Black Ops Recoil Plots + MW3, COD4 links

[psijaka]

Thanks for the feedback.

Any suggestions for other recoil stats/plots?

I have chisen to plot 3 and 10 round bursts, as 3 round bursts are typically what would be used for longshotting, and that is what you get with the M16/G11 in any case; nd 10 round bursts give an indication of what happens if you open up at a medium range target on full auto.

I am planning to do a simulation of a 30 round burst; probably the Famas as it has such biased recoil, to see if there is a lot of difference between a 10 and a 30 round burst. I suspect that most of the bulets from rounds 11-30 will not stray much beyound the 10 round burst limits, but we will see.

[mannon]

Going up to 30 you're likely to start hitting the min and max viewkick limits, I believe. I don't know what those figures are for the weapons, though. I'm sure somebody has them.

[psijaka]

Maybe, and I don't have any idea what they would be.

I think that I'll stick to 10.

[mannon]

If you're on PC you could probably open up the files and get them yourself. I think there's a thread that tells you how. Of course we'd also have to make an assumption as to what the units of measurement are. Prolly degrees, or something known, though.

[psijaka]

Have added plots of the AK74 with various attachments.

I am not intending to do this for all combinations of gun and attachment (!) as the overall effect on recoil will be similar.

[psijaka]

@ mannon

Wii !

[mannon]

LOL nvm then. hehee

[psijaka]

Snipers and Secondaries now added to recoil chart.

Edit - I wasn't planning to tackle shotguns, as they are close range weapons with a wide pellet spread. Should I include them for completeness?

[MastaQ]

Feb 22, 2011 16:48:27 GMT -5 psijaka said:

Snipers and Secondaries now added to recoil chart.

Edit - I wasn't planning to tackle shotguns, as they are close range weapons with a wide pellet spread. Should I include them for completeness?

Can't see what good anyone could get out of a shotgun recoil plot, TBH. I think a shotgun recoil chart could be slightly misleading, if anything, since the chance of you hitting the target with X pellets is random, which wouldn't really make sense in the "% on target" column.

[mannon]

Actually I think most of the shotguns recenter all the way anyway, but it might be worth noting which ones don't. I think the ones that don't REALLY don't. Let me see...

Yeah the Stakeout (I just misspelled it steakout... man I could use a steak), and masterkey always recenter completely. The Olympia recenters horizontally, but has a nasty upkick that will always bee too high to recenter. The SPAS12 and HA10 sometimes recenter horizontally and sometimes don't, but always kick too high to recenter vertically.

I think it's at least worth plotting them just with a note that each dot would represent the center of the potential scatter pattern rather than a pellet. It would be more accurate to plot translucent circles instead, but then you would have to somehow size the circles correctly for the spread, and that's pretty much impossible ATM.

I don't really care if some people get confused about what it means. If you tell them what it actually represents and they don't listen it's their own fault. The data could give you an idea how spammable the shotguns are, though.

Note: I based all my observations above on the recoil stats and firetime alone. If there are more stats relevant to the time it takes to fire each shot on a shotgun then I did not figure them in, because I don't know them. ;p I'm just saying... Hopefully the firetimes listed here are all, though.

[psijaka]

I will not be doing any plots of semi automatic weapons and definately not shotguns; I just don't see the point, as we have no way of knowing when aa second shot would be fired, let alone a 3rd, and as for shotgun pellet spread...

[psijaka]

Have decided to add Shotguns to the Recoil Chart, as I think the informtion is useful.

That's all guns/attachments covered (let me know if I have missed any)

[psijaka]

Have added Assault Rifle ACOG recoil plots.

[tacit]

Informative charts, and cements a great deal of the thoughts I had about the weapons from the outset. It's interesting how people will, even without the hard data, gravitate towards the weapons that are [for their intents and purposes] better.

I do love the Skorpion, mostly because of accuracy and minimal weapon strength loss from silencing. It fits my play style well because I like to pick targets off at moderate range [especially when they don't know I'm there] rather than engaging them head-to-head in a battle of recoil, ping and hit detection luck.

In the whole FAMAS vs AUG thing, it's basically a matter of preference: do you prefer a weapon that's likely to kick in two directions, or three? I'll take the FAMAS any day [if I'm not using the M60 w/grip].

It's kind of disappointing to see so many weapons with the same kick and recoil - I like the idea of weapons with their own niche, because players tend to enjoy beating other players with weapons that aren't typically considered overpowered.

