Advanced Warfare Plots: Assault Rifles

[probaddie]

Bal-27

[probaddie]

AK12

[probaddie]

ARX-160

[probaddie]

HBRa3

[probaddie]

IMR

[probaddie]

MK14

[probaddie]

AE4

[probaddie]

There was a serious error with the AE4 plots: I used the pre-patch version of the weapon. That is now fixed.

[ninopettis]

Hey probaddie, thanks again for this. I spoke to you in your post on the subreddit.

I went through these and put the numbers in an Excel file. I dunno if you have those already... I probably should have asked. If you do could you post them please? Anyway, I noticed a few weird things going through the weapons...

Why is the Bal-27 Assaulter any different to the base Bal and the Tactical? Range, ADS time, Reserve mags, and Hipspread are the only differences between the weapons.

Why does the Bal-27 .308 have such a weird plot? Its shape looks different to the others. It also has a better inaccuracy score than the base Bal and Tactical even though it has less accuracy and the same fire rate. Surely there's a mistake here?

Some Bals seem heavily affected by having a higher or lower fire time, despite the fact that they'd get frame rounded at 60 fps. For example, the Obsidian Steed doesn't see a lower fire rate at 60fps, yet it's significantly more accurate than the Carbon, and they both have the same accuracy. The Bal Virtue and Marksman have the same thing between them at a higher accuracy. How come? There's other instances of this with other weapons, but the differences aren't as severe as they are with these Bal examples.

Regarding the burst weapons, there's a few weapons with a shorter or longer burst delay; the ARX Head Shot has the same accuracy as the base ARX but a longer delay and thus a lower RPM; the IMR Hunter has the same accuracy as the base IMR but a shorter delay and thus a higher RPM (before frame rounding only); and finally the IMR Feedback and the Lightweight have the same accuracy but the Feedback has a shorter delay and thus a higher RPM. The thing is, the longer the delay and lower the RPM is for these weapons, the worse the inaccuracy score is. Is this right? Shouldn't it be the other way around?

Finally, how come you shot the MK14 Signature and Desecrator at the same RPM as the base MK14 and some other variants? They both have higher fire rates.

edit: I just read your main post on semi-auto and burst-fire weapons. So is it not humanly possible to fire the Desecrator and Signature as fast as their fire rate limit? I think I've read previously that it IS possible but I'm not sure. I can't even test it myself because I don't have MK14 unlocked. My point about the burst weapons still stands doesn't it; the weapons with a longer burst delay shouldn't be less accurate, right?

[probaddie]

Feb 23, 2015 3:20:10 GMT -5 ninopettis said:

Hey probaddie, thanks again for this. I spoke to you in your post on the subreddit.

I went through these and put the numbers in an Excel file. I dunno if you have those already... I probably should have asked. If you do could you post them please? Anyway, I noticed a few weird things going through the weapons...

Why is the Bal-27 Assaulter any different to the base Bal and the Tactical? Range, ADS time, Reserve mags, and Hipspread are the only differences between the weapons.

Just another one of many oversights. When I compiled the data, I started with a copy of Marvel4's spreadsheet - which had every variant listed - and cropped the list as I found identical weapons. So, 90% of the mistakes I made left orphan variants in the sheet that should have been merged with another (or the base version).

Why does the Bal-27 .308 have such a weird plot? Its shape looks different to the others. It also has a better inaccuracy score than the base Bal and Tactical even though it has less accuracy and the same fire rate. Surely there's a mistake here?

I accidentally gave it the wrong introFireTime (0.11s instead of 0.09s). More on how I'll fix this below.

Some Bals seem heavily affected by having a higher or lower fire time, despite the fact that they'd get frame rounded at 60 fps. For example, the Obsidian Steed doesn't see a lower fire rate at 60fps, yet it's significantly more accurate than the Carbon, and they both have the same accuracy. The Bal Virtue and Marksman have the same thing between them at a higher accuracy. How come? There's other instances of this with other weapons, but the differences aren't as severe as they are with these Bal examples.

