Planetside 2 - Page 144

There are certainly a lot of "non-optimal" gunplay mechanics in the game, but flinching has to be top contender for the worst. I hid this discussion away because it's somewhat off-topic. The rest of this section provides context with which to view the nanoweave and gun performance analyses at the end of the post - this little section here is for me to basically bitch about this terrible mechanic.

What makes flinching so bad? The main issue is that flinching (as of the time of this writing) is a culling mechanic which invalidates a large amount of content present in the game. Imagine playing an RPG with a huge amount of diverse abilities and ways to build your character. Now imagine that you can build a character around stunlocking - i.e. you can load this character up on stuns so that it can constantly stun opponents. Even worse, imagine that the enemies in the game are consistently designed to stunlock. All character builds that aren't stunlock-oriented are invalidated.

What do I mean by invalidated? I mean that all other character builds are suboptimal in comparison. Given two groups of enemies to fight, any player-based character would struggle more with a stunlocking group than a non-stunlocking group. The game has made the player choose between being efficient and being allowed to explore the diversity of the rest of the combat mechanics.

Suppose we have a non-stunlock character build. It has balanced fights against non-stunlock opponents - in these fights we can see all the glorious diversity allowed by the game mechanics. However, when this character build is met up with stunlocking opponents, it simply loses - in efficiency if not altogether. Stunlocking is a mechanic which, by design, overrides other game mechanics.

Flinching is Planetside 2's stunlock. Take two players of equal skill. Give one of them a high-ROF weapon and the other a low-ROF weapon. For the purpose of this experiment, make the weapons deal the same DPS (in Planetside 2, low-ROF weapons always deal less DPS) and have their other stats scaled appropriately. Have them duel 100 times. I can guarantee that the player with the high-ROF weapon will win the majority of duels, primarily because of the flinch mechanic. Player 1 will have better flinch coverage (more reliably flinches the target) and a better defense/reaction against flinch (more reliably flinches the target while being flinched). This means Player 1 is more likely to cause Player 2 to miss shots than vice versa.

People offer all kinds of "advice" about how to "counter" flinching. The problem is that all the "counters" to flinching (e.g. "don't get shot first, take cover and flank, run away if you take the first hit") can be utilized by players who are "built" for flinching - i.e. players who are using high-ROF weapons. This is exactly the situation with the RPG where stunlocking is the king of builds - stunlockers are the strongest opponents of other stunlockers.

It's not just low-ROF automatics that get hurt by flinching, either. The worst casualties of this problem, the guns that have been the "most invalidated", are the semi-automatics. Why are the battle rifles so bad? Primarily because of flinch. There are other reasons (their low bullet velocity is a big one), but flinch is the knife in the gut.

I play primarily infiltrator, and one of my biggest frustrations with the game is how it punishes me for trying to use the Nyx - the VS semi-auto scout rifle. There has never been any other gun in the game where I've consistently had the experience of firing on someone from behind and being unable to land that last killing bullet because the target literally turned around and flinched me until death + Show Spoiler [ironically...] +. It happens when I play my NC character where I've been using mostly the NC6 Gauss SAW but not nearly as frequently or blatantly. It's still a problem, though, even with guns like the Flare - you find situations where you lost the flinch war. You missed too many shots and died first even though you had the initiative and fired first.

Is this just me being bad at the game? Maybe to some degree, but it's mostly the game punishing me for choosing to use a sub-par weapon.

One other thing to note is that flinching doesn't only punish low-ROF weapons; it punishes aiming in general (as does screen shake). Flinching when not ADS is barely noticeable and has no real effect on the victim. This makes it so that hipfiring is the default preferred method of firing rather than being a sidegrade-oriented playstyle or a situational benefit (going ADS takes precious milliseconds that sometimes will get you killed).

All that said, what does the game reward? If we should avoid low-ROF and semi-auto weapons, what should we go for to properly participate in efficient, effective combat? Probably anyone with enough experience playing this game will have the same answer: the high-ROF hip-fire weapons such as the VX6-7, the LC2 Lynx, and the GD-7F.

These carbines have extreme ROF, giving them among the highest flinching capabilities in the game. In addition, they also all come with Advanced Laser Sight upgrades, allowing them to bypass flinch from opponents very effectively. It's like a Sword of Flinching with a +5 Resistance to Flinch pommel. Those of you complaining about the "lack of metagame" - these guns are the metagame (sadly, I'm only half-joking).

