|
Post by roey on Nov 23, 2011 3:43:09 GMT -5
Both Den Kirson's work on many presentations of game's weapons metrics and Symthic's recent BF3 chart are superlative work at concisely presenting a lot of data in a useful way for making interesting comparisons. One thing I feel might help differentiate having two presentations of BF3 weapon metrics, and help keep both useful would be to offer graphic representations of weapon accuracy, as this is one of the more difficult aspects of firearms metrics for people to visualize comparatively from raw data. As there are a lot of things that impinge on weapon accuracy in BF3 it may be more productive to only look at idealized native accuracy of each weapon in it's best configuration and ideal fire position: supported fire from a bipod (if that weapon fits one), furniture mods, heavy barrels etc. -- anything that improves the best results for a particular weapon. This is one area where Den and Symthic could agree to differentiate their charts if both wanted to present this kind of data graphically; one could do native weapon performance, and another could do some sort of mean or weighted weapon accuracy graph based on the weapon handling metrics. There are a lot of simple graphic ways to present accuracy data that can be visually simple, appealing and informative -- and this is another place Den and Symthic could differentiate their charts. Simple ring graphs with three rings representing the weapons spread at three different and meaningful range, an accuracy vs distance x/y graph, or various range 'man targets' with 'hits at range' scatter graphs. I realize I'm suggesting a lot of work here for someone else, but it's only that, just a suggestion, and I won't be disappointed in the least if no one runs with it -- as I'm ever so grateful for all the effort that's already been made! Thank you Den Kireson and Symthic!
|
|
|
Post by bel on Nov 23, 2011 8:02:17 GMT -5
I hadn't actually realised Den had started to put a chart together.
But at the very least, the biggest issue with an idea like this is that we're not there yet. We (or at least I, perhaps Den is more clued in that I am) don't really know how the recoil and spread values translate to the game. I'm collecting this stuff in my (now quite outdated) recoil thread. Basically, the spread, amp and recovery values don't accurately describe what we see in the game. The values we can see are either lying, I'm interpreting them wrong, or there are other variables that modify them that aren't on the chart.
I've found a couple of extra vars (again, in the later posts in that thread) but even those don't give the whole story.
tldr; there's no point talking about accuracy charts until we actually know how accuracy works.
Also, aptitude test failed, bro.
|
|
|
Post by roey on Nov 23, 2011 13:51:52 GMT -5
I hadn't actually realised Den had started to put a chart together. But at the very least, the biggest issue with an idea like this is that we're not there yet. We (or at least I, perhaps Den is more clued in that I am) don't really know how the recoil and spread values translate to the game. I'm collecting this stuff in my (now quite outdated) recoil thread. Basically, the spread, amp and recovery values don't accurately describe what we see in the game. The values we can see are either lying, I'm interpreting them wrong, or there are other variables that modify them that aren't on the chart. I've found a couple of extra vars (again, in the later posts in that thread) but even those don't give the whole story. tldr; there's no point talking about accuracy charts until we actually know how accuracy works. This is why I suggested native, idealised single-shot weapon accuracy, not shooter accuracy metrics; if qualified this is very useful. There are far too many vars impinging on shooter accuracy in BF3 to make a visually useful per weapon graphic representations; that's why I suggested seperating nataive weapon accuracy -- just like it's done for real firearms. Also, aptitude test failed, bro. Is this kind of petty trolling really necessery? I've failed no test, you haven't challenged my intellect or aptitude in anything -- you however have failed the civility test, "bro"...
|
|
|
Post by jackson on Nov 23, 2011 15:33:58 GMT -5
I think I am working on what you are talking about. I've made a function in MATLAB that uses the recoil and spread data for the guns to simulate large numbers of shots fired to create shot density plots. You can select any weapon, any attachment combos, burst length (how many shots you fire before letting off the trigger and allowing the gun to reset spread and recoil), and number of bursts. There some limitations though. I am guessing at the background mechanics of firing, recoil, and spread. -I have assumed that in firing a shot the game looks at the spread value and picks a random value less than or equal to this and a random angle within the cone for the shot vector (basically a random distance from the center and at a random rotational angle around the aiming axis, random polar coordinate) -I have assumed recoil picks a random value with the range of horizontal recoil values and kicks the center aiming point that much. Same with vertical recoil. -I have assumed the maximum spread is increased with every shot. -I have assumed that there is recovery of both recoil and spread between shots. I am also assuming the recoil and spread values values ripped are in degrees. This seems to be ok for the recoil and spread added values. However the recovery values cannot be in degrees per second. If they would, most weapons would recover to first shot accuracy between shots even on full auto. But we know that recoil and spread do play a role so there must be a conversion factor if my other mechanics assumptions are correct. Here is an example plot: AK-74M, no attachments, standing position, aiming down sights, while standing still, firing bursts of 6 rounds, 100 bursts Is this what you are looking for? This is raw data. I would run it through another function that would plot it on a color scale of shot density (black is low density, white is high density) which helps see the plot when there are so many overlapping points. But this one only has 600 data points.