This is the biggie. Until this game, I simulated ViewKick in a way that was frame-independent. I did this because framerate is capped at 91 fps on PC (since MW2) and at 1000 fps (the Quake 3 engine's cap) prior to that. This method meant that the worst-case scenario was always represented (most recoil), regardless of platform.

Because of the way (we presume) the game internally handles heat accumulation for the AE4, EPM3 and EM1, I had to abandon that approach for this game; tests showed that the game was clearly using frame-dependent arithmetic to determine the amount of heat accumulated per shot for these weapons. I wrote new code that took a given time between frames (17ms) and bumped up the coded fireTime to match the nearest multiple of that time.

In light of your observation, I did a test run with the Bal variants and double-checked that the fireTime was being rounded correctly: it was. I also checked the plots and, in this run, the two Bal variants you mention do have the same recoil. This leads me to believe that my initial run was bugged. My only explanation for this is that I inadvertently fixed the issue after the initial run of plots. I had been meaning to clear out some cobwebs in my code - making thins clearer and adding comments - for a while, and did so after my run of plots. This must have fixed the bug, though I can't tell myself what exactly the bug was or how I remedied the problem.

For these reason, I'm going to re-upload the entire set after fixing the errata in my spreadsheet.

Regarding the burst weapons, there's a few weapons with a shorter or longer burst delay; the ARX Head Shot has the same accuracy as the base ARX but a longer delay and thus a lower RPM; the IMR Hunter has the same accuracy as the base IMR but a shorter delay and thus a higher RPM (before frame rounding only); and finally the IMR Feedback and the Lightweight have the same accuracy but the Feedback has a shorter delay and thus a higher RPM. The thing is, the longer the delay and lower the RPM is for these weapons, the worse the inaccuracy score is. Is this right? Shouldn't it be the other way around?

Yes, strictly speaking, weapons with longer burst delays have a lower average RPM. However, the burst delay on the ARX-160 is so long and its recoil so negligible that the difference in burst delay is immaterial: the ARX always recovers its ViewKick and GunKick before the next volley of shots is fired. To take this to an extreme, if you waited a full three seconds before firing each burst from the ARX, you technically are firing the gun at a lower average RPM than if you fired them off, say, every two seconds. But you wouldn't expect your accuracy to be any better in either of these two cases.


The difference in inaccuracy score between those plots is likely insignificant; even over 1000 simulations, its always possible for two identical (or very similar) weapons to achieve slightly different inaccuracy scores, even ones that belie the true accuracy difference (i.e., giving the slightly lower inaccuracy score to the gun that is technically more inaccurate). That, or the fireTime bug caused the slight deviation.

Finally, how come you shot the MK14 Signature and Desecrator at the same RPM as the base MK14 and some other variants? They both have higher fire rates.

edit: I just read your main post on semi-auto and burst-fire weapons. So is it not humanly possible to fire the Desecrator and Signature as fast as their fire rate limit? I think I've read previously that it IS possible but I'm not sure. I can't even test it myself because I don't have MK14 unlocked. My point about the burst weapons still stands doesn't it; the weapons with a longer burst delay shouldn't be less accurate, right?

I know some people can fire past 625 RPM, naturally or through a modded controller. However, there does need to be a cut-off point - would a plot of the EPM3 fired at 1200 RPM make sense? Granted, a lot of people use the Desecrator specifically because they can abuse the higher firecap. I think for the re-upload I will separate the Desecrator and fire it at the highest possible firerate.

Thank you for observations. I'd like to be vigilant enough to spot some of these problems myself, but the sheer tedium of uploading these all by hand often saps my will to do any meaningful fact-checking. It is a blessing that there are people out there passionate enough to notice and care about these things.

[Megaqwerty]

Feb 23, 2015 6:05:48 GMT -5 probaddie said:

would a plot of the EPM3 fired at 1200 RPM make sense?

