A "Formula" for Comparing Recoil Stats

[probaddie]

May 29, 2012 7:05:37 GMT -5 Evan950 said:

1. these scores only match the stats of how accurate the first couple of bullets are
these scores in some way reflect recoil, but its only 1 factor.
2. as said before, recoil & kills are different for all players, some prefer higher recoil becuase of their bad accuracy, so they depend on luck off the recoil, others want the gun super accurate becuase they are super accurate.
3. in your tests the ACR didn't score well becuase of the fact that there is no chance that the bullets after the first will land in the same spot as shown by pjiaska
4. however,
what makes the gun so good is that it will bounce around sending alot of bullets in a tight area
unlike the g36c, the first bullets can all be on target, but the next stream after that will kick straight up with recoil

5. as said in a post above, it's dependant on the player, and then still the guns would have to be put in groups of how they work and then rated from there.

1. Strictly speaking, this method actually only rates the weapons on the probability distribution of your point of aim after the first shot. But the mechanic that governs the placement of the second shot (i.e. your point of aim after the first) is the same that governs the placement of all shots. Even the fact that cross-axial movement is not allowed is accounted for in the method. (On that note, I'll update my code soon in the OP.)

2. Sure, some players do prefer higher recoil. But I'm attempting to grade the magnitude of recoil inherent to each gun, not the preferences of players; doing that would simply be an exercise in polling. The fact that some people prefer to use guns with higher recoil does not change how much recoil those guns have and should not affect a weapon's ranking.

3. That's exactly why the ACR (without Kick) should score poorly. You said it yourself: there's very little chance that a stock ACR will return to center after the first shot and will thus tend to have a large deviation in shot placement. You'll notice that the ACR with Kick ranks as one of the best weapons in the Assault Rifles class in terms of recoil, and the recoil plots suggest that that is fair.

4. This seems to contradict your original statement. Those plots show a 10 000-fold simulation of firing 10-round bursts. They serve to illuminate the weapon's expected behaviour but in no way are these recoil plots a prescription for how the gun must behave in a given firefight. Maybe I am misreading what you're saying -- and apologize in advance if I am -- but it sounds like you believe that the weapon should, upon reaching the "boundary" of the recoil plot, move back towards center to "fill out" the rest of the pattern depicted in the recoil plot. If that is what you meant, then it is patently wrong. If you get unlucky and your gun fires three shots with near-maximum ViewKick velocities, the game will make no special effort to return your aim back to center any more than the CenterSpeed already does in order to achieve some sort of predefined pattern; you're going to have to deal with the fact that your weapon is far off-target on your own.

5. Of course. But again, I don't aim to rank weapons on preference, just raw deviation of their average shot placement. I'm not sure what you mean by "still the guns would have to be put in groups of how they work." Every weapon is governed by the same recoil model, and so it should be logical to subject them all to the same method of grading.

Perhaps you mistook this all as an attempt to "preach" which gun has the "best" recoil. I assure you that that is not the case. Even if a particular gun scores very well within this scheme, it is always up to the player to judge for themselves whether the gun works for them. We don't regard the Desert Eagle as the best pistol just because it has the lowest time-to-kill in the class. We all know it's the joke gun of its class, and we know this from experience. Nonetheless, the fact remains that it does have the best TTK in the Pistol class. Ranking these weapons is done only for informative purposes, much like someone might rank the guns on TTK. I expect players will always choose weapons that they know work for them regardless of any statistic.

[probaddie]

I've just had another epiphany that stems from the realization that led to me discarding the Euclidean metric: the biased statistics are flawed -- but not fatally. Here's the problem:

I introduced the vertical bias into the taxicab metric by simply multiplying the horizontal variance by the bias ratio (height-to-width) of 2.77. (Yes, the horizontal variance. That is not a typo.) The reasoning was that the horizontal variance should be punished -- i.e. should contribute more to the overall score* -- 2.77 times as much as vertical recoil. Again, remember that the variance is a "squared distance." But the bias ratio is not; it was derived using real measurements of average human proportions. So we have a vertical bias ratio that represents "real" distances being used to modify variables that represent "squared" distances. In essence, horizontal recoil was only being punished by a factor of 1.66, the square root of 2.77. What I should have done is to multiply the horizontal variance by the square of 2.77, i.e. 7.67. This would effectively punish horizontal recoil in the intended ratio of 2.77.

