Jump to content


Dispersion values are 3 times worse than claimed.

Proof

  • Please log in to reply
110 replies to this topic

Stevies_Team #21 Posted 11 September 2019 - 03:28 PM

    Second Lieutenant

  • Player
  • 11000 battles
  • 1,025
  • Member since:
    07-14-2016

Plus it's not really about the bits they tell you about

 

It's about the bits they don't tell you about

 

Its the mushroom system practiced by governments and corporations across the world

 

Keep 'em in the dark and feed 'em poop

 

 

Like I say, just play for fun and shoot tonks, it's a fun game


Edited by If_I_Die_You_Die_Too, 11 September 2019 - 03:35 PM.


Dorander #22 Posted 11 September 2019 - 03:42 PM

    Major General

  • Player
  • 21025 battles
  • 5,767
  • Member since:
    05-07-2012

View PostKarasu_Hidesuke, on 11 September 2019 - 02:10 PM, said:

Well, there definitely is something strange there, because even when playing with the same tank it varies a lot from battle to battle. I know this is my 'tin foil hat moment', but I could swear there is some random generator that determines your overall RNG values for each separate battle meaning sometimes you get good RNG, and sometimes you just get royally shafted.

 

The problem is that people are notoriously bad at predicting how many times improbable results actually occur and have a disproportionate focus on negative experiences compared to positive ones. We tend to think that high enough probabilities should always work, as if they were 100%, but they aren't. When they do work, we think "Well of course it did, it's a high chance" and when it doesn't we just go "How the *bleep* did that happen, what are the odds???" while we knew the odds all along: they were presented to us as a percentage.

 

Consider this for a second, not what the right answer is but what your personal first inclination would be: Would you rather be in a group that has 1000 euros and a 90% chance to win another 1000 euros, or would you rather be in a group that has 2000 euros and has a 10% chance to lose 1000 euros?

 

The odds of those scenarios are completely identical. Yet I can guarantee you that most people from the second group who lost 1000 euros, are going to feel a hell of a lot worse and curse their luck a whole lot more than most of the people who didn't win 1000 euros from the first group, because the people from the first group never lost something they actually had. The second-worst feeling group of people are going to be the ones from the first group who didn't win but felt like they ought to have won, who thought success was guaranteed, because they feel robbed of something they genuinely thought they had, even if in reality they never did.



Laukaus #23 Posted 11 September 2019 - 04:04 PM

    Corporal

  • Player
  • 10053 battles
  • 146
  • Member since:
    03-27-2013

View PostVela_ZG, on 11 September 2019 - 12:47 PM, said:

Yes, that's why I did this also. The accuracy is completely wrong. 

 

Remember that this is a r_____n "game". Dont expect anything to be right. 



XxKuzkina_MatxX #24 Posted 11 September 2019 - 05:13 PM

    Brigadier

  • Player
  • 53231 battles
  • 4,987
  • Member since:
    04-02-2016

The data in the first post checks out. The machine gun ball (Kugelblende) of the tiger is around 30-35cm in diameter. The effective accuracy (crew, equipment, etc.) of the CAX was not tampered with. The scaling of the tiger in WOT is correct to the real life dimensions.

 

I believe the issue here is comparing the actual size vs. the apparent size of the tank. This machine gun ball gets smaller as the distance increase but the aiming circle doesn't because that's your reference or constant while aiming.

 

From the wiki...

 

"In other words, for a gun with 0.32m effective accuracy at 100m, 95.45% of all shots will land within 0.32m of the center of your aim at that distance. Dispersion amount increases linearly with distance, i.e. 0.32m effective accuracy at 100m translates to 0.64m at 200m and 1.28m at 400m."

 

"The aiming circle (also called the dispersion circle, dispersion indicator or reticle) describes the area of 2 standard deviations 2σ from your aim point."

 

The diameter of that ball isn't 30 cm at a 100m distance while your aiming circle is constant in size as long as you're stationary. Think of that accuracy definition as the minutes of angle used in scopes. MOA has different sizes at various distances.


Edited by XxKuzkina_MatxX, 11 September 2019 - 06:35 PM.


Vela_ZG #25 Posted 11 September 2019 - 07:22 PM

    Warrant Officer

  • Player
  • 27859 battles
  • 735
  • [CBS87] CBS87
  • Member since:
    10-09-2013

Interesting approach XxKuzkina_MatxX! I understand what you mean, but then again, something is not right. Since my accuracy is 0.29 and at 100m the aiming circle should be 0.554m because our shells fly randomly within that circle. I do understand that what you're saying could be right, but I also see that the accuracy stats are completely wrong. Here, I fixed my radius mistake from the first post, you can see it in attachments. The green circle is the actual in game reticle size, and according to WG it should be the size of the blue circle, red is the reference.

 BTW, my FoV is stock at 95.

Lets not make this thread about RNG please, let's talk about measurable facts that we can test. It's not about how many times you can hit the bull's eye, but about the size of the circle with regards to the target. And the difference is huge!

Attached Files

  • Attached File   WOT dispersion radius.png   5.35MB
  • Attached File   shot_006.jpg   328.69K


XxKuzkina_MatxX #26 Posted 11 September 2019 - 07:39 PM

    Brigadier

  • Player
  • 53231 battles
  • 4,987
  • Member since:
    04-02-2016

The difference you see is a problem of interpreting the numbers not the numbers themselves. The machine gun ball which is, let's say, 32cm in diameter will get smaller on your screen as the distance between you and the tank gets bigger. Your aim circle will not change, it's the constant part of the equation. Notice the wording in its definition...

