Weapon Discussion Thread

[Dioxine]

What's idealized about it exactly? And how is a model which says that a brain buster is as likely as a body shot closer to reality than a model which says the former is significantly less probable than the latter? Even if your issue is with perfect accuracy not being a thing, that's a separate issue with a separate solution.

I've explained it 2 posts ago, so I won't repeat myself.
However -
If faced with a problem of reality not conforming to a model, you can do any combination of the following:
- Abstract the known and unknown unknows away, using a rough function, based on a rule of thumb that more or less accounts for all problems (the direction I favor in this case)
- Try to get closer to reality, by summoning new, more accurate models. This will give you a more accurate model, but you have to travel paths of ever-increasing complexity. Sometimes it's needed, sometimes it's not.
- Ignore a set of problems to idealize the model towards some perceived elegance.

[Surrealistik]

I can see you feel you have explained it, and that you favour the first case while assuming I'm all about the third rather than vice versa, but that is not at all the case.

Gaussian is the rough, albeit imperfect summation and abstraction that addresses all problems; alternately it's number two if you feel that an accuracy cap is required, thus requiring a small additional step/level complexity to the solution.

Flat ignores an obvious and glaring problem with all injury outcomes not being anywhere close to equally probable for the sake of (excess) simplicity, based on an erroneous assumption that this somehow compensates for the potential for perfect accuracy. Personally I think the reality is you just love the RNG too much to let it go.

[Dioxine]

I can see you feel you have explained it, and that you favour the first case while assuming I'm all about the third rather than vice versa, but that is not at all the case.

I never said that much. I could, but it isn't so simple as that; that's why I said it's always a combination of these three ways. Also, you may, naturally, discard my explanations.

Gaussian is the rough, albeit imperfect summation and abstraction that addresses all problems; alternately it's number two if you feel that an accuracy cap is required, thus requiring a small additional step/level complexity to the solution.

Only assuming you sweep all the irregularities that effectively flatten the curve (to some degree) under the carpet. In the wider context, Gaussian is idealized, as it assumes that when a bullet hits, it hits more or less true. It makes a clear distinction between target and environment, which is also idealization.
Also, Gaussian, rough? Just look at the function. It's perfect, round and symetric. Not that drab and boring flat line

Flat ignores an obvious and glaring problem with all injury outcomes not being anywhere close to equally probable for the sake of (excess) simplicity, based on an erroneous assumption that this somehow compensates for the potential for perfect accuracy. Personally I think the reality is you just love the RNG too much to let it go.

Woo, the Master Psychologist and Love Counsellor in action! Are you interested in mechanics, or people? Since if the latter, there is nothing to discuss.
As for your obvious and glaring problem - yes, flat 'ignores' it, because it puts the whole affair in a wider context than just simply 'you got a direct hit by a bullet'. So the goal is to abstract the problem away, not to simplify it. It is imperfect by far, but going deeper would require adding mechanics for critical fails and critical successes.
Also, what error is in my assumption that chaotic events flatten the Gauss distro? I've seen no reply to that. Research into chaos had shown that unlikely outcomes are far more likely than they should be. Eg. for reasons unknown, our Moon has far more large craters, as compared to smaller ones, than you could expect, basing it on standard distribution. That's the reality, not your futile and insulting lust for finding out how does my brain tick.

[Surrealistik]

You keep saying that vague irregularities flatten the curve assuming you get hit by a bullet but you've yet to really demonstrate that this is the case.

I separate this out into two clear components: Accuracy, the probability of actually hitting, and Outcome, the probability of what happens when you actually get hit.

Accuracy should probably have a bearing on Outcome (like accuracy above 100% granting a bonus to damage for example) as it makes more advantageous Outcomes more likely. Barring called shots/especial aim, you generally get hit in accordance with whatever features the most surface area, which typically means the chest/torso. Gaussian models this contingency better than flat because flat assumes a graze or a headshot or a footshot is equally as likely as a body shot; it is not. Where does the flattening come into effect? The adverse conditions that weigh upon Accuracy that the game doesn't quite account for (outside of injury)? Gaussian can be argued to factor those in as well. Yes, you might be tired/distracted/etc, causing your aim to go awry, but ultimately a body shot is still much more likely than a headshot, and it is much better to represent these factors, if you must, by impacting Accuracy itself rather than distorting Outcome in such an untenable way and calling it a representative abstraction as opposed to a verisimilitude defying horror.