So for the fifth draft it's a step back. Later today (when I get back to my home computer) I will post one spreadsheet that contains both the unbiased and biased taxicab metrics, this time with the proper ratio embedded into the code. I will also include statistics for Rapid Fire LMGs -- I surprised no-one called me out on missing those -- and will include a warning for those "impossible" double attachment combinations, with emphasis on their effect on the LMG rankings.

I know some of you disagree with using a biased statistic on principle, but those of you just looking for the statistic which agrees most with your intuition will probably have something to say. Barring the discovery of other kinks in the code, I will expect that we will be able to make a final decision as to which statistic is best. (That or you can scream at me for dropping the Euclidean metric. Either or.)

* When I say "contribute to the score", I mean contribute to the penalty that brings it down from a perfect 100.

Edit: I somehow managed to copy my post and paste it onto itself. Don't ask.

[banana]

i predict its going to be the new biased taxicab

[probaddie]

May 30, 2012 17:43:05 GMT -5 banana said:

i predict its going to be the new biased taxicab

We'll see . I'm caught up in tweaking the code again (forgot about RF Mk14 as well as RF LMGs), so this may take a bit more time. I'm revising the code to deal in standard deviations instead of variances (en.wikipedia.org/wiki/Standard_deviation) as well as give percentile rankings to go with the rankings. Using standard deviations just avoids all the problems that cropped up earlier. Stay tuned.

[banana]

Sounds promising

[probaddie]

Here it is (finally), the fifth draft:

Unbiased taxicab: docs.google.com/spreadsheet/ccc?key=0AnFeORnwotj_dHRoRnU3QjByNTlFbWRzRllLUWZIb2c#gid=1
Biased taxicab: docs.google.com/spreadsheet/ccc?key=0AnFeORnwotj_dEc2a053c2paQWlObTZZLXoyRlJiWmc#gid=0

I've also uploaded the current version of my code (Mathematica 8.0.1) for those interested:
docs.google.com/document/d/1StDNw_DiTH74wjJGzBkYTDD2jAKwQfbobBZZnHMEhmQ/edit

Let me know what you think.

[Indy_Bones]

I would let you know but I can only access the Biased taxicab, the other is blocked and requires access be provided...

That being said, I can't agree with the biased version so far, I just hope the unbiased is significantly different without the bias.

[probaddie]

May 31, 2012 10:10:35 GMT -5 Indy_Bones said:

I would let you know but I can only access the Biased taxicab, the other is blocked and requires access be provided...

That being said, I can't agree with the biased version so far, I just hope the unbiased is significantly different without the bias.

Fixed. You should be able to access both sheets.

Update: I've edited the OP to be more friendly to people who may just now be finding this thread, so that they may contribute without having to read all of its pages.

[Indy_Bones]

Thanks for the update, the unbiased one is much better IMHO and a fairer representation of the guns in-game.

I've copied the info and re-organsied the rankings to remove 'impossible' combinations and I'm going to try testing this further tomorrow if I get chance.

[mmacola]

Psijaka's thread link doesn't work, at least for me.

[probaddie]

Jun 2, 2012 20:26:28 GMT -5 mmacola said:

Psijaka's thread link doesn't work, at least for me.

Fixed... again. (It seems to break randomly for some reason?) I've not forgotten about this by the way. I've just been very busy, and my gaming time has been spend actually playing MW3 to decide how I feel about the stats myself.

For me, the G36C vs. ACR is where this thing will be decided. I'll be using those guns exclusively whenever I play. I will come back on Monday at the latest to discuss my own opinions and we can try to decide a winner (stat-wise).

[Marvel4]

IMO, the biased formula is better. The G36C with Kick feels much more accurate than the ACR with Kick.

[banana]

I agree

[Indy_Bones]

Just goes to show how much personal opinion affects things like this.

I like the G36C (even if I have to use an RDS) but even with kick I don't think it stands up to the ACR for accuracy.

But maybe that's the really good thing about there being 2 versions - pick the one which suits you best.

[asasa]

I agree. I think its far from close as accurate. I had an instance today where I had a good half second of extra firetime on a guy, closish range, -- I dropshotted, but the recoil was so high and random that i went from below his feet to above his head. Not a single hit. That is why it cant be on par with the ACR. Pulled out the ACR after that and went from 0-3 to 9-3 in the last 2 rounds of the S&D. That gun is so sickeningly OP I feel dirty using it :/

[probaddie]

Jun 3, 2012 5:00:14 GMT -5 Marvel4 said:

IMO, the biased formula is better. The G36C with Kick feels much more accurate than the ACR with Kick.

(Let me preface everything I say by reminding everyone that I am a PC player.)