Yes.

[probaddie]

Feb 23, 2015 14:12:08 GMT -5 Megaqwerty said:

Feb 23, 2015 6:05:48 GMT -5 probaddie said:

would a plot of the EPM3 fired at 1200 RPM make sense?

Yes.

Only for Cheaty McCheatersons users with modded controllers.

[LeGitBeeSting]

Feb 23, 2015 14:15:47 GMT -5 probaddie said:

Feb 23, 2015 14:12:08 GMT -5 Megaqwerty said:

Yes.

Only for Cheaty McCheatersons users with modded controllers.

What is a mouse.

[probaddie]

Feb 23, 2015 14:57:32 GMT -5 LeGitBeeSting said:

Feb 23, 2015 14:15:47 GMT -5 probaddie said:

Only for Cheaty McCheatersons users with modded controllers.

What is a mouse.

There's no way you and your hooves can put out 1200 RPM with a mouse.

[Megaqwerty]

Bind Fire Weapon to Mouse Wheel Down.

You're not the dog of me.

[ninopettis]

The reason I'm going through the numbers is because I'm putting together a spreadsheet that calculates HTK and TTK, as well as displaying other weapon stats, also using weapon accuracy, accuracy by range, and attachment modifiers.

Thanks for updating the plots. I've gone through them again and they look better than before. The only real issue is that the ARX Damnation is missing. Another thing, and this may be insignificant and/or a coincidence, is that for all the weapons that don't see their coded firetime increase/decrease at 60fps due to frame rounding, they're helped or hurt more by the foregrip than the weapon that doesn't have a firetime change. For example, the Obsidian Steed is slightly more accurate than the 308 and the Carbon without a foregrip due to its lower firerate when there are dropped frames, and this is how I'd have expected it to be. However, it's less accurate than the 308 and the Carbon when it gets a foregrip. You can see the same thing with the AK12 Fluted vs Hair Trigger, and the ARX Tactical vs Superlite. The differences are only small, but they're there in all three instances, so I thought there could be an error.

This is the first FPS I've played and I haven't even used a semi-auto weapon yet for any period of time so I didn't really know about how fast a person can fire. I just watched drift0r's EPM3 video and he said most players fire at 400-500 RPM, with better players firing at 600-700. I also learned through a Google search that in previous COD games, some semi-auto weapons had a firecap of 625 RPM and the gun would jam if it went over this. I didn't know this and thought 625 was more an arbitrary number you chose. I think you did the right thing not going over 625. While some players are capable of firing faster than that, it doesn't mean that they're going to fire as fast as that all of the time or just on average. For example, a player might be capable of firing 900 RPM but his average RPM would be much lower, and average RPM is more important for the spreadsheet I'm putting together.

[probaddie]

Feb 23, 2015 22:01:25 GMT -5 ninopettis said:

The reason I'm going through the numbers is because I'm putting together a spreadsheet that calculates HTK and TTK, as well as displaying other weapon stats, also using weapon accuracy, accuracy by range, and attachment modifiers.

Thanks for updating the plots. I've gone through them again and they look better than before. The only real issue is that the ARX gosh darn golly gee whization is missing. Another thing, and this may be insignificant and/or a coincidence, is that for all the weapons that don't see their coded firetime increase/decrease at 60fps due to frame rounding, they're helped or hurt more by the foregrip than the weapon that doesn't have a firetime change. For example, the Obsidian Steed is slightly more accurate than the 308 and the Carbon without a foregrip due to its lower firerate when there are dropped frames, and this is how I'd have expected it to be. However, it's less accurate than the 308 and the Carbon when it gets a foregrip. You can see the same thing with the AK12 Fluted vs Hair Trigger, and the ARX Tactical vs Superlite. The differences are only small, but they're there in all three instances, so I thought there could be an error.