They all also have unidirectional recoil and thus aren't "capped" in their ranged ability like their balanced recoil "sidegrades". If they could be said to have disadvantages it's that they only have average First Shot Recoil (except the VX6-7 which has the worst in the game) and they have lower bullet velocities (except the GD-7F which has average bullet velocity [4th out of 10 velocities]).

In addition to being flinch-dealing, flinch-resistant weapons, they also are too good at range in comparison to the other carbines. If you compare the Solstice to the VX6-7, the Solstice just doesn't stack up except at longer ranges. These weapons which should be highly-specialized sidegrades meant for close-combat domination are too good in other areas, making them jack of all trades. They deal the best flinch, they have the best resistance to flinch, they are extremely strong in CQC and are quite good at medium range and workable at long.

There is no reason to use any other weapon because its advantages over these guns will be overshadowed by its disadvantage with respect to them - i.e. the opportunity cost of not using these guns is too damn high. Shotguns are better in close range but practically unusuable anywhere else. Some of the SMGs are arguably better than these guns in close range, but they fall very short at mid range and beyond. The "sniper" carbines (like the Pulsar C [one of my favorite weapons, btw]) are better at long range but will lose the flinch war at medium and are just plain worse in close. The versatility of these weapons, their power at every range, violates the sidegrades philosophy and instead makes all other guns niche weapons in comparison to their "ultimate soldier" status.

The fact that they are also the best flinchers and have the best defense against flinchers is just a joke and highlights all the problems associated with flinching and all the potentially cool and interesting content the game would otherwise have to offer.

I'm not trying to pick on these weapons. The major problem with them is the flinch mechanic which serves only to multiply their advantages to the point where they are no longer sidegrades.

SOE, if you are reading this, I want to play your game in its fullness. I want to spend money on the battle rifle and not regret it. I want to get Auraxium on the Nyx without it being a badge of suffering. Remove flinch from the game or at the very least remove the advantage that high ROF provides with flinch.

Just as an addendum, I'm totally fine with keeping flinch on sniper rifles, or even making it worse, for balance/gameplay reasons.

Note: All the calculations I do are in floating point precision (not double) to closer mimic what the game engine likely uses, but I show most values as 4 decimal digits for space and demonstrative purposes.

The given, known COF values that I tested are

COFs = 1.0000    1.2500    1.5000    2.0000    2.5000    3.0000    5.0000    5.5000    6.5000

and the measured values of the inner and outer radii (in pixels) are

outerRadii =  28    32    36    44    51    59    90    97   113
innerRadii =  16    20    24    32    39    47    78    85   101

Notice that the COFs start at one and grow slowly. I try to emulate this behavior by dividing each set of radii by its first entry.

innerRadii/innerRadii(1) =

    1.0000    1.2500    1.5000    2.0000    2.4375    2.9375    4.8750    5.3125    6.3125

outerRadii/outerRadii(1) =

    1.0000    1.1429    1.2857    1.5714    1.8214    2.1071    3.2143    3.4643    4.0357

Look how very close the innerRadii are in comparison to the COFs. They're surely the ones that represent the COFs - so I'm going to conclusively state that the target zone is contained inside the little crosshairs. This might be "obvious" to FPS players, but the kind of answers I'm trying to get at demand that I make no assumptions until I'm out of options.

There are some slight deviations between the scaled radii and the COFs.

COFs - scaledRadii =

         0         0         0         0    0.0625    0.0625    0.1250    0.1875    0.1875

At first it appears that the COF values and the measured radius of the crosshairs has a direct one-to-one (linear) relationship, but the measured values and the known COF values are growing farther apart as the COF value increases. To understand why, we need to work out some of the relationships involved in the triangle using some simple voodoo devilry trigonometry. We can see that the relationship between r and alpha is not a direct proportionality, there is a sine function in the way.

This is evidence that the COF stat might actually be the angle 2*alpha. Anyone with engineering or physics background is likely to know about the small-angle approximation that sin(x) = x if x is sufficiently small. It takes the sine function a bit to get going so zoomed in towards x=0 sin(x) looks very much like x. This would explain how r and alpha appear to have a one-to-one relationship at first but start to diverge farther away.

Now we can start making some assumptions. Let's assume that COF is the angle 2*alpha. We have a measured relationship between r and alpha and a theorized relationship as well ( sin(alpha) ~ r or arcsin(r) ~ alpha). So let's compare them. Using a computer it's simple to construct our theorized model and to automatically account for conversions (remember, the measurements we have are in pixels, not in whatever elemental units the game engine or original designer intended). It's not an exact fit, but it's damned close. There are considerable sources of error each step along the way - even the precise measurements made of the radii suffer from being converted from floating point into discrete pixels.