|
|
tiesieman
True Bro
mental lagger
Posts: 1,401
|
Post by tiesieman on Nov 23, 2011 15:52:58 GMT -5
Is this kind of petty trolling really necessery? I've failed no test, you haven't challenged my intellect or aptitude in anything -- you however have failed the civility test It is a custom that newborn bros show their apptitude by ending their first thread with 'you are my bro, bro'. It is so stated in the rules, (it's just a lil' thing that shows you've read the rules; and bel probably wasn't attacking you)
|
|
|
Post by natsuterran on Nov 24, 2011 0:31:51 GMT -5
Is this kind of petty trolling really necessery? I've failed no test, you haven't challenged my intellect or aptitude in anything -- you however have failed the civility test It is a custom that newborn bros show their apptitude by ending their first thread with 'you are my bro, bro'. It is so stated in the rules, (it's just a lil' thing that shows you've read the rules; and bel probably wasn't attacking you) Which is really idiotic in my opinion. Who actually reads the rules before making threads? Especially for a forum that is this underpopulated. I've failed the test because I don't read the rules of forums. They are generally the same principles in all forums and I've been in enough and never had any issues before. You can't fail an aptitude test if you never received the test in the first place.
|
|
|
Post by roey on Nov 24, 2011 1:18:12 GMT -5
I think I am working on what you are talking about. I've made a function in MATLAB that uses the recoil and spread data for the guns to simulate large numbers of shots fired to create shot density plots. You can select any weapon, any attachment combos, burst length (how many shots you fire before letting off the trigger and allowing the gun to reset spread and recoil), and number of bursts. Cool! There some limitations though. I am guessing at the background mechanics of firing, recoil, and spread. -I have assumed that in firing a shot the game looks at the spread value and picks a random value less than or equal to this and a random angle within the cone for the shot vector (basically a random distance from the center and at a random rotational angle around the aiming axis, random polar coordinate) -I have assumed recoil picks a random value with the range of horizontal recoil values and kicks the center aiming point that much. Same with vertical recoil. -I have assumed the maximum spread is increased with every shot. -I have assumed that there is recovery of both recoil and spread between shots. I am also assuming the recoil and spread values values ripped are in degrees. This seems to be ok for the recoil and spread added values. However the recovery values cannot be in degrees per second. If they would, most weapons would recover to first shot accuracy between shots even on full auto. But we know that recoil and spread do play a role so there must be a conversion factor if my other mechanics assumptions are correct. Here is an example plot: AK-74M, no attachments, standing position, aiming down sights, while standing still, firing bursts of 6 rounds, 100 bursts Is this what you are looking for? Not exactly, but that's not a bad thing, especially so as you're the one with the tools and making the effort here! All your assumptions seem quite reasonable based on what's known about BF3 so far and BF games in general. There are two different approaches to the data that might yield more useful graphic representations for making comparisons I'd like to suggest for discussion and exploration: The first is along the lines of what I've already suggested, where only native, single shot weapon accuracy is graphed vs distance taking all weapon handling metrics and time out of the equation. While this has obvious limitations of removing an immense amount of data from how a weapon will typically perform 'in game', it does give the native accuracy of the best that might be achieved with a particular weapon if it's handled properly and not fired full-auto for getting an idea of absolute range of effective fire. The most useful mathematical representation I can think of would be to include all three dimensions of weapon performance (damage, accuracy, time/ROF) on an standardized target area (6 inches, or ~15 centimeters) at different ranges; this is the only objective way to show effective time/damage on a 'man target' for lethal effect. This latter is complicated by having to normalize the 'shooter accuracy' vars as some mean or average function based on the particular weapons handling constants, but once done and if the results look reasonable you have something very useful. Time has to either be represented as mean time to lethal effect, or as a percentage of lethal effect based on some time constant like 2 seconds of full-auto, or rapid semi-auto fire on a standard target. If all this has gone past anyone that has tried to stay with the thread this far; think of it this way; we take a standard target that's slightly smaller head size, not because we're evaluating head shots but because we want a standard target slightly smaller then any critical area we're going to be shooting at. We try to come up with a graphic representation of what each weapon