This is the first FPS I've played and I haven't even used a semi-auto weapon yet for any period of time so I didn't really know about how fast a person can fire. I just watched drift0r's EPM3 video and he said most players fire at 400-500 RPM, with better players firing at 600-700. I also learned through a Google search that in previous COD games, some semi-auto weapons had a firecap of 625 RPM and the gun would jam if it went over this. I didn't know this and thought 625 was more an arbitrary number you chose. I think you did the right thing not going over 625. While some players are capable of firing faster than that, it doesn't mean that they're going to fire as fast as that all of the time or just on average. For example, a player might be capable of firing 900 RPM but his average RPM would be much lower, and average RPM is more important for the spreadsheet I'm putting together.

I suspect that it probably is just a coincidence. Realize that the difference between these inaccuracy scores is on the order of hundredths of a degree. For reference, I plotted the Bal-27 variants you mentioned using 100,000 simulations:





The error now seems to be on the order of thousandths of a degree. In retrospect, I probably shouldn't have reported six decimal digits of accuracy in the scores .

[ninopettis]

Ok, I can accept that it's just a coincidence. I know it was only minor, but it happened in all three cases where there's frame rounding with the ARs. It even happens (albeit a very small amount) in the example above. I dunno if it'll happen with the SMGs and LMGs but I plan to look at them too to put them in my spreadsheet so we'll see.

I didn't mean to say "gosh darn golly gee whization" above. One of the two elite ARX-160 variants is missing, and it isn't the Steel Bite.

[asasa]

Well as far as I know probaddies simulation doesn't round. It operates as if the game ran perfectly and as coded.

It's also totally possible for a gun with a higher firerate to have less recoil than the same weapon with a lower firerate because viewkick is a velocity and not an xy change. Being able to see this by a glance at the numbers is not easy. With enough simulations, it could be confirmed without a reasonable doubt that the hairtrigger does have less recoil than stock, but I'd doubt this to be the case.

At some increased firerate, all guns will begin to become more accurate, and smoother in recoil, eventually approaching zero recoil

[probaddie]

Feb 24, 2015 9:33:44 GMT -5 ninopettis said:

Ok, I can accept that it's just a coincidence. I know it was only minor, but it happened in all three cases where there's frame rounding with the ARs. It even happens (albeit a very small amount) in the example above. I dunno if it'll happen with the SMGs and LMGs but I plan to look at them too to put them in my spreadsheet so we'll see.

I didn't mean to say "gosh darn golly gee whization" above. One of the two elite ARX-160 variants is missing, and it isn't the Steel Bite.

I just fixed it. Thanks for your help.

All the confusion stemming from the inaccuracy scores might tempt me to do a spreadsheet of them where I simulate a higher number of trials to churn out more accurate numbers. I'll let you know if that idea comes to fruition (i.e., I can will myself to code something that will do the spreadsheet compilation for me.)

[asasa]

Feb 24, 2015 10:03:13 GMT -5 Mousey said:

you need a pretty high fire rate for shots to fire before the peak though

Yeah probably in the thousands.. but the smoothness improves constantly, so even with more recoil, a gun could be preferable.

Example: bo1 m14 - no way to compensate for recoil until high firerate

Also what changed probaddie? I see new plots

[probaddie]

Feb 24, 2015 10:58:52 GMT -5 asasa said:

Feb 24, 2015 10:03:13 GMT -5 Mousey said:

you need a pretty high fire rate for shots to fire before the peak though

Yeah probably in the thousands.. but the smoothness improves constantly, so even with more recoil, a gun could be preferable.

Example: bo1 m14 - no way to compensate for recoil until high firerate

Also what changed probaddie? I see new plots

I just posted those to illustrate the expected error in the inaccuracy scores when simulating 10000 trials per weapon. (I assume you're referring to the tetrad of Bal-27 plots?)

[asasa]

No, I thought the AK12 plots changed but I must be crazy.

[probaddie]

MK14 (Post-03/03/15 Patch)

[asasa]

LOL. Thanks!

[probaddie]

STG44