[HelmetHair]

Removed for conscious flamebaiting - Dioxine

[hellrazor]

I actually only wanted to know what kind of purpose Heavy Machine Guns could have in openxcom, but ok i think someone is not understanding the difference between a mod and the game enginge.
Regarding this Accuracy OpenXcom

Regarding Damage look here
I do not understand the argument you both have, since in openycom we have clear explanations and well known formulas.
So back to my original idea how is the tactical purpose and usage of hmg? And coulf a implementatiin look??

[Surrealistik]

I actually only wanted to know what kind of purpose Heavy Machine Guns could have in openxcom, but ok i think someone is not understanding the difference between a mod and the game enginge.
Regarding this Accuracy OpenXcom

Regarding Damage look here
I do not understand the argument you both have, since in openycom we have clear explanations and well known formulas.
So back to my original idea how is the tactical purpose and usage of hmg? And coulf a implementatiin look??

Our argument is over whether a gaussian multiplier for damage is more representative of RL than the typical 0-2x flat modifier; we both understand how accuracy and damage are calculated by the engine.

[hellrazor]

A Gaussian curve ... istick with random.

[Surrealistik]

They're both random.

[ohartenstein23]

I don't think you can safely make the assumption that a hit to the body is the most likely outcome, considering that the game models the hitbox as a cylinder; 100% accuracy is hitting within the target's tile/cylinder.  A Gaussian distribution would be great if you knew exactly where you hit on the target, as a given projectile and velocity does have an expected amount of tissue damage it will inflict.  But then we have to make a distinction of what you hit, as you'd expect different amounts of damage for multiple body parts.  To account for that, we could make a damage distribution for each body part.  The result is that any given bullet falling within the cylinder of the hitbox will fall on one of these distributions - if for simplicity's sake we want to make only one roll for damage, the amount done will have a probability distribution that looks like a weighted sum over many Gaussian distributions, each with a different mean and standard deviation.  The final distribution is flatter than a Gaussian, and while it may be biased toward the body shot result, the width of the distribution would look much more like the flat case.

The point is, short of coding in better hit detection - body vs limbs vs head, influenced by firing accuracy - or adding in some method of critical hit/miss, a single Gaussian alone is a gross oversimplification based on a very particular set of assumptions.  Flat is a great simplification too, but Gaussian is too elegant, clean, and symmetric for something as messy as RL combat.

Now if you average over many flat damage rolls, that does start to look more Gaussian - this was more the route gone by the reboots; you never just fire a single shot from that assault rifle, and you can probably assume your sniper is hitting center of mass.  Critical hits and kill shots are even animated as hitting the head.

Edit: @Surrealistik  The likely reason you're getting so much flak over the suggesting switching the mod to Gaussian damage model is that it is balanced for the flat distribution, and making the switch means each weapon has to be looked at compared to armor values, and one or both would have to be changed, for some ~300 weapons/ammunition types and ~50+ armors.  This is a massive amount of work for Dioxine, who'd probably rather spend the time implementing new features.  The other option would be a change to the executable such that no rebalance is necessary - adding critical hits/misses, some manner of firing accuracy influencing where you hit on target, or implementing a Gaussian distribution with added probability in the tails to account for headshots/deflections, which would be more work for Meridian, who is busy with coding in other new features.  If you really want Gaussian distributions to be the norm, I'd suggest trying to mod in the balance yourself, or make a fork of OXCE+ to test out how to make it a better model.

[HelmetHair]

Sol,

I'd like over penetration. It could be a great mechanic for some weapons etc... as well as an explanation as why sometimes bullets don't kill... they simply wound and over penetrate.

This happens quite a bit with rifle rounds to extremities like 7.62x39 where it zips on through without really doing structural damage, but a shit load of soft tissue damage.

Also, 7.62x39 goes through bones creating a butterfly break. Look it up its pretty neat in a medical sense.

Not to mention terrain definitions could be wonderful. Tin building... awful cover. Miss a shot?  Well it could kill a cookie.

Plasma weapons equal huge penetration values I would think. Considering a 200 amp plasma cutter can blow through a 7/8 inch(13ish mm) of steel like butter.

Thoughts?

[Arthanor]

the amount done will have a probability distribution that looks like a weighted sum over many Gaussian distributions, each with a different mean and standard deviation.  The final distribution is flatter than a Gaussian, and while it may be biased toward the body shot result, the width of the distribution would look much more like the flat case.

[...]

