Ogre posted 11-27-98 05:32 PM ET (US)        

Hiya all,

I've been playing around with some gunnery/ballistics calculations lately in an effort to understand the real-life WWII gunnery environment (mostly in the areas of bullet scatter and energy loss). I've also contrasted this with how I think the Warbirds model works. If you are interested, I've put together a rough page that details some of these results.

http://www.cris.com/~reaper/gunnery/gunnery.html

-Ogre

ik Jagdgeschwader 77 posted 11-27-98 06:11 PM ET (US)            

Wow ogre, i'm VERY impressed! I'm not going to say the results here cause I want everyone to read them for themselves. Another thing you may want to adress in your study is how warbirds shots are 3 feet in diameter, and not bullets at all! Again, wtg ogre, your page is first rate.

Tchüß,

ik

sick posted 11-27-98 07:09 PM ET (US)            

excellent piece of work, ogre. i did some similar calculations (although nowhere *near* as detailed or thorough!), with exactly the same conclusions.

for reference, i did my calculations in a range of 10 to 20 mils (i cant find my copy of shaw right now, but i seem to recall he says 10 mils is prety much the theoretical limit for modern firearm accuracy). 10 mils is about 0.6 degrees deviation! that means its certainly within reason that your work is on the conservative side. 20 mils is nearly 1.2 degrees of variation; at that level the results would be almost boggling.

thanks for the detailed work, ogre.

sick

CO 487th FS

Bombom posted 11-27-98 08:23 PM ET (US)            

WTG Ogre! This verifies everything we've felt in our bones. Devastating information indeed.

As an aside, I really don't understand the mindset of someone who opts for the d10 shot. It is SO much more FUN to get in CLOSE and see parts fly off, and to really enjoy the detail and lighting while blasting 'um.

BB is out, take cover

Kun JG54 posted 11-27-98 09:19 PM ET (US)            

Whoa!!

Way to go squaddie!! I'm impressed, this is clearly the authority. I have dificulty observing any future gunnery arguments without referencing to this page.

Kun

Hooligan posted 11-27-98 11:05 PM ET (US)            

Although I believe your model is valid and I respect the work you did and the results you posted I have to say I am extremely skeptical of the range of offset angles you used. The basic value of +/- 0.5 degrees seems high to me for the following reasons. At the closest range of 50 yards convergence/target distance, your program shows that all of the bullets from a single gun will end up impacting in a "target group" about 3 feet in diameter. I have a hard time believing that a WWII fighter aircraft was this horrible a gun platform. I can personally equal or exceed that performance with a pistol, and frankly I am far from an excellent pistol shot. Much of the factors, such as wing movement, which should cause "gun movement" in the fighters under discussion would tend to move the entire gun (either up/down/back/sideways etc) as a whole rather than "tilt" the barrel. If wing flutter were to move the entire gun up 1 inch, at 200 yards that would only result in the bullet impacting the target at a point 1 inch higher, as opposed to the results of tilting the gun up half a degree which would move the bullet up about 5 feet at the target. To put this another way: Although .5 degrees seems "small", if in WB your aim at a target 300 yards distant is .5 degrees off either up or down then you will miss the target entirely. Clearly we adjust our aim by increments smaller than this. I have no hard data handy on this, but IMO .5 degree offset angle seems way too high. Somewhere there is some information in print which would tell us what kind of pattern they would get when they sighted and test fired guns at their convergence range and this would certainly be useful baseline data. I would be interested in seeing your charts for "at convergence" dispersion pattern charts for lower values (perhaps 0.05 and 0.1 degrees). Thanks for posting your stuff. It provides extremely interesting food for thought.

Ogre posted 11-28-98 12:41 AM ET (US)            

Hiya Hooligan,

Thanks for the feedback. Just a quick response to your comments. Like I said in my writeup, I figured that value would be the most discussed

WRT offset angle, I threw in a bunch of causes for why the rounds would scatter, some of which would produce a lateral movement. But I think the stongest would be recoil and buffeting effects that change the pitch and yaw angle which should both be angular effects.

For example a change in pitch that results in a 1.5" elevation or depression in the aircraft's nose would generate a 0.5 degree offset.

In the area of recoil effects, I've seen old footage of gun synch tests (alas no footage of the target tho) and when they did closeups on the barrels firing, they really oscillated. This leads me to believe that the barrel is basically cantilevered (or weakly supported) and should result in a change in the centerline angle relative to true.

