Black Ops Recoil Plots + MW3, COD4 links

[asasa]

^OP'd lol

---

Random request:

M14 with 70 up kick and 1600 centerspeed [1800 gripped] @ 8SPS + gripped
FAL with 1600 centerspeed @ 8SPS

[psijaka]

I will have a look at the "fantasy" guns next week, but will post them on a separate thread. I want to keep this thread for more relevant stuff.

[asasa]

Hmmm well I did some more testing of M14 grip vs FAL and there is definitely something odd about that M14.

When you consider the vertical recoil differences of 80 to -40 and 70 to -20, you would expect them to similar. M14 has more downwards recoil and does it more often. Slightly more upwards, which is then completely balanced out via 1600 centerspeed. [1/7th more kick, 1/7th faster recenter]

In testing, though, the FAL usually recoiled no more height of the windows on Nuketown, while the M14 quickly shot from the 2nd floor to over the roof, hitting the recoil cap.

Also, the FAL recentered fully quite often. Maybe 25% of the time? [@8sps for both btw]. M14 never recenters.. though every now and then does begin to recenter.

This is far more than the numbers would say. I hold my stance that the recoil is not fully understood. My guess is that the "negative" downwards is actually a minimum upwards.

EDIT

Since the CoD4 M14 was 40-80 up @ 1200 centerspeed, could anyone try CoD4 M14 vs Black Ops M14 Acog+Grip @8sps? I dont have CoD4 anymore :/

[Disgruntled Jigglypuff]

Aug 27, 2011 17:25:57 GMT -5 asasa said:

EDIT

Since the CoD4 M14 was 40-80 up @ 1200 centerspeed, could anyone try CoD4 M14 vs Black Ops M14 Acog+Grip @8sps? I dont have CoD4 anymore :/

If I had a modded controller, I would offer to do this. Sadly, I only have a trigger finger.

[asasa]

Tried 6SPS and damn the differences in recoil are astounding. FAL almost always fully recenters, M14 Grip never recentered.


FAL:
-average horiz roll = ~31,
-average vert roll = ~30

M14:
-average horiz roll = ~20
-average vert roll = 60

This is how I assume it works. Doesn't account for recentering between shots though.. but w/e.

Would explain the lack of full recentering between shots; insane vertical recoil versus the FAL, and all around being a worse gun than the FAL.

Ends up that the M14 still has longer average recentering even with grip, despite the FAL having 2 chances for 70 kick rolls.

[psijaka]

Hey, I've got some new plots for you.

These show the time taken to kill when longshotting on full auto. I have stuck to the LMGs as these have been the subject of a fair bit of discussion lately, and I have just finished an LMG prestige. I may include other guns once I have had feedback.

The charts are based upon 100 x 12 round bursts, and show the percentage of bursts that would have resulted in a successful kill at a given time. A longshot hit is defined as taking place when the gun has fully recentered. Also shown in red is the number of failed kills after 12 rounds.

Here goes:



HK21 = deadly
RPK = lol
M60 grip = good for short bursts
Stoner = worrying

Feedback please; is this useful; is the format good etc

EDIT - I would like to take this further, for instance:

- simulate 1000 bursts of 15 or more rounds
- expand the range of guns (obviously). Skorpion + grip !!!
- look at medium range full auto. This would involve counting some near misses as hits. We would need to have a lengthy discussion about what would count as a hit, and what range should be covered.
- Look at shorter range full auto (ADS). This would involve counting even more near misses as hits. More discussion!

I have more ideas, but we will see where this leads first.

Feedback please.

[phale]

Very interesting. I can't wait to see how these compare to some of the assault rifles. And the RPK is absolutely OP.

One thing I'd like to mention is the time scale on the x-axis. Because each gun has a different firetime, you have used different scales for each weapon. Perhaps it would be effective to use the same scale, so we can put things into perspective.

For example, the Stoner looks pretty meh on this chart compared to the HK21, but if you look closely the Stoner's percentage at 0.768 seconds is pretty close to the HK21's at 0.72 seconds (both around 60%).

[psijaka]

Thanks for the feedback phale.

I agree with your comment about the variable time scale on the x-axis, and am trying to work out how to implement a standard scale. Not as easy as it sounds with a bar chart. Watch this space though.

The ARs are pretty poor; only the AK47/Commando consistently get above a 30% kill rate with a 12 round burst at long range; hardly remarkable. Much better to burst fire or single shot.

With one obvious exception, just forget it with the SMGs. But just wait until you see the results for the Skorpion!