does to this target when aimed dead center from a standardized shooting position, at standardized distance(s), firing full-auto or rapid semi-auto for a standard amount of time. The results can be quite telling when you have a very lethal round where most shots miss, and a less lethal round were most are on target and damage is critical... This kind of 'shooter' representation also makes for some good comparison to native weapon data in determining the kind of approach best to take with a particular weapon, at a given range, when confronting a threat. Is this kind of petty trolling really necessery? I've failed no test, you haven't challenged my intellect or aptitude in anything -- you however have failed the civility test It is a custom that newborn bros show their apptitude by ending their first thread with 'you are my bro, bro'. It is so stated in the rules, (it's just a lil' thing that shows you've read the rules; and bel probably wasn't attacking you) I'm sorry, I missed it in the Terms Of Service, which was the only thing I 'agreed' to when registring, and in fact still can't find anything about being anyone's 'bro'; no harm, no foul, and no hard feelings I hope. Edit: Aha, found it... So, per the 'Aptitude Test': " You are my bro, bro; dude bro, most bro dude of broness, like totally, bro, and like dude, like, don't be that way bro..."
|
|
|
Post by turdferguson on Nov 25, 2011 1:49:31 GMT -5
Which is really idiotic in my opinion. Who actually reads the rules before making threads? Especially for a forum that is this underpopulated. I've failed the test because I don't read the rules of forums. They are generally the same principles in all forums and I've been in enough and never had any issues before. You can't fail an aptitude test if you never received the test in the first place. There is really no reason to get upset here. If you feel this strongly about it, go gripe somewhere else. ...Here is an example plot: AK-74M, no attachments, standing position, aiming down sights, while standing still, firing bursts of 6 rounds, 100 bursts ... Very cool. I was wondering; are the axes labeled in degrees? I was thinking that it would be useful if the axes could be converted to meters. Of course the scale would be dependent on the range. The utility of doing this conversion is that you could superimpose these graphs onto screenshots of a soldier in game at the appropriate range. It's just a thought; I don't know anything about the program you are using to generate these plots. The conversions would be pretty simple math. I remember seeing pictures of soldiers at different ranges in the first google doc weapon charts.
|
|
|
Post by bel on Nov 25, 2011 1:58:54 GMT -5
Very cool. I was wondering; are the axes labeled in degrees? I was thinking that it would be useful if the axes could be converted to meters. Of course the scale would be dependent on the range. The utility of doing this conversion is that you could superimpose these graphs onto screenshots of a soldier in game at the appropriate range. It's just a thought; I don't know anything about the program you are using to generate these plots. The conversions would be pretty simple math. I remember seeing pictures of soldiers at different ranges in the first google doc weapon charts. You don't really need to do that, though. The plot would obviously be the same, so you may as well just superimpose it as-is with a single value in m for how far away the guy is. Great chart idea. x axis in m or degrees doesn't really matter, because it doesn't make things any easier in-game to have it in meters. You care how big the guy is relative to your spread, which you get from the superimposition regardless of what's on x.
|
|
|
Post by turdferguson on Nov 25, 2011 2:45:27 GMT -5
Very cool. I was wondering; are the axes labeled in degrees? I was thinking that it would be useful if the axes could be converted to meters. Of course the scale would be dependent on the range. The utility of doing this conversion is that you could superimpose these graphs onto screenshots of a soldier in game at the appropriate range. It's just a thought; I don't know anything about the program you are using to generate these plots. The conversions would be pretty simple math. I remember seeing pictures of soldiers at different ranges in the first google doc weapon charts. You don't really need to do that, though. The plot would obviously be the same, so you may as well just superimpose it as-is with a single value in m for how far away the guy is. Great chart idea. Yep, you are correct. Mental lapse there. It isn't necessary. I suggested the meter scale because it is more intuitive to have all three axes in a common unit (the z axis extends toward the soldier next to an MCOM that the game marks with a measurement in meters). That way you could determine e.g. that for a certain gun that x% of bullets will land within a 2m square centered on the target. That would be a useful metric for weapon accuracy. I suppose you could use that to calculate more realistic TTK values and to graph TTK vs range for each weapon. Again, these are just suggestions. Keep in mind all the assumptions we are making and their pertinence outside of our model.