Regarding your pistol analogy, I think it would be more applicable to say whether or not you could get a 3' diameter grouping from an automatic weapon while you are in a moving car on a *fairly* flat road (or perhaps a boat in still water). FWIW, I've seen 0.50 caliber weapons on steady mounts that generate this kind of grouping as well at pretty short ranges.

Just my opinion, good discussion.

-Ogre

Fletchman posted 11-28-98 11:23 AM ET (US)            

Nice work Ogre. Although the .5 degree offset is of course a bit arbitrary, I don't think that it is off by much. (Of course the offset would vary depending on how much/well the guns are mounted, buffeting and other factors.) I suppose there must be definite study out their on the dispersion of automatic weapons fire. From personnal experience I do know that the third round of a burst never ends up the same place as the first round! (:

The bottom line of course is that your tests are much more in line with actual pilot reports of WWII than the Warbirds Model is.

avin posted 11-28-98 12:00 PM ET (US)            

What are the units for the "RL versus WB section"? The y axis on that graph is:

overestimation (WB/RL hit probability).

Do you simply assume that in WB, 100% of the rounds will arrive on target for a perfectly aimed shot at convergence? So the value of 100 for an offset of 0.625 degrees at 3000 feet implies a hit probability of 0.01 in the model?

I'm still not sure I really understand the model. It would help to see the results for smaller offsets, as well as the case where the offset is zero.

It would also be interesting to see the results for the case where bullet diameter is 3 feet, no?

avin

Ogre posted 11-28-98 01:00 PM ET (US)            

Avin,

(is there anyway to respond to a post directly?)

Yes, you are correct wrt the WB/RL hit probability calculation. Essentually there are no units, it's simply 100%/1%= 100 for the 0.625/3000' case. I have a link on the main page to a seperate page that provides these probabilities for several offset angles (down to 0.25 degree). I'm waiting to see if this discussion will provide any new quantitative data to support a different offset value before running off and doing up a whole full up series like I did before. It's very time consuming

I plan to do a 0 offset case for the out-of-convergence situation. At convergence a perfectly aimed shot hits perfectly (so 100% across the board).

The large bullet size (if they are still using it) will affect the out-of-convergence situation as well (and be perfectly hitting within convergence). I imagine it will help primarily in the middle-ranges where many rounds might just miss by 1.5 feet. I'll try and get to that one too.

Thanks for your comments and suggestions.

-Ogre

Dave posted 11-28-98 04:09 PM ET (US)            

Hi Ogre,

Interesting stuff.

Your charts show that at 300 yards convergence, you'll get more hits by firing under the convergence range than at conv. By definition, convergence is the point in space where all bullets theoretically meet. It must be the point in space where the most hits will be scored. Since your chart doesn't show this, I've got to assume that there must be a flaw in your model.

I think your offset angle of +/- .5 degrees is too much. I'm not sure how long a machine gun is from the end of the barrel to the chamber, but let's say it's 4 feet. Your 1/2 degree has the centre line at the end of the barrel jumping around by about 13/32nds of an inch in all directions, making a circle 26/32nd in diameter. That in itself is too big IMO. However, you are also assuming that the bullets will exit the barrel at any of an infinite number of points within that circle, and evenly spread out. It seems to me that most would be concentrated at a point closer to the centre and become fewer and fewer towards the edge of the circle. Of these 2 points, I think the 2nd is likely more important than the first, and if adjusted would give the results at convergence that I'd expect.

Baal

avin posted 11-28-98 05:03 PM ET (US)            

Ogre,

In your previous post, you say that in this model, with zero degrees offset:

"At convergence a perfectly aimed shot hits perfectly (so 100% across the board)"

giving results identical to the WB gunnery model [that last bit is mine].

That tells me that the only source of scatter in the model is the offset. You don't model dispersion from any other cause. What this means is: assuming reasonable offset, your calculations are likely to be conservative. In RL, there's even more scatter, from things you haven't modelled.

The other thing you model that's not in the WB model is loss of energy with distance. This is where things like atmospheric viscosity, etc. that you mention come in. The only reason I mention this is to confirm that these are not sources of additional scatter in your model.

Have I understood this correctly? If so, your results are fascinating. Your estimate that long-range gunnery on WB is too generous is, if anything, conservative. Add the loss of energy with distance, which WB doesn't model, and the WB model is even more generous.