[psijaka]

Thanks for the feedback, tacit.

I agree with your point about a lot of guns having very similar recoil; treyarch really didn't show a lot of imagination here.

I generally prefer the guns with more random recoil, rather than the "upwards and to the right" biased guns. But why couldn't we have had a gun that went upwards and to the left? Or pulled upwards and only slightly to the right? A missed opportunity.

I too am a fan of the Skorpion; although I tend to use it unsilenced, to maximise the range that you get a 2HK. It's medium range performance really is remarkable, despite the low damage. With over 50% of shots being recentered, it is best to just aim and then hold down the trigger!

Pity that scavenger is essential, due to the low medium range damage and small clip size. I would love to use this gun with Lightweight or Ghost.

[tacit]

Re: the Skorpion with Ghost or Lightweight - it's really a question of being realistic about how long you last in the game, as to whether you really need Scavenger at all.

I'm not a great player, so I tend not to bother with Scavenger, and I also get a bit of a kick out of picking up someone's gun when I run out of ammo anyway.

The interesting thing about that is, out of the thousands of times that I've picked up someone's gun, maybe only a handful of times it had low ammo in it. Usually they're packed with over 100 spare rounds that never got used.

[psijaka]

I just hate having to hunt for a gun, and like to just concentrate on getting the kills without worrying about finding another gun, and then having to change my playing style mid game because I pick up (say) an RPK.

[arcanine2009]

Dumb question but... Shooting duel wielding weapons at the same time, increases recoil by exactly 2X, right? That seems logical to me, but I wonder whats behind the actual coding..

Oh and looking at the chart again, I'm wondering... Is grip on m14 even necessary? No grip is .286 recenter time, and w/ is like .25. Thats only like a .03 difference! And I usually shoot 2-4 bullets a sec, max. I don't think I ever went beyond 300 RPM on a regular basis. To do 600 RPM, you'd have to spam 10 bullets a sec, and that is insanely fast.

[n1gh7]

Mar 16, 2011 16:40:51 GMT -5 arcanine2009 said:

Dumb question but... Shooting duel wielding weapons at the same time, increases recoil by exactly 2X, right? That seems logical to me, but I wonder whats behind the actual coding..

actually, I think only one view kick is counted because testing shows that a shot after another will over take the original viewkick. That is assuming that duel wield guns are counts as one gun which I think it is.

[psijaka]

Mar 16, 2011 16:40:51 GMT -5 arcanine2009 said:

Dumb question but... Shooting duel wielding weapons at the same time, increases recoil by exactly 2X, right? That seems logical to me, but I wonder whats behind the actual coding..

Oh and looking at the chart again, I'm wondering... Is grip on m14 even necessary? No grip is .286 recenter time, and w/ is like .25. Thats only like a .03 difference! And I usually shoot 2-4 bullets a sec, max. I don't think I ever went beyond 300 RPM on a regular basis. To do 600 RPM, you'd have to spam 10 bullets a sec, and that is insanely fast.

I use M14 from time to time; without grip. I choose extended mags indtead, as on the Wii this gives you more total ammo.

I would suggest trying without. From what you say about the rate at which you fire; I would doubt that you will notice much difference.

On the subject of dual wields; as these are for hip firing at close range, and have poor hip accuracy, recoil is the least of your worries!

[psijaka]

Is Viewkick a Displacement or a Velocity?

My initial simulations assumed that viewkick was a displacement, and that centerspeed a velocity, working against the kick to recenter the gun.

However, it is more than likely that viewkick is a velocity and that centerspeed is applied as a deceleration to this, to recenter the gun.

I have simulated both, assuming that for the latter, if the gun is not recentered, any residual velocity left over from the previous round is cancelled and overwritten by the viewkick velocity from the new round.

The good news is that the shape of plots from both methods is the same!

BUT the displacement is greater for the velocity model for guns with longer fire times, and less for guns with shorter fire times, as displacement = velocity x fire time (minus the effects of recentering).

As the shape for the 2 methods is exactly the same, the tabulated probablilties that 2nd and subsequent shots will take place when the gun has recentered after the previous rounds are UNCHANGED.

HOWEVER, the recoil plots need to be modified.

Plots for the faster guns, eg fire time <0.08s (Famas, 30-20 SMGs, Stoner etc) should show less recoil spread.

Plots for longer fire times (M60) will show more spread. Please bear this in mind when making comparisons between the recoil spread plots.

Looks like I have a lot of work to do!

[jesterv01]

Sorry - put these in the wrong thread.