|
|
|
Post by jackson on Nov 28, 2011 12:58:26 GMT -5
The axes are nondimensionalized by range. Multiply the distance to target by the number on the axes by range to get the deviation. If you are shooting at a target 100m away and the shot is at .02 then it would miss by 2m.
|
|
neverlast74
True Bro
Clan:DarkBadEvil Fightaz of DeathCommando - Squad: Elite Ninjas
Posts: 236
|
Post by neverlast74 on Nov 29, 2011 10:19:35 GMT -5
this picture is great - and if we had them from all/other guns we could start comparing...
one question jackson - do you see any way we could indicate visually which are the "early" bullets and which are the "late" bullets?
...because if the spread adds hitmarks only on the outer area, then this means "learn to counter the spraying" on PC. if the hitmarks jump around all over the ... this means we have to burst.
|
|
|
Post by bmwtx on Nov 29, 2011 11:16:24 GMT -5
This thread is awesome, I look forward to more comparisons whenever folks are courteous enough to provide additional pics using the same method.
|
|
|
Post by rudybojangles on Nov 29, 2011 13:12:42 GMT -5
|
|
|
Post by jackson on Nov 29, 2011 14:17:32 GMT -5
I've updated my code a bit to reflect recent findings on how recoil and spread behave. Changes: -First shot recoil modifier -Vertical recoil is no longer random -Reduced recoil recovery while firing to reflect the 1-2 degrees/second approximation -No spread recovery while firing -Colored points based on shot number 1. Blue 2. Green 3. Red 4. Cyan 5. Magenta 6. Yellow 7. Black (Then repeat order) Without the exact code, this can only be taken as a very VERY loose guide. I don't really know if any conclusions can be drawn from this yet. There are still too many assumption for me to put any confidence in the plots. The only thing I would put some faith in is the first shot accuracy since recoil and recovery haven't come into effect yet. There is an album of plots of all automatic primary weapons. No attachments. Standing position. Aiming down the sights. Stationary. Firing a 4 shot (or 3 shot for the burst-only guns) burst. 200 trials. Axes are unitless. Multiply by the range to target to get the x-y deviation of the shot in whatever units of range you want to use (meters, yards, feet, AU, lightyears, arshins, whatever). s80.photobucket.com/albums/j166/PvtDBJackson/BF3%20Accuracy
|
|
sleep
True Bro
Posts: 10,189
|
Post by sleep on Nov 29, 2011 15:20:26 GMT -5
wow, great work!
|
|
Lexapro
True Bro
PSN: Lexa_pro
Posts: 1,066
|
Post by Lexapro on Nov 29, 2011 15:59:16 GMT -5
Can we get some labels for which gun each graph corresponds to?
|
|
sleep
True Bro
Posts: 10,189
|
Post by sleep on Nov 29, 2011 16:40:58 GMT -5
labels would be cool, but in the meantime you can see it in your browser's status bar when you hover over each image, or in the title after you've clicked on it.
|
|
|
Post by doctorbrain on Nov 29, 2011 16:57:12 GMT -5
Awesome, thanks for this. Didn't realize how accurate the AS VAL could be.
|
|
|
Post by jackson on Nov 29, 2011 23:12:43 GMT -5
Awesome, thanks for this. Didn't realize how accurate the AS VAL could be. Yup. It is the only automatic weapon with a minimum spread of 0. Your first shot will go directly where you want it. Labels and such are coming. Might be able to get something hosted here. Photobucket is less than ideal.
|
|
|
Post by ZeroKelvin^ on Nov 30, 2011 9:05:36 GMT -5
schweet info, it would be cool if we got the effect of attachments of one weapon
|
|
neverlast74
True Bro
Clan:DarkBadEvil Fightaz of DeathCommando - Squad: Elite Ninjas
Posts: 236
|
Post by neverlast74 on Nov 30, 2011 12:06:35 GMT -5
great pictures - looking at it, from a decision perspective; which weapon should I use as an "class X" player: (assuming - attachments do not change that - which I am not completely sure)
Assault:: I think as Assault m416, m16A3, are good - and if you can live with 3 round bursts F2002 as this is also very accurate with grip.