From what I have experienced in the game, it is only viable to longshot on full auto with the Skorpion and (with care) the HK21, although with the HK you have to be prepared to take a pause if the cumulative kick takes you off target. The results for these 2 guns match my experience, but the plot for the Stoner troubles me.


I have tried looking at the results at medium range (25-37.5m), counting near misses within 1 ViewKick Unit vertically and 0.5 VKU horizontally as hits (this is an arbitrary choice of hit tolerance and I would welcome suggestions as to what would be a realistic margin of error at say 30m - any idea Den?). The ARs start to look much more promising on full auto as they get a 3 hit kill. The more random guns look better than the biased with these criteria, but of course skilled players can compensate for predictable recoil....

[Megaqwerty]

I agree with phale: a standard time scale would be great. My suggestion on that note would be to just make each graph larger.

That said, given that the RPK and the HK21 do have the same time scale...it shows how much the HK21 shits all over the RPK.

Also, I started making a spreadsheet the other day that showed TTK as a function of accuracy and this makes me want to finish that as that's slightly related to what you just posted.

Overall, this is an interesting way of displaying each gun's accuracy. It does make the HK21 appear to completely outclass the Stoner, but that's a function of the Stoner's extreme RPM.

[psijaka]

I do need to sort out a common time scale if I am to include these plots in the OP. Working on it. I may try to plot several guns on one graph, to allow comparison.

[psijaka]

OK, think I have nailed it. Cast your eyes over this...

EDIT - please see next post for improved plots



I have managed to incorporate different fire times into the same plot. LMGs to follow. This is based upon 1000 long bursts of up to 15 rounds.

Just look at the Skorpion! The rest of the SMGs are just not suitable for this tactic, but I have included the best; the AK74 grip for comparison.

LMGs to follow.

[psijaka]

Hey, some more plots.

Just to recap, these plots give the probabilities of getting a long range kill within various times when firing on full auto. I have assumed:

- A hit only takes place if the gun is fully recentered.
- No cover.
- No headshots.
- Your aim is steady and the target does not move.





I have decided that the steppy graph I posted previously is not easy to follow, so decided to try a more smoothed plot; just point to point between the different data points, which represent when a round is fired. So do not try to read off the graphs from the lines; go for the data points instead. Despite this drawback, the plots are much easier to read, and more clearly allow comparison between the guns, so this is the format I shall be sticking with.

Now the results

- the Stoner looks kind of scary when plotted in real time against the other guns

- the Skorpion is devastatingly effective on full auto, even with RF. Other SMGs, just forget it.


Once I get sufficient feedback I may move these plots on to the OP.

Next steps -

- possibly look at single shotting V burst firing V full auto on selected guns (quite a bit of work)

- look at medium range full auto. This is where I expect the ARs to shine (they are all poor at long range full auto, which is why I haven't bothered posting). I do need to count instances when the gun if fired whilst almost recentered as hits though, and am looking for guidance here. I would suggest a tolerance of +/- 1 vertically and +/- 0.5 horizontally (viewkick units, degrees whatever they are).

Suggestions would be most welcome.

[asasa]

Wow, those graphs are awesome. Can't wait to see the ones with some recoil tolerance.

Me thinks you should bump it up a bit to maybe 2.5/1

[psijaka]

Thanks asasa

I have already created an AR plot with tolerances of 1 vert, 0.5 horiz; but am happy to try 2.5 / 1; be interesting to see the difference. Will post shortly. Here they are; tighter tolerance first.
SUPERCEDED

SUPERCEDED


EDIT - perhaps the lower graph should be regarded as short range?

[psijaka]

RANGE AND HIT TOLERANCE - I really need to get to grips with the ranges and hit tolerances, or "hit box" size. Looking at ishbane's damage charts, I think choosing the following ranges would be good:

Long - 70m - an easy choice; just allow min damage and zero tolerance on misses - only a fully recentered shot will count as a hit.

Medium - I would suggest 35m - at this range the ARs still get max damage, but the SMGs and the M60 are on minimum damage. Hit tolerance +/-1 vertically; +/- 0.5 horizontally. Or 1.25 vertcally to be half the size of the short range tolerance at 17.5m?

Short - this one is quite easy. 17.5m is the maximum range at which all SMGs are on max damage, with one obvious exception. And the M60 gets a 2 hit kill up to 25m ;D It will be interesting to see how the AK74 RF grip compares with the FamAug at this range. etc. Tolerance +/-2.5 vertically; +/-1 horizontally.

Should I adopt these as a standard?

I would like to do some tests, but am limited in what I can do on the Wii.

[asasa]

Awesome again. Thanks psijaka. Yes to making them a standard! Not sure what kind of tests you are thinking of, but if they are something I could do, I'd be willing to do them.

