How exactly does centerspeed work?

[psijaka]

Mar 3, 2011 22:07:07 GMT -5 Contrary said:

Mar 3, 2011 4:00:37 GMT -5 psijaka said:

NO CONTEST

but but but *snivel* I loooove my fal.

Actually that is a fairly convincing argument. gosh darn golly gee whizit I would try stealth M14 if it weren't for the fact M14 doesn't have Dual Mags.

I just can't bring myself to try the FAL, especially with that poor hip accuracy. With a semi auto, you need all the help you can get if you meet a Famas wielding thug coming around a corner!

Extended mags is good; only a bit less total ammo than Dual mags (90 v 100, I think).

[mannon]

FAL in MW2 was overly critisized. It was definitely a harder gun to learn than the others, but by the time I got mine I'd been using burst fire weapons and they had sort of taught me how to use a semi auto better than I did back in W@W. I loved the FAL even before we discovered Holo+SP made it a 2HK sniper rifle. ;p (Though if you really want that you could always go for an M21EBR+CB at the cost of gaining idle and longer ADS time.) I dunno about BO yet, but doubt I'll like the FAL as much as I did in MW2. It was just fun in MW2.

As for the M14 I'm not really convinced there's anything wrong with the math using the information Den has given us. Even with the ACOG the chances of a second shot kicking down at all instead of recentering are a mere 17.33% and with only a maximum of -20.8 units of vkick left over. I don't know what that works out to in degrees off axis, but it's not much compared to the 60.8 maximum up kick units left over.

Presuming what Den said about centerspeed is right then our math should be too. But even so it would be a pretty rare thing for it to kick down, and you'd really only notice on the second shot, because you're going to have so much leftover kick that all your other shots are going to make a line strait up. Basically a 3rd shot has to deal with whatever leftover up kick remains uncentered from the first shot as well as the kick it got from the second.

We could devise a test, perhaps. If you can find say a flat wall with some sort of line across it or some other way of marking height. You could aim right at the line then fire a series of 2 shot bursts. 3rd and further shots simply have higher and higher odds of going high so are not desired for this test. You could just sit in the same spot and burst over and over again. Do enough bursts and eventually a second shot should (theoretically) land slightly lower than the first shot. I would advise using quite a bit of range to ensure that any deviation would be as visible as possible as the downward kick could be very small, depending upon the way the view moves during a kick. (Which we've been debating here.)

I would suggest a large sample size of tests before concluding that it never kicks down. Since even at 20 it would only on average kick down 3 times and it's well within the realm of possibility that it could simply kick up or recenter 20 times in a row at random. I'm not even sure I'm too comfortable with 50 as a sample size as that would only on average give us about 8 down kicks. I'm particularly concerned because a lot of uncentered kicks are going to fall so close to centered that we mere humans wouldn't notice the difference. The formula we used does not count an area as being on target, but only exactly, precisely, zero deviation. Even 0.001 degree would be considered not centered. So that means that even though the gun may downkick 17.33% of the time an unknown number of those will be effectively indistinguishable from a recenter.

I'd say go ahead and bump the test up to 100 two shot bursts. I still don't know if that's enough to really be definitive, but on average you should see about 17 downkicks. Between the indistinguishable kicks and random chance maybe that's enough to reliably spot a few down kicks... maybe.

Luckily you don't have to be totally precise about 2 vs 3 shot bursts. A 3 shot burst should give as valid a result as a 2, it just wastes more ammo. But you absolutely must recenter completely between bursts. So it would be a slow burst rhythm.

Forgive me for being skeptical of your skepticism. It's just that generally Den is right about these things, and really the numbers show down kicks to be rare enough that I don't think mere anecdotal evidence is enough to disprove their existence. We need some actual scientific testing before we can conclude something like that. Sadly I don't have BO so I can't do any testing, myself. ;-(

[asasa]

Mannon; the M14 Acog has a "glitch" like the Barrett from MW2. Its max yaw is set very low. [Well, not really, but relatively low] This means with Acog, all vertical recoil can stop after about 5 shots. Continuous fire never once drops below that point, meaning, no shots ever kick down.

Extended mags is good; only a bit less total ammo than Dual mags (90 v 100, I think).

Ext Mags doesnt give extra ammo.. or does it on the FAL? FAL is usually 80 bullets; 120 with dual mags.