As soon as you decide on a reasonable and defensible offset, we'll have an answer. And I'm sorry for labouring what may be obvious, but this was really interesting, and I wanted to be sure I understood.

avin

itmo Lelv24 posted 11-28-98 07:59 PM ET (US)            

if we added this 0.5 or whatever degree factor to accuracy because of barrel chatter etc etc etc. the hit% would still be same..why? because of the 3ft bullets..lets make the bullets the size of bullets first and then see if we can hit anything=)

AND another thing we need is kinetic E modeled..or even a near approximation.

Ogre posted 11-28-98 08:11 PM ET (US)            

This one is for Baal,

Regarding your question about the higher hit percentage at 50 yards with a 300 yard convergence, than at 300 yards for a 300 yard convergence:

This origionally caught me off guard as well. The reason it occurs is because we have competing mechanisms at work here, namely convergence and the scatter phenomenon I'm trying to quantify in my study. Because of scatter, as the target gets further away, your rounds become more "dilute" thereby less effective (less rounds on target). Meanwhile, as you stated, convergence is trying to concentrate the rounds by bringing the centreline of the barrels (and the average position of the burst pattern) towards the same point. However, if the scatter effect outpaces or equals the convergence effect, the final result can be as you see in the charts.

For example, look at the bullet pattern for the 50 yard case (300 yard converge) and that for the 300 yard case (300 yard converge) on the "outside convergence" page. You can see that in the 50 yard case, the rounds are tightly concentrated because scatter is low. Also, they are in two widely seperated groups hitting each wing because we are well within convergence. Since I'm simply counting hits, this situation gets a good hit percentage. Meanwhile, at 300 yards, the two groups have come together (converged) but scatter is beginning to take hold and the rounds are more spread out b/c we are now 250 yards farther out. Hence, numerically, the 50 yard shot is more effective. In a way, it's a quirk, if you miss hitting those wings because you are slightly rolled, your effectiveness goes down for the 50 yard case and the profile will peak at 300 yards as you expect.

This effect could also occur even if we didn't have any scatter at all. We could have two point strikes on the wings at 50 yards and a single point strike on the tail at 300 yards. Numerically, both yield 100% hits.

As far as the "correct" value to use for the offset, I'm trying to figure that out as well. I'm not necessarily saying its all due to variations in the barrel centreline due to recoil though, but rather a combination of events that in the end produce this offset. For example, a 1 inch oscillation of the nose in the pitch and yaw directions will give you more than half the 0.5 degree offset, leaving a 5/32 variance in the position of the muzzle to take up the rest.

I don't want to come across as defending the 0.5 degree value to the death. It's a somewhat arbitrary value based somewhat on my real-world experiences with vibration and pointing systems. I'm hoping something quantitative will come of these discussions to flesh out a well defendable and historical value. At this point, it's the trends I find interesting.

-Ogre

Ogre posted 11-28-98 08:36 PM ET (US)            

Avin,

The only scattering mechanisms I'm assuming are those that affect the pointing direction of the gun barrels wrt the target. At present, I'm doing this indirectly, as the maximum "offset" angle I mention could be a combination of several things: for example, movement of the barrel exit during firing or changes in pitch/yaw due to non-pilot influences. Imagine your eye is the barrel exit looking at the center of the target, how would the target appear to move during firing? How the target centre moves relative to the centre of your view is what I call the maximum offset.

Once the round exits the barrel following a randomly determined angle between 0 and 0.5 degrees off the centreline, it is free to be acted on by drag and gravity, but is not affected by any other influence.

-Ogre

Hooligan BK posted 11-30-98 05:03 PM ET (US)            

First of all:

Ogre when I tried to email you directly it didn't work. Maybe you need to redo your profile.

I have some suggestions for improvements in your model. First instead of using a uniform random distribution between -offset and +offset use a Gaussian (normal) distribution. The run-time library for your compiler probably has a function you can use. If not you can create your own by taking N samples (5 or more would be a good number for N), summing them and dividing by N to get the final value. Since this kind of distribution can give you very small incidences of relatively large offsets, it would make sense to name your output based on the CEP (Circle of Equal Probablility): the circle where exactly half the results fall in the circle and half fall out. For example: your -.5 to +.5 degree offset has a CEP described by -.3535 to + .3535 degrees. That is, half of your random degree offsets fall within the .3535 degree offset circle. Actually a Radius of Equal Probability it probably what you want to categorize your results by. Another thing you can add in addition to angular differences is positional differences. i.e. represent the case where the wings flex upwards a couple of inches at the gun locations. You can add random small offsets in X,Y and Z for the initial conditions for the bullets.

cmos posted 11-30-98 06:50 PM ET (US)            

Ogre,

Excellent work.