I think there should be some room for partial recenters even on the 70m. 1.25 vert for Medium would be good. .25/.625 would be the obvious answer [if you feel some tolerance should be there], but something like .2/.5 is possible as well. It's really amazing how much faster they kill at [our version] of 17.5M than they do at 35M.

If we could find out what 1 unit of deviation was, we could find the width of a player, and the corresponding sizes at 70/35/17.5M.


Would you be able to do some calcumalations to find out, and we could work off of that?

I was trying to think of how to do so... but then realzied I'm way too tired for math. Lol. Also, M60 is gonna look super OP'd haha.

Oh, and.. do you have any comment on the M14 being 40 min and not -40?

[psijaka]

To be consistent, I suppose I should choose +/-0.625 vertical and +/-0.25 horizontal tolerance at 70m. But I really don't think it will make much difference. Will try and see.

Thanks for the offer of helping with the tests; I would have in mind trying to work out how big a body or torso is in terms of viewkick units, degrees, whatever at 17.5m, 35m and 70m. So we could make an informed decision on the hit tolerances at these ranges. Not sure exactly how this could be done, but would probably involve modding weapon stats, so PC version needed I suspect.

As for the M14. I have used it a moderate amount and I do not recall having ever seen it kick downwards, either in gameplay or whilst doing tests looking for downwards kick. hmm.

Will be interesting to repeat the M14 V FAL tests at different fire rates and seeing how they stack up aainst each other on the new graphs. Watch this space.

[psijaka]

@ asasa

I have gone back to my schoolboy mathematics, and worked out the hit tolerances from arctan(target/2 / range).

If we choose a target of 1m h * 0.4m w, representing the torso, deviations of less than 0.5 or 0.2m count as hits.

Plugging this into the formula above, we get (in degrees):

70m
+/- 0.41 vertical
+/- 0.16 horizontal

35m
+/- 0.82 vert
+/- 0.33 horiz

17.5m
+/- 1.64 vert
+/- 0.65 horiz

I think that we should go for a target a bit smaller than the whole body, as arm, leg and head hits are a bit random, and the target may be partially obscured or crouching.

EDIT - I would appreciate some feedback on this from other Bros too, before I really get busy.

[psijaka]

OK, here are the SMGs at 17.5m range.



A few points to note:

- The MP5 plot is virtually identical to that of the AK74, so is not included. There really is very little between the two.

- The MAC11 is the same as the Spectre. PM63 and Kiparis slightly worse (omitted for clarity)

- I have not tried to include all possible combinations of attachment; would be far too confusing. Grip generally adds around 4% - 6% to kill probability.

More to follow.....

[phale]

Wow, the Up/Right recoil weapons do not fare well with this analysis. Looks like compensating might be more important than we think.

[psijaka]

Agreed. I haven't yet tried to simulate a skilled player moving his aim downwards and to the left whilst firing. Or aiming at the bottom left of the torso. But I may.

Also, we also must remember that these plots are for stationary targets; enemies at this range are rarely so obliging as to stay in one place whilst we shoot them.

[Megaqwerty]

Dude, I don't know about you guys, but the Uzi is totally more accurate than the Spectre.

This is really my only problem with this analysis. With your charts in the first post, this is easily compensated for by ignoring anything past the second shot as the difference between random and biased recoil only begins to manifest in the third and subsequent shots.

[psijaka]

We have to be careful how we interpret these plots. The Spectre has a much tighter recoil spread than the Uzi; problem is, it is biased upwards and to the right, so most will miss, especially as time progresses. If the player is skilled enough to compensate for this, them more will be on target.

Whereas with the Uzi, the shots are all over the place. Although 2nd shot accuracy is poor, accuracy for subsequent shots does not drop off as quickly as for the Spectre.

Both guns are poor for longer bursts when compared to the MPL, AK74, MP5, and especially the Skorpion.

Also, this is the maximum range at which the SMGs get their maximum damage. At closer range, more of the Spectre's shots will hit. I may do some plots for 8.75m, but I am not sure how valid these would be, as at this range hip firing becomes a more viable option, and people are much more likely to be moving fast.

EDIT - I do plan to look at the response of the guns when they are fired in short bursts of 3 rounds, with a gap of (say) 0.25 seconds between bursts. I would expect the Spectre to outclass the Uzi if used in this way. Likewise the Famas V Aug or Galil V AK47.

EDIT 2 - even at close range (8.75m and 5m) the Uzi has higher kill probabilities than the Spectre!
I really must mod the spreadsheet to allow for recoil compensation.