Dual mags gives 2 extra mags.

[mannon]

When the gun kicks up to maximum centerspeed will slightly bring the gun down from that before the next shot fired, and that would be your highest possible point. At that point any kick that hits the maximum will cause the next shot to land in exactly the same spot. However, the gun need not kick down to result in a lower shot. The gun could actually come up with a null kick at 0 or a very small vertical kick that doesn't quite make it to the maximum, and that too would result in the gun dropping.

For the gun not to drop at all, ever suggests a viewkick more like the M60 which is from 10 to 60 and never goes negative, or else that something about what Den says about centerspeed and viewkick is incorrect. Unfortunately I can't test it myself, but it is interesting.

[asasa]

Right, and not once has it dropped... so, its always getting enough to cover for how much the centerspeed... centered.

I probably calculated it wrong, but, 40 kick @ 7SPS; or 28 @ 10.

What I find interesting is 7SPS. 40 kick? Thats quite a bit. I think I can set this mod lower, down to 5 or 6.. so we'll have to see if its covering even more than 40, or if it starts to drop at that point.

[mannon]

hmm... Interesting...

[mannon]

I hadn't actually thought too much on the M14 being semi-auto. I've just been using it's firetime. (To be fair it's not that easy to tell from the charts alone which weapons are semi, full, or burst, and alas... I lack the game it-self to verify.)

That makes a pretty big difference, though. The maximum side kick is exactly the same as the maximum down kick. The maximum amount of time needed to recenter from a kick to either side or down is merely 0.2 seconds with the ACOG. Thus a rate of 5 shots per second would be enough to negate every single side or downward kick leaving nothing but up kicks and recenters. (Note that a rate of 5 shots per second would produce 6 actual shots in the first full second then 5 per second thereafter unless you double count shots. It's due to the same math as the hits to kill. The first shot is effectively instantaneous and thus you always get 1 more shots than the number of delays in a burst.)

A firetime that produces a rate of 7 shots per second ( 1 / 7 ~= 0.143 ) is about 71% of the time needed for the full down/side recenter. Still, you would expect some downkicks at about 9.52% of the time, at that fire rate. Well... that is 9.52% of second shots should kick down rather than up or recentering. That number does not translate to 3rd or later shots, and certainly not to when it's pegged at the max kick.

What I'm finding particularly interesting here is we may not be understanding exactly how the view interacts with the maximum. If all vertical movement stops and it begins to immediately start recentering upon hitting the maximum that's one thing. Another possibility is it could alter the viewkick before it is applied and thus simply make the maximum the apex. This would result in less of a jarring motion upon reaching it and would actually just look like the gun keeps kicking up and recentering, just that the up kicks are never higher than a certain point. Another possibility is that it could still be more or less attempting to kick and recenter in the same amount of time, but peg the view at the top of it for the duration somehow. The third option could actually result in the view sticking right at the top and not jumping up or down. I kinda doubt that's how it works, but I guess it's possible.

I think we may need to know more about how exactly the view interacts with the max viewkick. It is possible that the interaction somehow nullifies any downward kicks. I don't know. *shrug*

I'm glad this was spotted though. I'm sure it is something simple, but it's worth thinking about it as our assumptions may prove to be false.

[asasa]

I think I'm misunderstanding you in your second 'paragraph.'

When I hit the max yaw; there is horizontal recoil, even gripped. You can see the shots going to the left and right, so it is not fully recovering.

All of this has me wondering, though -- does ANY weapon VISIBLY kick down, or kick down at all?

Yes, the odds are lower of kicking down, bla bla... but you can definitely see the vertical kick every time. Downward? Hmm..

For one, I'll be testing that "max recovery time" on a few weapons. Will be interesting to see whether or not that works out as we would hope. I'll exclude semi autos, esp the M14 .

[mannon]

The more testing the better. Are you using a script or programmable controller to set a ROF? Should be a gosh darn golly gee whiz good way to test the max recenter times based on the centerspeed formula given by Den.

As for the second paragraph I was just saying if you go down to 5 shots per second then the horizontal and downward kick should always be compensated by centerspeed. But, you were testing at 7.

Basically the slower the ROF the closer to center we get, but also the less likely we are to get a downward kick at all and at 5 shots per second it becomes impossible to get anything but recenters and strait up kicks. Though once we've kicked up already that kinda changes things in the vertical arena.