My only comment is to echo Hooligan in that a gaussian distribution of offset angles is statistically more accurate.

If I read you correctly, the assumption that you are making now states that "any offset angle between 0 and X degrees is equally probable", which is *probably* not the case ... extreme offset angles are more likely to be less probable than low offset angles.

As Hooligan said, you should be able to code that fairly readily (Excel even has a macro for it!). Then, of course, we would get to argue about what an appropriate 1-sigma distribution is!

I agree with all of your comments about the state of WB and the gunnery model. Superb game, gunnery model could use some improvement. However, we might not want to toughen it up ALL the way to RL levels, many players might lose interest because kills would be so difficult.

Again, great work.

Regards,

cmos

Pale Horses Squadron

"and it's rider's name was death ..."

jedi posted 11-30-98 09:12 PM ET (US)            

Hehe well math-geek technobabble aside, I think the key here is that with almost ANY induced vibration (and I'm convinced that any large automatic weapon MUST have some significant inherent vibration), the rounds are going to scatter out quite a bit at any range outside about 400-500 yards, and severely limit the amount of damage you could inflict on a real airplane at long range.

Now contrast that with WB, where you can set your convergence at d8, and if you can figure out where to aim, you can readily achieve the same damage level with the same sort of bursts you use at "proper" firing distances. With a lot of ammo and tracers visible to d10, there's no reason why one couldn't become an excellent long-range marksman; yet almost every ace is quoted as saying that long-range shooting was foolish.

If the model doesn't match reality, it needs work, pure and simple.

--jedi

Wells posted 11-30-98 09:41 PM ET (US)            

How hard would it be to limit the firing distance to 500 yrds? Seems like it would be one of the simplest things that they (IMOL) could do. Simply change it from d10 to d5? It would certainly help a bit?

Kats posted 12-01-98 12:15 AM ET (US)            

CC wells, this is the joke about the matter. My probelm is that the difference in lethality between ranges is not very pronounced at all. I would just say make the 200 yrd shots much more lethal and the 800 yrd shots less lethal than they are now.

As far as programming and resources - side step all that mess and reduce long range lethality without modelling the why's. Leave the lazer shooting but make it less lethal.

suso posted 12-01-98 02:30 AM ET (US)            

Baal said:

Your charts show that at 300 yards convergence, you'll get more hits by firing under the convergence range than at conv. By definition, convergence is the point in space where all bullets theoretically meet. It must be the point in space where the most hits will be scored. Since your chart doesn't show this, I've got to assume that there must be a flaw in your model.

This is my response to Baal:

No, there is not necessarily a flaw in the model. Convergence is the point where all bullets theoretically meet ON THE AVERAGE. The bullets from a single gun actually fan out in a cone (a very skinny cone!) the central axis of which points at the convergence point. However, there will be a scatter of the bullet impacts from a single gun around the convergence point.

At a short distance (but outside convergence) the cones coming from two separate guns have a small diameter, but the central axes of the two cones are at different points. This produces scatter.

At a longer distance (but right at convergence), the central axes of the two cones intersect with each other. However, at this longer distance the diameters of the cones are bigger, and therefore you get a scatter AT the convergence distance. The optimum distance for minimum scatter will be some compromise distance which will be somewhat SHORTER than the convergence setting.

suso

BadMan 4thFG posted 12-01-98 12:30 PM ET (US)            

I'd just like to add my agreement with CMOS and Hooligan. The only thing that keeps this analysis from being near perfect is the lack of a probability curve in the offset angle.

In my limited experience in bench shooting, (shooting from a stable platform, with the rifle locked in clamps) the shot pattern shows a definite bias to the center.

Ram1 posted 12-01-98 12:54 PM ET (US)            

Ogre:

Nice work, just one question. If your gun convergence equaled the distance you shot then why is there a dispersion pattern? I thought if you set convergence to 800 then all rounds will converge at that point in space. Now if convergence was fixed, say at 300 and you shot out to 800 then you would get dispersion causing many rounds to miss their target.

What did I miss?

Daff RSAF posted 12-01-98 01:02 PM ET (US)            

Raml: That he added a slight dispersion due to buffeting, engine vibration, etc.

Daff

bcnu posted 12-01-98 02:06 PM ET (US)            

Up until recently I've been a critic of the "long range gunnery flaw" theory, attributing most such incidents to lag.

For instance, a SpitXIV hit me at (my perceived) d7 -- and I wasn't flying straight-and-level. Hmm, said I, until my wingman said that the Spit in question was very warpy.