Support: tough decision - probably the RPK74 if your plan is shooting a lot without bipod, with bipod proned I would even go for the PKP as it has the lowest max recoil of all 3shot LMGS (m60/m240B/PKP) 1.0 vs. 3.0 which can not be seen in the test as this was done standing. however if you plan shooting mid range single tab/dualburst for suppression - it might be a different weapon better. (namely 240b lowest recoul on first shot)
Engineer: also tough decision - scar? - have no experience with these weapons./dont play engineer so I do not know what to look for...
|
|
sleep
True Bro
Posts: 10,189
|
Post by sleep on Nov 30, 2011 12:41:51 GMT -5
with bipod proned I would even go for the PKP as it has the lowest max recoil of all 3shot LMGS (m60/m240B/PKP) 1.0 vs. 3.0 which can not be seen in the test as this was done standing. however if you plan shooting mid range single tab/dualburst for suppression - it might be a different weapon better. (namely 240b lowest recoul on first shot) good eye, i didn't notice that lower prone ADS max spread for the PKP. the PKP is also the only one of those three with a working silencer
|
|
|
Post by joeskeppi on Nov 30, 2011 13:59:25 GMT -5
I had to register just so I could say good work on those images, assuming they're evn remotely accurate that's some impressive work.
The best thing to do, would be to create a small app where you could select the gun and attachments, and then generate the image of the spread. Not sure how easy that is, I only know a bit of c# focused on web stuff.
|
|
|
Post by jackson on Nov 30, 2011 15:31:35 GMT -5
I had to register just so I could say good work on those images, assuming they're evn remotely accurate that's some impressive work. The best thing to do, would be to create a small app where you could select the gun and attachments, and then generate the image of the spread. Not sure how easy that is, I only know a bit of c# focused on web stuff. Thanks. Without some in-game screenshots of bullet impacts, I can't say how accurate these are. Maybe a PC bro could help out with that? I'm on PS3 which means no private servers and no easy way to get screen captures. If someone could fire at a flat wall from some distance and get some pictures that would rock. Probably easiest to use the M16A4 or another burst weapon. Fire a burst without compensating for recoil. Let the sights settle. Repeat ad nauseam. I'm just a mechanical engineer with some extra time at work and access to MATLAB. I wouldn't even know where to begin trying to make an app. All I know about MATLAB is from data analysis in college. Though the actual firing code is like 40 lines. The biggest part is selecting weapons, assigning weapon and attachment parameters.
|
|
|
Post by cthuzael on Dec 1, 2011 3:59:39 GMT -5
thanks for the great work, jackson. i really enjoy those files. however, having a different colorcode would make it even easier to estimate recoil.
i think since one usally uses burst fire instead of full auto, a representation with a 4 bullet color-code would be by far easier to read?
btw. first post: You are my bro, bro!
|
|
|
Post by ecchi on Dec 1, 2011 7:34:03 GMT -5
Love what you're doing here! Can't wait to see how much accessories affect the spread.
You are my bro, bro.
|
|
neverlast74
True Bro
Clan:DarkBadEvil Fightaz of DeathCommando - Squad: Elite Ninjas
Posts: 236
|
Post by neverlast74 on Dec 1, 2011 11:21:35 GMT -5
I am an PC bro - but not knowledgeable I do not know how to setup a local server - is that done in the console ? do I need some special software - but this might be possible. But the question I have jackson - all guns with all combinations needs to be tested? I do not have them yet unlocked. only some of them. 20guns x 5 attachements x 3 postions x 3 firemodes x 3 bipod/grip/nothing = a lot of testing... what exactly are you looking for?
|
|
|
Post by jackson on Dec 1, 2011 12:46:02 GMT -5
For now literally just one gun, no attachments. Just so I have some baseline. Any gun with a burst mode would be best so you could do consistent bursts.
|
|
|
Post by rudybojangles on Dec 1, 2011 12:59:25 GMT -5
Gah...
AS VAL is a monster. You know exactly where that first shot is gonna be, and most likely were the second and third will be too.
Great Great pics bro.
|
|