[psijaka]

Mar 4, 2011 17:27:06 GMT -5 asasa said:

Mannon; the M14 Acog has a "glitch" like the Barrett from MW2. Its max yaw is set very low. [Well, not really, but relatively low] This means with Acog, all vertical recoil can stop after about 5 shots. Continuous fire never once drops below that point, meaning, no shots ever kick down.

Extended mags is good; only a bit less total ammo than Dual mags (90 v 100, I think).

Ext Mags doesnt give extra ammo.. or does it on the FAL? FAL is usually 80 bullets; 120 with dual mags.

Dual mags gives 2 extra mags.

Ext mags gives you 1.5x ammo per magazine (or 2x for LMGs or 10/7x for the Galil!). Usually you get the same numbwer of mags so total ammo is increased by the same factor. 200 + 200 for M60 extended mags!

The exception seems to be with rapid fire, where you don't get more ammo (an attempt at balance?).

All this is on the Wii, by the way (not sure why other consoles would be different).

[aequinox]

Mar 5, 2011 23:22:41 GMT -5 asasa said:

All of this has me wondering, though -- does ANY weapon VISIBLY kick down, or kick down at all?

Try a Stoner 63 or HK21 with ACOG (same recoils) on any wall and you can see that it does in fact kick down lower than the original aim point.

EDIT: Although the maximum negative yaw might be seriously reduced.

[Contrary]

Mar 6, 2011 11:43:28 GMT -5 psijaka said:

All this is on the Wii, by the way (not sure why other consoles would be different).

Yeah Imma guess this is just the Wii. For PS3 at least, you get less clips when you use emags so that you get the same amount of ammo. With the M60 you get no reserve ammo, just the starting clip.

[Zero IX]

Yes, on the 360/PS3, Extended Mags doesn't give you any extra ammo. M60, for example, gets 200 bullets with none in reserve.

[asasa]

Tried to test if the "max recenter time" really worked out, but many speeds ending up 2 shot bursting, or 1,1,1,3, etc. Seems to be a mix of the game and the controller. I would imagine this could be tested on PC without issue.

[psijaka]

This interested me; as much for the recoil pic as for the point Den is making.

Mar 12, 2011 16:25:56 GMT -5 Den said:

And here, I modified the COD4 M4 Carbine so that it has:
adsViewKickPitchMin 30
adsViewKickPitchMax 40
adsViewKickYawMin 20
adsViewKickYawMax 30
adsViewKickCenterSpeed 1200

What happens with these amounts?
This. A steady upwards climb with a slight pull to the left.

Those groups are fifteen to twenty shot bursts.
The one on the right, five shot bursts.
I'm probably about two melee lunge ranges from that wall.
There's still a tiny bit of (intentional) randomness. But it is very predictable while still being... recoil-ly.

So I tried to simulate it, using the Viewkick = displacement; centerspeed = velocity model. This is what I get (5 ten round bursts).



Pretty close; I would say. Perhaps the pictured recoil is slightly more vertical than the simulation.

[aequinox]

Possible to simulate with velocity/acceleration model?

[psijaka]

Mar 13, 2011 17:12:47 GMT -5 aequinox said:

Possible to simulate with velocity/acceleration model?

Trying!

[Den]

I like this recoil. Here's another.

I modified the G3 to be... kinda like a SCAR. The red lines are a rough approximation of the path it took.

PitchMin 25
PitchMax 40
YawMin -5
YawMax -35
CenterSpeed 700
FireTime 0.12, Full Auto

With a slightly greater kick per shot than the M4, this expectedly more powerful weapon will have a harder time landing hits consecutively at long range without pausing.
The slightly larger horizontal randomness also prevents it from being as easily controlled, also a means of preventing it from being too good at a distance.

I didn't measure it for any kind of balance. Killing speed and the recoil are probably both favoring the M4 greatly.

[psijaka]

This is what I get:


compared to recoil screenshot

[mannon]

You know I've not looked at the actual stats to any of these as I don't have the games on PC. The X axis is reversed? Right is negative? I don't guess I really have a question there, it's just funny how things are sometimes. I dunno why anyone would program it that way, though I'm sure it made sense to someone at the time.