[psijaka]

OK, here is a plot at Close range, which I define as 8.75m; half tthe distance I chose for Short range (17.5m). This shows the Uzi Vs Spectre, both with and without crude recoil compensation.

The method I used was to assume that the Spectre user shoots at the middle of the bottom left quarter of the torso; that is 25% of the way up and to the right of the bottom left corner of the target. There is no compensation once the burst starts; in reality, the player may shift their aim down and to the left, but that is beyond my spreadsheet as yet.

Even though this is a pretty crude method, it clearly shows the benefit of compensating for the Spectre's recoil, and hints at the true strength of the gun. More to follow....

[asasa]

Hmm, yeah, not really sure that simulating compensation is going to be possible. Personally I would have just gone with a centerspeed increase, or a constant X,Y movement pushing down and to the left [but, unlike centerspeed, it would continue over the lines]

[psijaka]

Sept 6, 2011 20:47:19 GMT -5 asasa said:

Hmm, yeah, not really sure that simulating compensation is going to be possible. Personally I would have just gone with a centerspeed increase, or a constant X,Y movement pushing down and to the left [but, unlike centerspeed, it would continue over the lines]

I can now do down and to the left, will post soon

[mannon]

Okay this is starting to get REALLY interesting. Okay so now you are able to simulate plain full auto as well as aiming for the bottom left area of a target as well as shifting aim over time and burst firing... In other words 3 complete recoil compensation methods.

Here's my thing. We can certainly try to predict the various compensation methods used per each weapon to simulate player usage and thus compare one weapon from another, performance-wise.

But... I'm also thinking we might have another very nice capability here. Namely to try out simulations of lots of different burst lengths, compensation styles, and combinations in order to discover empirically which compensation methods actually perform best per weapon. O,O Sure it may not make a huge amount of difference, but then again... It would be kind of interesting to know just how effective different techniques are and it'd be nice to know whether I should burst fire 2, 3, 4, or more rounds per weapon for the optimum performance. ;3

I'm just saying.

[psijaka]

Working on it mannon. At present, I can:

-Simulate kill time probabilities for 1 second long burst on full auto at any range.

-Crudely simulate aiming at the bottom left of the target; only valid for up and to the right biased recoil guns (the usual suspects).

-Simulate dynamically shifting your aim down and to the left whilst firing a weapon. Or just down, or any way, in fact.

-Simulate single shotting at any rate (easy; just change the full auto fire time to 200ms or whatever).

-Simulate firing in bursts of 3, with a variable gap between bursts.

What I cannot do yet is handle bursts of 2 or 4 shots; but I could add this (if I can be bothered!)

I am also toying with the idea of including ADS time; would be easy if bros think it would be useful.

My problem now is that there are so many different combinations of plots that I am struggling to know where to start.

What I am planning:

- Straight comparisons between guns within each class at Short, medium and long range (17.5, 35 and 70m), shooting at a target 1m x 0.4m (representing the torso).

- do some comparison of guns from different classes, at each range.

- do some plots showing the benefit of using different recoil compensation methods (none, aim low, shifting aim, burst firing, single shotting). These would include 2 guns from a class with different recoil types (Famas v Aug or MPL v Spectre).

I will post some of these on the OP eventually, but would like some opinions on them first.

I am working away from home at the moment (away from my family/Wii) so I do have time to concentrate of this. Either that or head for the bar....

[psijaka]

OK here is a plot of the Spectre showing the effect of various simulated methods of recoil compensation. Also plotted are the Spectre, MPL and Uzi without compensation, for comparison.



Some points :

- Aim offset is aiming at the middle of the bottom left quarter of the target. EDIT - aim for their appendix!

- Aim shift is moving your aim downwards and to the left, each at a rate of around 6 degrees per second. Move too fast and you miss (and die)

- offset and shift is same offset, but shifting at half speed. The most deadly, but would take more skill to pull off.

- Spectre burst fire is with a gap of 200ms between bursts. 3 bursts get the kill; slow but sure (if you can aim and don't die first)

I have tried other rates of aim shift, and 6 degrees /sec down + 6 left seems to give the best results. Much above this and the kill time starts to drop off. A tricky technique to really master.

At greater range, you would have to shift your aim more slowly; double the range, half the shift speed I suspect. Even more tricky to get it right, and probably less effective (will check).

EDIT - these plots are now based on the simulation of 5000 long bursts of 17 rounds. 1000 bursts was just a little bit too variable for close comparison. My laptop breaks out in a sweat when I open the COD.xls file!

[psijaka]

mannon
I can now do bursts of any length. Any particular gun you would like me to look at?