you just have to move the minimum and maximum bound (in case of uniform law) of the distribution AND the variance too (in case of Normal laws)

Mmm I did something similar in a bayesian approach with markov chains (Monte carlo simulation) with somthing about pi and a circle, I will look thru my papers see if I get this back.

Peace[/QUOTE]

Yeah in effect you have a joint distribution but if you generate 2D polar coordinates and project those into a 3D trajectory only the radius does determine the spread, that is if I define spread by distance to center of circle. Looking at where the bullets go in the game I guess the angle is pretty much a uniform random number and the radius is a bit more complex.