[psijaka]

Mar 15, 2011 11:38:59 GMT -5 mannon said:

You know I've not looked at the actual stats to any of these as I don't have the games on PC. The X axis is reversed? Right is negative? I don't guess I really have a question there, it's just funny how things are sometimes. I dunno why anyone would program it that way, though I'm sure it made sense to someone at the time.

Yes, strange isn't it.

I have been looking closely at the comparison between the Viewkick=displacement, centerspeed=velocity, and the Viewkick=velocity, centerspeed=acceleration (or deceleration) model. I think that I have sorted out my errors.

The shape of the recoil plots from the 2 methods is exactly the same, but I find that recoil for the velocity|acceleration model is greater than for the displacement|velocity model as the fire time increases. Thus, if I scale the 2 plots to have equal magnitude for (say) the AK47 with a fire time of 0.08; the recoil for the M60 with a fire time of 0.112 is worse using the velocity/acceleration model. And better for guns with a faster fire time.

The velocity|acceleration model uses

x=ut+1/2at²

x=displacement at the time a subsequent shot is fired
u=velocity (viewkick)
t=fire time
a=centerspeed/2.5 (to give centerspeed/5=recentering per second). A deceleration, in other words.

Will post some comparative plots shortly, but want to do some more checking.

[n1gh7]

In no math that I can think of does the right direction ever have a negative value. Maybe it is a leftover from the game's engine being on a default of inverted? But this would mess up the Y axis too, and it is not like that.

Also, did anyone ever think of the recoil model as a sphere? Maybe we are thinking too linearly.

[mannon]

Actually it sort of is a sphere in a way. What I mean is the viewkick pushes your view sideways and up and down on the X and Y axis, so naturally you are merely rotating around the camera's origin.

If you think of a globe the longitude and latitude lines give you a pretty good idea of the two axes. That is if you picture the player being at the core of the globe. Y kick pushes the view up or down towards one of the poles, while X kick simply pushes you around the equator. If you're looking up then X kick pushes you along an arc of a circle above the equator.

The catch is that Den discovered that centerspeed works on the X and Y axes independently rather than in concert. In other words when kicked off at a diagonal the centerspeed doesn't necessarily bring the view back to center on a direct path and unless the view is kicked at a 45, 135, 225, or 315 degree angle then one axis will recenter before the other one does. Thus kicks not on those angles should describe arcs somewhat like what I graphed earlier for a 40, 80 kick.

Since centerspeed works on them independently we really only have to undesrtand how it works in 1 dimension to figure it out. Once we can model that the 2D model is merely an extrapolation.


BTW psijaka, I was wondering if your velocity|acceleration simulations are keeping leftover velocity when the next shot is fired or overwriting it?

At the moment I'm leaning towards a displacement|velocity model simply because the velocity|acceleration model makes it impossible to kick across zero on the axis without reentering at zero first and even then it would seem to require something of a special case while at zero to allow the viewkick to happen at all.

Well... You could allow it to kick across zero, but you'd have to allow velocity to carry it across zero rather than stopping it, and unless you modeled some sort of friction that would result in the view perpetually orbiting center after just one shot. Even with friction in the model the view wouldn't just return to center and stick there like we observe in game. It would return to center and go past it, then come back after a smaller arc, and slowly wobble back and forth like this until it settled down. Of course, with enough friction it wouldn't necessarily wobble too many times, but at least once is pretty much guaranteed, especially for big kicks like the Barret in MW2.

On the other hand, this does not in and of it-self invalidate the model. Given the (to me) surprising amount of recentering going on there is ample opportunity for the view to return to zero on one axis and thus be free to kick in either direction. We could, however, debunk this model with careful observation, if we can simply find a clear example of the view being off center to one side from a previous shot fired and kicking across the zero, then returning to zero via centerspeed. Careful testing with a low ROF weapon that has a lot of kick should reveal it one way or the other, methinks. Video evidence of it would really be great to have rather than just observation.

Another possible way to test it would be to fire 3 shot bursts being sure to note the exact position of the first shot. If it's not possible to kick across zero then the 2nd and 3rd shots will always fall to the same side if not recentered and it would be impossible for the 3rd shot to ever fall on the opposite side of the 1st. It would be possible for the gun to kick to one side, recenter, then kick the other way, but in such a case the 2nd shot would fall at center and again no two shots would fall on opposite sides of center. This should actually be a pretty easy test, especially if you just test left/right kick and use a vertical line on a wall as center.