I'm of the same opinion. At short to medium ranges it feels like a wash, but at long ranges the ACR's moderate horizontal recoil put me off-target far more than the G36C's vertical recoil (both guns with Kick). The ACR definitely feels like a better weapon overall, and generally will take more shots to kill at a given range. This may contribute to a feel of better recoil management because you're not having to keep your aim on-target for as long.

Just goes to show how much personal opinion affects things like this.

I like the G36C (even if I have to use an RDS) but even with kick I don't think it stands up to the ACR for accuracy.

But maybe that's the really good thing about there being 2 versions - pick the one which suits you best.

And, perhaps, your platform. I'm only hypothesizing, here, but I imagine that, once the G36C's recoil takes you off-target -- thus removing the benefit of auto-aim -- it's probably much harder to return your aim on-target. This problem is probably worse with lower sensitivities. Using a mouse with a large mousepad, I simply adjusted how quickly I pulled down to match the G36C's recoil.

I agree. I think its far from close as accurate. I had an instance today where I had a good half second of extra firetime on a guy, closish range, -- I dropshotted, but the recoil was so high and random that i went from below his feet to above his head. Not a single hit. That is why it cant be on par with the ACR. Pulled out the ACR after that and went from 0-3 to 9-3 in the last 2 rounds of the S&D. That gun is so sickeningly OP I feel dirty using it :/

Just to be clear: you were using both guns with Kick, yes?

All this opinion notwithstanding, I believe that, without being able to come to a consensus, we should use the unbiased stats simply on philosophical grounds -- even to my own chagrin. I believe that, If we can't agree which model represents the usefulness of the weapons realistically, then the statistic should at least be the most mathematically sound representation of the weapons' raw deviation from center.

[banana]

Jun 4, 2012 9:42:50 GMT -5 probaddie said:

Jun 3, 2012 5:00:14 GMT -5 Marvel4 said:

IMO, the biased formula is better. The G36C with Kick feels much more accurate than the ACR with Kick.

(Let me preface everything I say by reminding everyone that I am a PC player.)

I'm of the same opinion. At short to medium ranges it feels like a wash, but at long ranges the ACR's moderate horizontal recoil put me off-target far more than the G36C's vertical recoil (both guns with Kick). The ACR definitely feels like a better weapon overall, and generally will take more shots to kill at a given range. This may contribute to a feel of better recoil management because you're not having to keep your aim on-target for as long.



And, perhaps, your platform. I'm only hypothesizing, here, but I imagine that, once the G36C's recoil takes you off-target -- thus removing the benefit of auto-aim -- it's probably much harder to return your aim on-target. This problem is probably worse with lower sensitivities. Using a mouse with a large mousepad, I simply adjusted how quickly I pulled down to match the G36C's recoil.

I agree. I think its far from close as accurate. I had an instance today where I had a good half second of extra firetime on a guy, closish range, -- I dropshotted, but the recoil was so high and random that i went from below his feet to above his head. Not a single hit. That is why it cant be on par with the ACR. Pulled out the ACR after that and went from 0-3 to 9-3 in the last 2 rounds of the S&D. That gun is so sickeningly OP I feel dirty using it :/

Just to be clear: you were using both guns with Kick, yes?

All this opinion notwithstanding, I believe that, without being able to come to a consensus, we should use the unbiased stats simply on philosophical grounds -- even to my own chagrin. I believe that, If we can't agree which model represents the usefulness of the weapons realistically, then the statistic should at least be the most mathematically sound representation of the weapons' raw deviation from center.

that would be great if we were attacking square targets which we aren't so the biased formula is the most mathematically sound

[probaddie]

Jun 5, 2012 1:00:24 GMT -5 banana said:

Jun 4, 2012 9:42:50 GMT -5 probaddie said:

(Let me preface everything I say by reminding everyone that I am a PC player.)

I'm of the same opinion. At short to medium ranges it feels like a wash, but at long ranges the ACR's moderate horizontal recoil put me off-target far more than the G36C's vertical recoil (both guns with Kick). The ACR definitely feels like a better weapon overall, and generally will take more shots to kill at a given range. This may contribute to a feel of better recoil management because you're not having to keep your aim on-target for as long.



And, perhaps, your platform. I'm only hypothesizing, here, but I imagine that, once the G36C's recoil takes you off-target -- thus removing the benefit of auto-aim -- it's probably much harder to return your aim on-target. This problem is probably worse with lower sensitivities. Using a mouse with a large mousepad, I simply adjusted how quickly I pulled down to match the G36C's recoil.



Just to be clear: you were using both guns with Kick, yes?