Well, yesterday I'm in my Spit I doing the Eagle Squadrons thing with conv set at 180 and chasing a 109 at d4 ... looked about mid-d4, so 450 yards.

Just for a fluke, and hoping a stray ping might make the 109 break, I lined up on this steady target and let loose a moderate burst.

To my great surprise, there were hits on both wings (remember this is over 2x my conv and the 109e has a 32ft wingspan) and to my greater surprise the 109's engine started losing glycol. A further short burst caused an explosion, but whether this was from a pilot hit or something else I couldn't tell.

Without getting into the details of a 0.5 degree deviation vs. 0.6 or 1.0 or 0.2 I'd say Ogre's results are spot on, provided one also take into account the host setting for round size...

Bino posted 12-01-98 02:13 PM ET (US)            

...If the model doesn't match reality, it needs work, pure and simple...

Right on Jedi! Cut to the chase!

And bravo, Ogre! I knew there was a reason that Fletch made you Geschwader Adjutant!

Ogre posted 12-01-98 06:13 PM ET (US)            

Hooligan said:

I have some suggestions for improvements in your model. First instead of using a uniform random distribution between -offset and +offset use a Gaussian (normal) distribution. The run-time library for your compiler probably has a function you can use. If not you can create your own by taking N samples (5 or more would be a good number for N), summing them and dividing by N to get the final value. Since this kind of distribution can give you very small incidences of relatively large offsets, it would make sense to name your output based on the CEP (Circle of Equal Probablility): the circle where exactly half the results fall in the circle and half fall out. For example: your -.5 to +.5 degree offset has a CEP described by -.3535 to + .3535 degrees. That is, half of your random degree offsets fall within the .3535 degree offset circle. Actually a Radius of Equal Probability it probably what you want to categorize your results by. Another thing you can add in addition to angular differences is position differences. i.e. represent the case where the wings flex upwards a couple of inches at the gun locations. You can add random small offsets in X,Y and Z for the initial conditions for the bullets.

CMOS said:

My only comment is to echo Hooligan in that a gaussian distribution of offset angles is statistically more accurate. If I read you correctly, the assumption that you are making now states that "any offset angle between 0 and X degrees is equally probable", which is *probably* not the case ... extreme offset angles are more likely to be less probable than low offset angles.

Yes, I agree completely with both of you FWIW, it wasn't my intention to keep the uniform distribution model but to include a centre-weighted one like you both suggest. In hindsight, I regret "releasing" the URL before I concidered it completed (got caught up in the wave of gunnery posts and then my job hit hard). I broke my cardinal rule that this study wouldn't contain anything that would make RL look worse than it is. Even more so, I failed to adequately explain that I was providing trends wrt to the hit probablities (I didn't think I was going to release it early . I've changed the writeup some as an intermediate step, until I finish my study of what distribution to use. I'm also making changes to my software to allow for non-uniform distributions (which is what stopped me from using one intitially). For the application (non-game) that the software was created, everything is uniform.

Anyway, on to the math

I was thinking of a "Gaussian-like" radial distribution similar to the behaviour that the free end of a cantilevered beam undergoes during random vibration. It is somewhat similar in amplitude at the centre to a the N=5 Gaussian that Hooligan mentions, but with a broader distribution at the ends (more peaked at dead centre). My reasoning (which I'd appreciate comment) is as follows. I think it is applicable for both muzzle chatter and pitch/yaw oscillations of the aircraft:

While the position traced out by the endpoint (muzzle or nose) will follow a pattern that is centre-weighted, there is also a question of the *time* spent at each position as well. At the terminus of the arc, the endpoint must slow, stop, and reverse. Therefore, it spends more time there than at the centreline. This throws *some* weighting away from the centre towards the outer reaches.

Hooligan, I haven't looked at it in depth yet, but I suspect the N=5 Guassian is extremely sharp towards the extremes. Also, my guess is that excursions beyond half the range are negligible. Therefore, while we get the "hump" in the centre we don't get the full range of motion either.

Interesting discussion. Please continue your comments. My goal is to make this as "correct" as possible.

I think we are getting close now.....we are now just haggling over the indices

-Ogre

Hooligan posted 12-01-98 07:29 PM ET (US)            

Hehehehehe After you update your model then we can argue about the standard deviation values you use. This will be a very fertile source of passionate and unsupported-by-real-data arguments

GunsGG posted 12-02-98 08:07 AM ET (US)            

The old marine corp manual for heavy machine gunnery would be a good sourse of information for you. It is based on the old water cooled .30 calibre locked down on a solid base. These guns were meant to be fired at ranges of 4-1200 yds, from a rear position, with the bullets arching over our marines at the front lines(usually in excess of 100 ft). Because the guns were locked, they could not be aimed , but instead left a "footprint" of many yards in width and length due to barrel vibration.