I will say, though. It is possible that a velocity/acceleration model that somehow does allow cross zero kicks could work. I just don't see how you'd do it in a way that's simple like we would expect, but then there are all sorts of simple bits of code with complex results, and I wouldn't rule it out entirely.

[mannon]

Okay I'm trying this with a UMP and I'm not sure how many tests it'll take to be statistically significant but I've yet to get 2nd and 3rd shots on opposite sides of the first. O,o

The TAR works even better for testing with more horizontal kick and less vertical. It's not really necessary to use vertical lines since you can just note the position of the sights once you recenter on the first decal.

I've tested probably a couple hundred 3 shot bursts now and not a single one has put the 2nd and 3rd shots on opposite sides of the 1st one. Still, even if it is possible I suppose it wouldn't be common with such a small burst. But if you got a small kick to one side and a big one to the other then it should happen... if it can.

Well, I giveup for now. I did a few hundred more 3 shot bursts. At one point I got a little burst happy and moved on really quick then though the burst I fired may have put a bullet on either side of my first one, but I wasn't using the lines and moved on so fast I couldn't really say which was the first bullet. Aside from that and a couple of accidental 4 shot bursts there was not a single case of bullets landing on both sides of the first shot fired.

Not definitive, but certainly interesting. And the implications are interesting as well. For example, if the view cannot be kicked across to the other side without recentering first then this means that weapons that recenter less often will tend to kick back across to the other side less often.

Case in point... M14 vertical kick. If a downward kick while the gun is already above zero only ever results in the gun hitting zero and stopping, and considering that even while at zero it only has a 1 in 3 chance of going down at all, (and most of those will recenter before the next shot is fired) then I'm not surprised that we aren't seeing it kick down.

It also means that guns that don't have a directional bias gain one as soon as the first shot is fired until it recenters on that axis... Now that, to me, actually does start to sound like behavior I've seen. It always seems to me that when I fire bursts they tend to go to one side or the other rather than actually clustering around center.

Mebbe I should test Famas or M16 since they can't help but fire in bursts. Though 3 shot bursts with the TAR are easy with it's fire rate and the sound makes it pretty easy to tell when I fudge up... which is actually pretty rate.

Doing this does make me think about burst technique, though. I'm finding that it's easiest to get a good consistent burst by squeezing and releasing the trigger at a certain speed, rather than trying to squeeze all the way, hold, then release. For the TAR it's actually a pretty slow squeeze to get 3 shots. Though this slows down the first shot a bit since I'm not mashing it as quickly as I could.

I wonder what technique other bros use. I've never really paid too much attention to my burst firing technique in detail because it's not exactly what I'm thinking about when I'm in game. ;p

At any rate these seem to be some pretty interesting test results... I'm getting tired of firing bursts though. heh

[psijaka]

@ mannon - You have been busy!

In answer to your question, I assume that the viewkick velocity is set to the new value for the next shot, with no addition of any residual velocity from the prev. shot. I think that there is a post by Den on this subject (must hunt it out).

Anyway, here goes.....

The plots below are for the Galil, with fire time modified to show the effect upon the 2 different methods of calculation. Note that the centre plot is scaled so that the 2 plots overlay exactly, by multiplying the displacement plot by the normal fire time! Also note that the same set of randomly generated viewkick figures are used throughout.1 square = 5 units (degrees?).



More to come. If I am keen enough, I may try including residual velocity, but I am not aat all convinced that this is how it is handled in the game.

[psijaka]

Some more plots and discussion at

denkirson.proboards.com/index.cgi?board=general&action=display&thread=2781

or (plots only)

i1225.photobucket.com/albums/ee391/psijaka/DisplacementvVelocityMP5withwithoutRFV2.png

i1225.photobucket.com/albums/ee391/psijaka/DisplacementvVelocityKipariswithwithoutRF.png

Note that I have scaled the displacement plot so that it is exactly overlaid by the velocity (green) plot when the gun is firing at it's normal rate (rightmost plots)

[mannon]

Ahh good. I can't really say whether the game would overwrite velocity or compound it. Either would seem reasonable if building a recoil model from scratch. Though compounding it would definitely have some effects that could complicate recoil patterns. Consider the M14 with it's 80 vs -40 Y Kick. A lot of the time at it's max fire rate the gun would still have some upwards velocity and if you keep firing at a rate that usually still has some upwards velocity then you're gonna be heading strait up faster and faster with each pull of the trigger, actually accelerating as velocity compounds.