Now if you average over many flat damage rolls, that does start to look more Gaussian

That is a great way of putting both why flat makes more sense for single shots and why autoshots are more reliable. Nicely done I could see sniper rifles being given a gaussian damage distribution on single hits, to show that the sniper is more likely to hit significant parts (do average damage), but that also reduces the odds of a sniper doing critical damage (odds of a 200% damage roll are smaller with gaussian than with flat). Presumably, something like a switched log-normal distribution (peaking on the high-side instead of low-side) would be best for snipers. For other weapons that are not as accurate as sniper fire, flat is fine.

@HelmetHair:
Over penetration would be cool, especially shooting through some obstacles/thin walls. I'm not sure about plasma though. As a gas of super heated particle, a plasma bolt has little momentum, it just has a LOT of energy.  Because it lacks the momentum to punch through and instead delivers 100% of the energy on the first thing hit (that's why cutters go through steel, they apply the energy locally and melt the steel. Firing a bit of plasma would melt whatever it touches, make a big whole that allows you to see through, but it wouldn't go through). It even makes sense as a weapon for genetically engineered super soldiers: They hit what they mean to hit, so they need a weapon that deals the most damage to what is hit, no wasting by overpenetration.

To me, Gauss would be the best penetrator (since its energy is purely kinetic, no explosive/heat), whereas plasma would be the best first hit damage. Regular ammo and laser would be the lower tech pair of penetration vs first hit damage.

[Dioxine]

Edit: @Surrealistik  The likely reason you're getting so much flak over the suggesting switching the mod to Gaussian damage model is that it is balanced for the flat distribution, and making the switch means each weapon has to be looked at compared to armor values, and one or both would have to be changed, for some ~300 weapons/ammunition types and ~50+ armors. 

No, that's not it. Comparing it to the work I've done on the mod already? Nothing, nothing at all. He's getting flak for the tendency to treat people like they were his personal bitches.
The reasons - well, your extended explanation covers the accuracy aspect in more detail than mine did. But it basically amounts to what I said. Gaussian distro ignores the reality of imperfect accuracy model, a reality better accounted for by a flat distribution model.

@Helmethair: you're conveniently forgetting that plasma cutters have a range counted in centimetres at best. The issue of plasma WEAPON overpenetration would largely depend on how it manages to deliver plasma to the target without discharging it in all directions just off the barrel's end.

@Hellrazor: the purpose of HMG is multiple, accurate, high-damage fire, at the price of having to deploy it in the place a turn before. It is naturally in most ways inferior to rocket launcher, as it was designed as a lower-tier weapon.

[Arthanor]

No, that's not it. He's getting flak for the tendency to treat people like they were his personal bitches.
The reasons - well, your extended explanation covers the accuracy aspect in more detail than mine did. But it basically amounts to what I said. Gaussian distro ignores the reality of imperfect accuracy model, a reality better accounted for by a flat distribution model.

lol When it's not calling me stupid/lazy (which luckily doesn't happen too often), your honesty is quite entertaining.

@Helmethair: you're conveniently forgetting that plasma cutters have a range counted in centimetres at best. The issue of plasma WEAPON overpenetration would largely depend on how it manages to deliver plasma to the target without discharging it in all directions just off the barrel's end.

Plasma containment fields, of course! Whatever that is...

And/or ways to spin the plasma such that it forms a self-containing magnetic field. Or maybe the aliens have found a way to use dark matter/energy to bind the plasma together (ex.: you fire a bullet of dark matter at the enemy, which in itself does nothing but make them feel queasy because of the gravity disturbance, but said gravity is enough to keep the plasma together. In reality, doesn't work since gravity wouldn't be able to fight the pressure unless it's approaching black hole density, at which point you might as well make a "micro black hole" gun! Indeed, plasma sounds dangerous until you realize how volatile it is, but it's part of the cannon so we gotta find a way to make up some techno-babble that makes it sound good enough to non-physicists.

@Hellrazor: the purpose of HMG is multiple, accurate, high-damage fire, at the price of having to deploy it in the place a turn before. It is naturally in most ways inferior to rocket launcher, as it was designed as a lower-tier weapon.

HMGs also have the huge advantage of not blowing up loot/corpses/civilians unless a stray shot hits them (which is relatively unlikely, compared to the assured destruction of being near a rocket launcher blast). With high accuracy soldiers, it is a much safer alternative to a rocket. And it is also cheaper per shot. (until you get destroyer outfit and spam big mind fueled 'splosions!)

[Solarius Scorch]

Can't we just have Gaussian distribution as a mod (extremely easy to do) and let people play as they like?
I really don't understand this little private war. This is not how it's supposed to look like, it doesn't lead to understanding or problem resolution. *stares disapprovingly in Polish*