1 million rounds each in 10 round bursts, contours show cumulative %. It's the irregular shape of the high hit regions that's interesting. I think it's caused by the non-linear behaviour of the centering motion never over correcting.






Matlab code here

dl.dropbox.com/u/710852/guntest.m

[mannon]

On dual wielding (duel wielding would be cool too heh) it seems to me that it would be very dependent on when you pulled the triggers. I believe that what you are using is actually some strange hybrid that in some ways acts as two weapons, but fires both through the same hipfire cone. If you pulled the triggers at exactly the same time then there it seems likely that the viewkick for one would overwrite the other completely. You are more likely to be off at least a little. This would mean that one of the weapons would allow one weapon to kick just a little before the other weapon's kick negates it, provided the firetimes stay in sync. The more off you are the more additional jitter you would get.

It could actually be beneficial to fire them in such a way that they fire one then the other as this would be exactly like doubling the ROF. It would most likely throw the view from center more quickly, but could also put more bullets closer to center during the first burst. I wouldn't really recommend it given you cannot ADS and thus won't achieve any real accuracy anyway, though. It's probably just better to minimize recoil and fire them at the same time. Plus that puts more lead downrange faster for the optimal TTK. If you could ADS a dual wielded weapon, however, it might prove interesting to see how exactly certain weapons perform.

[psijaka]

fascinating plots jesterv01

one thing troubles me; the famas should always recover from leftwards or downwards kick before the next round is fired.

edit-I cannot open the .m file. What method do you use?

[mannon]

The .m file is source code for matlab. I don't really know anything about matlab, but I googled it. Sounds interesting, though. It looks like you need the matlab compiler or interpreter. I don't know if it is a compiled language or not.

You can just open the .m file in a text editor to see the code.


PS. Just looking at the code and wow! I assumed it would be pretty complicated. A lot of it is still greek just because I don't know matlab, but I'm surprised such a short bit of code produced those graphs. I may have to look into it.

Is um... matlab something you have to pay for? ;p

[mannon]

NVM... Yeah it's definitely something you have to pay for. ;-( They don't even give you a damn price, you have to contact them first. lahme...

I hate companies that do that. I always get paranoid that they're just going to give me the shaft and charge me a lot more than they usually do for one reason or another. I don't see a good reason why a company should keep their prices private like that... it just seems... shady.

Anyway I'm sure I can't afford it. Maybe once I'm back in school if they have a cheap enough student version... but I doubt it's very cheap at all. heh

[jesterv01]

OK - I'm still stuck. I've fudged the model so that it seems to compensate for ROF correctly - M60 / Spectre comparison seems right. Fudge is to scale the viewkick by the time between shots. But - this makes ROF *always* make guns less recoily - obviously wrong. Any ideas? (.m file is just text if you want to see the full model)

Some plots,













dl.dropbox.com/u/710852/NewPlots/guntest.m


Matlab is very powerful tool, but also rather pricey (~£2000 a license). To get similar functionality for free try the SciPy distribution of Python (www.scipy.org/)

[psijaka]

I am quite fascinated by your results, jester, especially the unusual anomolies away from the axes; I just don't know where these come from.

I have a few questions about your code, especially the bits highlighted in various colours below:

gundata = {'FAMAS',[60 -20 60 -10],1600,0.064
'AUG',[60 -30 60 -60],1500,0.064
'M60',[60 10 80 -80],1500,0.112
'Skorpion',[30 -20 50 -50],1500,0.08
'Spectre',[60 -20 60 -10],1600,0.064
'SpectreROF',[60 -20 60 -10],1600,0.048};

close all

for gx = [1:6]
name = gundata{gx,1};
kick = gundata{gx,2};
cs = gundata{gx,3};
ft = gundata{gx,4};

nburst = 100000;
nshot = 10;
posx=zeros(1,nburst*nshot);
posy=zeros(1,nburst*nshot);
for jx = 0:nburst-1
for ix = 2:nshot
kickx = kick(4)+rand*(kick(3)-kick(4));
kickx = kickx * ft / 0.112;
if kickx>0
kickx = max([0 (kickx - (cs * ft)/5)]);

elseif kickx<0
kickx = min([0 (kickx + (cs * ft)/5)]);
end
posx(ix+jx*nshot) = posx(ix+jx*nshot-1)+ kickx;

kicky = kick(2)+rand*(kick(1)-kick(2));
kicky = kicky * ft / 0.112;
if kicky>0
kicky = max([0 (kicky - (cs * ft)/5)]);
elseif kicky<0
kicky = min([0 (kicky + (cs * ft)/5)]);
end
posy(ix+jx*nshot) = posy(ix+jx*nshot-1)+kicky;
end
end
% figure()
% plot(posx,posy,'.')