On the other hand if you're coming back down then it would actually help the gun recenter even if you are still high above center.

I am definitely leaning back towards velocity after seeing your input alongside the 3 shot burst tests I've been doing. I would think that at the very least there would be a difference in velocity vs displacement models when going across the axis, though you could always build a displacement model that also doesn't allow kicks across axis. I suppose in that case we would have approximately four competing models. Displacement with and without cross axis kicks, and velocity with and without compounded velocity. If my testing is any indication then we can probably eliminate displacement with cross axis kicks. I had hoped it might more definitively indicate displacement vs velocity, but while it would be more complicated to allow the velocity model to cross the axis, making a displacement model not cross the axis actually seems pretty easy to me.

Still, I'm not 100% sure of my testing. Maybe I should go fire some more bursts. Any other suggested tests that could be done without video as I do not have a PVR? BTW I found a Raffica actually works quite nice for these tests. The burst pattern actually doesn't seem to be any wider than the AR's in my limited testing, BUT the bullet decals are much smaller. The only annoyance is having to manually reload because you wind up with 2 bullets left instead of a full burst.

Maybe I should look at W@W for testing. We actually have the stats on those weapons. Are there any maps that are good for that? My memory of W@W maps isn't great, but they all seemed pretty much blown to hell already with relatively few clean places to make bullet decals.

Regardless of model I'm about 85% convinced that is IS impossible for a gun to skip from one side of the axis to the other with a viewkick without recentering first.

[psijaka]

My velocity|acceleration simulation certainly lets me skip across the axes. Perhaps this is an error (?). If so, a relatively minor one that I could fix easily enough.

I really am leaning heavily towards the velocity model now.

What I would like to see is some recoil screen captures for an SMG with and without RF; if RF has much worse (wider) spread, then it would suggest that the displacement model is correct (or the velocity model with cumulative recoil); if RF has similar or even slightly less spread, then this would be strong evidence for the velocity model, without cumulative velocity.

I think I will do some tests myself; although these will be "by eye" as my video capture attempts have failed to date.

[mannon]

I think that a velocity|acceleration model probably should not skip across the axis. The reason is in this model the viewkick is applied as a velocity bump to the view at the moment a shot is fired and from there it is simply allowed to float around under the effects of centerspeed until it reaches zero. Because of this even if you kick with a high velocity towards zero as soon as you hit it velocity should be expunged.

To allow it to pass across the axis would require you to remove this zeroing out effect, at least temporarily. We do know that the view doesn't bobble or orbit when it returns from a single kick. It goes to zero and sticks there. But we don't really know the bahavior when there are multiple kicks. It is possible that this sticking at zero behavior is only turned on after the apex of the last kick. This would allow it to kick across the axis, but would still stick there when centerspeed brought it back to zero. Of course, a cross axis kick would actually require some additional calculating because your direction of gravity (centerspeed) is going to switch directions half way through. Basically you would have to calculate the fall to zero with centerspeed helping it get there and get the residual velocity at zero then do another calculation for the other side with the direction of deceleration reversed. Of course you may be doing that. I'm not sure.

At the moment I'm mostly convinced crossing the axis doesn't happen, but I'll do more testing. I have fired a few hundred 3 shot bursts apiece with the UMP, TAR, M16, and FAMAS, and a bunch with the Raffi too, and I have yet to see a single burst where shot 2 landed on one side of the shot 1 and shot 3 landed on the other. There are a lot of recenters and a lot of 2 or 1 shots to one side, but never a shot on either side of the 1st.

BTW if you can't kick across the axis this would potentially increase recenter rates on subsequent shots as any kick that would have gone across the axis but not recentered before the next shot will stick at center immediately upon reaching it.

[psijaka]

I take your point, mannon. As you say, not crossing zero would increase the probability of 3rd and subsequent shots being recentered.

But having said that, I really don't think it would be too difficult to code the game to allow crossing of the axes in the direction of the last kick.

Maybe I'll do some BO cross axis tests with the M60; nice slow easily controllable fire rate with big horizontal recoil.