All this opinion notwithstanding, I believe that, without being able to come to a consensus, we should use the unbiased stats simply on philosophical grounds -- even to my own chagrin. I believe that, If we can't agree which model represents the usefulness of the weapons realistically, then the statistic should at least be the most mathematically sound representation of the weapons' raw deviation from center.

that would be great if we were attacking square targets which we aren't so the biased formula is the most mathematically sound

I think you're confusing "most mathematically sound" with "most practically useful." We are still arguing over which statistic is the latter, but if we are looking for the statistic that best measures "how spread out is the recoil plot with respect to the center," then the unbiased statistic is the most mathematically sound. It simply follows from the fact that we have not introduced a bias into it.

My only point here is that, if we cannot agree which statistic best represents the usefulness of the guns, then we should choose the statistic that best represents the spread of the plot, i.e. the unbiased statistic.

As an aside, I agree with you that that biased statistic is more useful. But my opinion is only one of many.

(Also, did you get a response from XboxAhoy regarding the GunKick numbers!?)

[banana]

no not yet but his last post was on sunday and i sent him the message on monday so its possible that he just didnt get on yet

[banana]

also btw since our ratio is a guestimate wouldnt a good way to find out the correct one be with using the crosshair to measure it but then again ours is probably close enough

[banana]

Are you planning to make one of these for black ops

[probaddie]

Jun 11, 2012 20:16:46 GMT -5 banana said:

Are you planning to make one of these for black ops

Yes; I don't see why I shouldn't. I'm guessing you didn't have any luck with XboxAhoy?

[probaddie]

Jun 11, 2012 20:28:13 GMT -5 probaddie said:

Jun 11, 2012 20:16:46 GMT -5 banana said:

Are you planning to make one of these for black ops

Yes; I don't see why I shouldn't. I'm guessing you didn't have any luck with XboxAhoy?

Sorry, did you mean BO or BO2?

[banana]

actuallly both but black ops 2 would obviously have to wait until launch or whenever we get the stats

[probaddie]

Jun 12, 2012 8:45:24 GMT -5 banana said:

actuallly both but black ops 2 would obviously have to wait until launch or whenever we get the stats

I can do that. Maybe producing results for BO will help settle the biased vs. unbiased statistic debate. Since GunKick is a negligible mechanic in BO, hopefully the results will be more, let's say, "agreeable" with personal experience. (Here's hoping some of you have played BO lately!)

I realize that I forgot statistics for the Rapid Fire M16A4 and Type 95. I'll update these in the fifth draft later today.

(chosenone: maybe you could try using a non-Facebook approach to getting Stuart's attention. I imagine it's easy for him to ignore most of what comes his way through Facebook, as popular as he is on Youtube. Maybe try reaching him through [e-mail]? )

[Marvel4]

The handguns in Black Ops also have gun kick, except the ASP and Python.

[TheHawkNY]

I would not hold out hope for him providing any statistics, as he is almost certainly under NDA. Most people would not jump at the opportunity to break their NDA and potentially anger their employer for no tangible benefit to themselves.

[Marvel4]

I'm sure he's allowed to release the stats, since he shows them in his videos.

[probaddie]

Jun 12, 2012 13:14:58 GMT -5 Marvel4 said:

The handguns in Black Ops also have gun kick, except the ASP and Python.

Fair enough. That is a far smaller subset of guns than that which is affected in MW3, so a footnote for that subset of weapons should be sufficient.

I'm sure he's allowed to release the stats, since he shows them in his videos.

That would be my argument. I think he would only be in trouble if he provided us with some way to decode the weapon files themselves, which we should never expect from him. Remember: IW themselves released the ViewKick numbers to us for MW3, though they are hidden behind the same encryption that the GunKick numbers are.

[banana]

Jun 12, 2012 13:06:41 GMT -5 probaddie said:

Jun 12, 2012 8:45:24 GMT -5 banana said:

actuallly both but black ops 2 would obviously have to wait until launch or whenever we get the stats

I can do that. Maybe producing results for BO will help settle the biased vs. unbiased statistic debate. Since GunKick is a negligible mechanic in BO, hopefully the results will be more, let's say, "agreeable" with personal experience. (Here's hoping some of you have played BO lately!)

I realize that I forgot statistics for the Rapid Fire M16A4 and Type 95. I'll update these in the fifth draft later today.

(chosenone: maybe you could try using a non-Facebook approach to getting Stuart's attention. I imagine it's easy for him to ignore most of what comes his way through Facebook, as popular as he is on Youtube. Maybe try reaching him through [e-mail]? )

ok