nbin = 30;
z = zeros(2*nbin+1);
yoffset = -10;
binWidth = 5
for ix = 1:nburst*nshot
binx = median([1,2*nbin+1,round(posx(ix)/binWidth)+nbin+1]);
biny = median([1,2*nbin+1,round(posy(ix)/binWidth)+nbin+1+yoffset]);
z(biny,binx)=z(biny,binx)+1;
end

z=z./(nshot*nburst)*100;

[FF,I] = sort(z(,1,'descend');
FFsum = cumsum(FF);
YY = ones(length(FF),1);
YY(I) = FFsum;
YYY = reshape(YY,size(z));

% Plot the contours of the cumulaitve weight
xax = [(-nbin*binWidth):binWidth:(nbin*binWidth)];
yax = [((-nbin-yoffset)*binWidth):binWidth:((nbin-yoffset)*binWidth)];
figure
clf
pcolor(xax,yax,YYY)
hold on
contour_levels = [[10:10:90] 98];
contour_labels = [[10:20:90] 98];
[C,h] = contour(xax,yax,YYY,contour_levels,'k');
clabel(C,h,contour_labels,'labelspacing',500)
shading flat
axis equal
grid on
set(gca,'Layer','top')
set(gca,'YLim',yax([1 end]));
set(gca,'XLim',xax([1 end]));
colorbar

title(name)
xlabel('X Kick')
ylabel('Y Kick')
print('-dpng','-painters',strcat(name,'.png'),'-r80')
end


The pink statements I just don't understand the purpose of. Where does 0.112 come into the equation? This is the fire time for the M60 (one of my favourites), but what relevance does this have to the Famas, Skorpion etc?

The orange statements look as if they let the gun cross the axes if the kick is less than recenterspeed/5. Am I missing something here?


Rather pricey !!!!?

I'm afraid at £2000 I would expect a car, rather than a bit of software, so zero chance of me or I imagine any other Bros shelling out for matlab. May take a look at the free offering though.

[jesterv01]

The pink bit is my fudge for correcting for rate of fire. It is entirely empirical and basically just weights the viewkick by the intershot period, normalised to the M60. The justification for this is that if the viewkick figures are speeds then to convert to a displacement they should be multipled by a time, and the intershot period would seem to be the right time to use.

The orange statements don't let the gun cross the axis - thats what the max() and min() are for. The process is:

Work out kickx
if positive choose the bigger of 0 or (kickx-cs*ft)/5 This stops the centering over correcting - if kickx is negative after the centering has been applied 0 is return.
Add kickx to position of the last shot.

And, whilst writing that I've realised it's wrong... Centering doesn't work relative to the last shot but to the aim point...

Need to fix - hurrah for peer review!

I'm doing this during my lunch hour at work - it actually looks rather like what I should be doing. Wouldn't recommend investing in a copy of Matlab just for this....

Thanks! - will post back soon.

[psijaka]

Mar 23, 2011 17:11:56 GMT -5 jesterv01 said:

The pink bit is my fudge for correcting for rate of fire. It is entirely empirical and basically just weights the viewkick by the intershot period, normalised to the M60. The justification for this is that if the viewkick figures are speeds then to convert to a displacement they should be multipled by a time, and the intershot period would seem to be the right time to use.

The orange statements don't let the gun cross the axis - thats what the max() and min() are for. The process is:

Work out kickx
if positive choose the bigger of 0 or (kickx-cs*ft)/5 This stops the centering over correcting - if kickx is negative after the centering has been applied 0 is return.
Add kickx to position of the last shot.

And, whilst writing that I've realised it's wrong... Centering doesn't work relative to the last shot but to the aim point...

Need to fix - hurrah for peer review!

I'm doing this during my lunch hour at work - it actually looks rather like what I should be doing. Wouldn't recommend investing in a copy of Matlab just for this....

Thanks! - will post back soon.

Sorry, missed the significance of the max/min functions there; I see what you are doing in that respect.

I look forward to seeing your new plots.

I would recommend that you stick to one gun with and without a range of attachments, so we can see the effect of varying rate of fire and centerspeed. How about Spectre with and without RF and ACOG (4 combinations)?