This close relationship along with the other suggestive evidence makes me confident enough to declare that the COF stat is the angle 2*alpha (in degrees, not radians).

We'll keep in mind that we have a simulated relationship between the COF stat and the target zone radius (in pixels) because it will be useful later on.

+ Show Spoiler [TLDR] +

Why would we ever burst fire? I've seen some truly awful explanations involving recoil, but the real answer is that it dramatically reduces the average COF per bullet which causes our accuracy to skyrocket.

If I was asked, without having access to the accepted stats, what the effect of pausing between shots did to COF, I would have said that as soon as we lift our finger off the mouse button COF is set back to its base value. However, the keepers of the stats claim that COF drops by 20 per second in this case. To put that in perspective, the fastest-firing non-SMG in the game fires at 845 rounds per minute or ~14 rounds per second. That is a refire time (time between bullets) of .071 seconds or 71 milliseconds. If we pause for the duration of one shot between shots, we can shed up to 1.42 COF. That is a huge amount considering that most ADS COF when moving starts at .3 or .4 and only gains .05 per shot.

Let's do an example. Consider the TRAC-5 (the default TR carbine) which has .3 base ADS COF and .05 gain per shot. We'll simulate firing 20 rounds full auto and in 10-shot bursts.

After 20 shots full auto, the TRAC-5 has .3 + 20*.05 = 1.3 COF.
After 10 shots full auto, the TRAC-5 has .3 + 10*.05 = 0.8 COF.

What about the average COF per shot? Full auto it comes out to .825 per shot where burst fire comes out to .575 per shot. Burst fire has 30% less average COF per shot at the expense of taking the time of 21 shots instead of 20.

Burst fire is obviously great on any gun for this reason, but it's much better and more consistent on guns with low First Shot Recoil. Whenever a gun's stats don't seem to make sense in the grand scheme of things, look to its FSR. Take the NS-11A (NS medic gun) or CME (VS medic gun) for example. Subpar ROF without a damage increase. What's the point? The FSR multiplier is 3x which is the worst in the game. This is a signal that this gun is designed for full auto fire, and we can see evidence of that in its other stats. They both have unidirectional recoil and very low base ADS COF.

Look at the VX6-7, another 3x FSR. It's a hipfire-specialized high-ROF carbine. The FSR is an attempt to keep its potential for medium and longer range in check.

Look at low-ROF high-damage weapons. AC-X11, Pulsar C, GD-22S, TMG-50 - these all have very low 1.5 FSR. In fact the only notable gun in the game with higher ROF and low FSR is the SVA-88 (VS LMG with 1.5 FSR and 698 ROF) - definitely an interesting outlier since VS guns tend to have the worst FSR in their class. These guns are simply amazing at consistent burst fire and so allow a player to more easily make the most of them at longer ranges.

As of the time of this writing, the information regarding balanced horizontal recoil is not easily available. Without accurate numbers on a per-gun basis, I can't simulate the effect of balanced horizontal recoil on accuracy - so I've chosen to ignore it entirely. Just know that balanced horizontal recoil is ruinous to a gun's accuracy at longer ranges - the farther you go from 0m, the worse the accuracy penalty gets. In a worst-case scenario, even with perfect aim, balanced horizontal recoil is expected to cause an additional miss rate of 50%. That is drastic and something readers need to keep in mind.

The "journey" up to this point is filled with sources of error - i.e. we expect differences between our calculations and what will actually happen in the game engine. I've artificially limited the range at which we will look at weapon stats to 80m. In my professional opinion, things inside this range should be fairly accurate with only minor differences between game engine and calculations.

It's important to note that the sources of error we met with are not random. This means that the relative value of calculations is not going to be questionable - better stats will show better performance. This is just a warning not to treat everything as if it's one-to-one with the game engine.

Additionally, the examples shown will all assume 100% perfect aim, something even aimbots aren't sophisticated enough to do without altering weapon stats.

If we have chance to hit, how do we determine expected shots to kill? If bullet damage provides that a target will die in 10 shots, and we have 50% CTH, we expect that it will take 20 actual shots to kill the target. This does not mean it will definitely happen, it's just what we expect. Let's look at that another way.

x * .5 = 10

In this situation, 'x' is what we're looking for - the effective shots to kill, the number of shots we'll need to actually fire to get the kill. What if we change the probability?

x * .8 = 10

If we have 80% chance to hit instead, how many shots are we looking at firing? Just solve for 'x' again: 10/.8 = 12.5. Now before someone objects, yes we can't fire half of a shot in this game. This is just demonstrating that the expectation value is not necessarily a possible value - just like the expectation value for rolling a 6-sided die is 3.5. What this means in actual game terms is that we will often kill the target in 12 or 13 shots (though there are certainly other possibilities).