Hooligan posted 12-02-98 11:39 AM ET (US)            

A quick note about MG ammo for infantry. The powder load for MG rounds was purposely varied slightly to cause more dispersion in the rounds. These guns will certainly have lot more dispersion than aircraft mounted ones.

Ozymandias DoD posted 12-02-98 12:47 PM ET (US)            

A quick note about MG ammo for infantry. The powder load for MG rounds was purposely varied slightly to cause more dispersion in the rounds. These guns will certainly have lot more dispersion than aircraft mounted ones.

This is a new one to me. Got a source or better explanation? Did they make air-to-air quality ammo and "those bastards don't need to hit anything" ammo?

I think people are overlooking or not understanding the cone of fire concept / beaten zone. This basically gaurantees scatter at range. Don't think is anything to do with the ammo. Essentially, due to vibration and othber factors, at range you get an conical or egg shaped zone where the bullets impact.

At least, that's what I've always read, and my Ma seemed to bear it out for me.

-ozds- DoD

Hooligan posted 12-02-98 01:33 PM ET (US)            

I don't have the source handy. If you look into .50cal "sniper rifles" you will discover that they strongly advise against buying military surplus ammunition because the powder load varies slightly between rounds. The explanation for this is: Purposely tweaking the ammo for to create more random dispersion of fire from a MG firing at a platoon of infantry makes some sense. After all you don't really want to put 5 bullets into a single target (unlike aircraft), you want dispersed fire to hit a dispersed target. Whether or not this actually works as intended is another issue and this "feature" wasn't my idea. I'm just reporting it because I thought it had some bearing on the discussion .

LawnDart posted 12-02-98 02:24 PM ET (US)            

Hooligan wrote:

If you look into .50cal "sniper rifles" you will discover that they strongly advise against buying military surplus ammunition because the powder load varies slightly between rounds.

Of course "sniper rifles" wouldn't suggest off the shelf ammo. Even my Winchester M52 .22 match rifle isn't too accurate with the bullets from the $19.50 brick at Sports Authority -- and i'm shooting at 50 FEET indoors. The .50s that you are reading about were intended to generate very tight groupings, at range, in controlled semi-automatic (or bolt) firing speeds. The vibration from the barrel is elminated too.

ALL bullets that are produced en masse and not considered match grade have slight variations in powder loading, bullet shape, primers, even powder types! As a result, the only way you would get a very tight grouping is by pure luck.

FYI - a 800 yd match target is a 9 foot diameter circle -- and even then if you are off by as much as 1-2 clicks on your sight you may not even hit that. THAT is the goal of the match rifle (aka sniper rifle).

For an MG, you were going for a beaten zone and may even have charted out/boresighted in for that ahead of time <G>. The point is area coverage -- even for point targets. Now granted, the Ma Deuce made an excellent sniper weapon because of its single shot capability -- but even then most of the vibration would be gone (and it was tripod based vs bipod where even the heartbeat of the shooter must be taken into account).

Also of note for my diatribe here, everyone reads the stories about the .50 cal snipers etc in Nam and think Hollywood. Military snipers are area shooters. We aimed center mass of the target -- our goal was to hit them. Preferably kill, but chest wounds even if not fatal got the target out of action. The Hollywood sniper head shots, behind the ear crap -- that just didn't happen all that often. Law Enforcement snipers must hit those precise spots because of things like hostages and even problems with background (area behind the intended target). Why do I bring this up? Well, the .50 cal snipers using the M2 generally did not use match ammunition -- why? Because if they miss they can shoot again, and again, and again.... The typically talked about .50 shot from about 2000m was actually TWO shots. The first hit the kids bicycle, the second round hit the kid.

The modern military .50 snipers are using weapons like the McMillan or Barrett use match ammo because of the unique environments that they go into (like Somalia where background is/was important).

BTW, .50 was never my preferred. The M40 was/is much better than the M1987R (the .50 I used, not the infamous Barrett of modern video game fame like R6 and DF) simply because of weight. I couldn't stand lugging that big thing around. I could take the M40 out with twice the ammo though, thanks to the frequent workouts. Of course, the M40 couldn't do nice things like shoot through cars.