Let's look at this a different way - from the perspective of each bullet firing individually. Each bullet can either hit (and so it adds 1 bullet to our tally of shots we need to kill the target) or miss (and so adds 0). Let's go back to a 50% chance to hit. With each bullet comes a 50% chance to add 1 to our tally but also a 50% chance to do nothing. What does that look like for 1 bullet?

1*.5 + 0*.5 = .5

This is just restating what we said before. If we fire 20 bullets, we'd just add up the previous statement 20 times.

1*.5 + 1*.5 + 1*.5 + ... + 1*.5 = 10

This is the same result as when we used

x * .5 = 10

which it ought to be since the statistics shouldn't change based on how you look at them. I'm trying to lead to a subtler point, though, and that's the idea that each bullet we fire potentially has a different chance to hit because each bullet potentially has a different COF. What's the difference between firing two shots with 50% CTH and firing two shots, one with 30% and one with 70% CTH?

1*.5 + 1*.5 = 1
1*.3 + 1*.7 = 1

They have the same expectation value. The only difference between the two cases is we can't easily write down something and solve for 'x' in the second case - we have to add up each contribution from each individual bullet as there's no shortcut.

So all that said, how do we get effective STK? We need an actual STK (i.e. how many direct shots it takes the target to die). We need information on every bullet (potentially every bullet in the magazine) to determine its CTH based on COF. With these two we can just do the following:

1*CTH1 + 1*CTH2 + 1*CTH3 + ... + 1*CTHn = STK

Simple, right? Except there are some details we need to work out.

First of all, how do we know how many bullets it will take to add up to the correct STK? We don't so we have to start with the first bullet, ask whether it's enough to add up to the STK, if not add the second bullet and so on until we add up to the STK.

Secondly, what if the bullets don't add up exactly to the STK? Let's say the STK is 10, but we've added up a bunch of bullets and it turns out to be 9.99. We're going to reject that until it adds up to 10. What if the bullets add up to 10.2? Well that's more than we need, but sufficient is sufficient so that does the job. There are other ways we could have addressed these issues, but doing it this way sidesteps the "how do I calculate TTK for half a shot?" issue and neatly discretizes the bullets we need to fire.

Luckily, working out time to kill from shots to kill is very simple. Imagine the target is at a given distance standing still. From the time we pull the trigger to the time he dies is what we're trying to calculate. Each bullet after the first is separated by a refire time. Each bullet has a travel time, but we only need to calculate the last one since all others will arrive before it.

Let's say we fire 5 shots. The time between the first and second shot is just the refire time. Same for the 3rd and 4th and for the 4th and 5th. To put that a different way.

1<-refire->2<-refire->3<-refire->4<-refire->5

5 Shots and 4 refires. The behavior is repeated with any number of shots, so we are looking at n-1 refires for n shots.

Now what about travel time? Each bullet has a velocity, and we know the distance to the target - so we'll just work out the relationship between velocity and distance. The time-integral of velocity should give us distance traveled (from the differential relationship). Here we are looking for a specific constant distance, so x loses any time-dependence. We also are going to approximate that V is constant (it isn't constant in 1 dimension unless you have 0 bullet drop - but it's very, very close). An integral is just a linear operator so we can take constants outside of it leaving us with a trivial integral. In order to get the travel time t for a bullet fired at a target x meters away, we simply calculate x/V.

So what is the final formula for TTK? It's simply

(STK-1)*refireTime + travelTime

That's for full auto fire. What about burst fire? It's not the Best Possible Way, but the method I use to burst fire is timing a "dead" shot between shots. In other words, I take a break of ~1 refire time before resuming fire. This is the way I've found to burst fire most consistently not necessarily with the most efficiency (i.e. highest DPS).

To adjust the formula for this, we simply add 1 refireTime every n rounds where n is the number of shots in a burst.

As a note, these all show differences between the two. They are given in the form

gun1 - gun2 statistic

meaning that we take the statistic of gun1 and subtract the statistic of gun2.

+ Show Spoiler [Bullet Damage] +
+ Show Spoiler [Full Auto] +
+ Show Spoiler [5 Shot Burst] +