 

"95.45% of all shots will land within 0.32m of the center of your aim at that distance."

 

At that distance, which is 100m in your test, the apparent diameter of that ball (what you can see of it) is smaller than 0.32cm. And again your reticle/aiming circle didn't change in size. It's 0.32cm over there not where you're sitting and making these measurements. I hope that was clear enough, i know it's a bit confusing but it does what it says on the tin.



Vela_ZG #27 Posted 11 September 2019 - 07:54 PM

    Warrant Officer

  • Player
  • 27859 battles
  • 735
  • [CBS87] CBS87
  • Member since:
    10-09-2013
We should do the tests. I'm sure the shells will land within the green circle, not blue. 

XxKuzkina_MatxX #28 Posted 11 September 2019 - 08:03 PM

    Brigadier

  • Player
  • 53231 battles
  • 4,987
  • Member since:
    04-02-2016
Indeed you should. Take a ruler or a vernier caliper and try to measure how tall the buildings around you from your window. You'll find that either we live in a very small world or the scale on the ruler is not correct. :)

Edited by XxKuzkina_MatxX, 11 September 2019 - 08:09 PM.


Mannanan #29 Posted 11 September 2019 - 08:06 PM

    Major

  • Player
  • 28698 battles
  • 2,968
  • [KHIL] KHIL
  • Member since:
    07-28-2017
I would simply say that when 3D scene with a tank is rendered perspektive is applied to it, hence the tank is rendered smaller. But perspektive is not applied to rectile.

Vela_ZG #30 Posted 11 September 2019 - 08:16 PM

    Warrant Officer

  • Player
  • 27859 battles
  • 735
  • [CBS87] CBS87
  • Member since:
    10-09-2013
I understand, but the size of the reticle SHOULD change with the distance of the target. Because our shells hit inside that circle! Do you not understand that? If the circle doesn't change in size, our accuracy is completely off! There could be any number of reasons why the circle is not in sync with the stats, but this only shows how completely pointless any of the stats in this game are.
19:18 Added after 1 minute

View PostXxKuzkina_MatxX, on 11 September 2019 - 08:03 PM, said:

Indeed you should. Take a ruler or a vernier caliper and try to measure how tall the buildings around you from your window. You'll find that either we live in a very small world or the scale on the ruler is not correct. :)

I understand perspective. I own a company that does 3D prototyping. I create 3D models daily.



24doom24 #31 Posted 11 September 2019 - 08:20 PM

    Staff Sergeant

  • Player
  • 8248 battles
  • 379
  • [WW3] WW3
  • Member since:
    10-20-2012

View PostVela_ZG, on 11 September 2019 - 11:44 AM, said:

I did some measurements and tests and came to a conclusion that the claimed dispersion values are 3 times greater than WG claims. I took the Tiger1 as a reference since it's dimensions are easily accessible on the net. In the attachments you'll find the measurements. I measured with Centurion AX that has 0.29 accuracy against a Japanese Tiger. The model is identical to German Tiger1. 

I parked 100m away from it and let the gun fully aim, then took the screenshot. As you can see Centurion's actual accuracy is around 0.90 give or take, at 100m.

It would be interesting to do some comparisons against different tanks with the same accuracy on paper and see what the difference will be, if any.

In any case, the numbers are wrong with a ratio of 3:1. 

So now we have a reference point, and thats Tiger1's MG port thats roughly 30cm in diameter. Who ever wishes to do the testing is welcome. Maybe we can find out some other things and possibly improve them.

 

Accuracy stats are kinda pointless with the amount of RNG that ruins it. Your shot in the first screenshot would have hit the side of the turret and ricocheted:trollface:



Dorander #32 Posted 11 September 2019 - 08:29 PM

    Major General

  • Player
  • 21025 battles
  • 5,767
  • Member since:
    05-07-2012

View PostVela_ZG, on 11 September 2019 - 07:16 PM, said:

I understand, but the size of the reticle SHOULD change with the distance of the target. Because our shells hit inside that circle! Do you not understand that? If the circle doesn't change in size, our accuracy is completely off! There could be any number of reasons why the circle is not in sync with the stats, but this only shows how completely pointless any of the stats in this game are.

 

They're not pointless if they are relatively consistent with each other and behave consistently. I don't know anyone who calculates distance and dispersion on the fly in battle. I, and I suspect many other people, simply aim optimally and fire.

 

It's easy to see that your reticle fully aimed is consistent for each zoom factor. It's easy to see that target sizes are rendered consistently. Stats related to consistent behaviour aren't pointless, even when they don't reference real life concepts. This is why you can practice playing a tank, you get better at playing it when you learn how it consistently behaves. Not because you're doing mathematics mid-battle. Useful for post analysis, sure. Need to acknowledge that stats in this game aren't always straightforward and depend on mutual interaction and that experiments such as these are good to evaluate that, absolutely. Conclude that they're therefor pointless? That's leaping to conclusions.



burbage1 #33 Posted 11 September 2019 - 08:59 PM

    Corporal

  • Player
  • 964 battles
  • 129
  • Member since:
    09-09-2017
In this game it's conversion, not dispersion. Accuracy is weighted to the centre of the aiming circle rather than the perimeter. Totally unrealistic, but good to know. It's an arcade game after all.

Vela_ZG #34 Posted 11 September 2019 - 09:18 PM

    Warrant Officer

  • Player
  • 27859 battles
  • 735
  • [CBS87] CBS87
  • Member since:
    10-09-2013
OK, I know, I accept that. But the circle should be much smaller. Let them fly all around but IN that (blue) circle.

XxKuzkina_MatxX #35 Posted 11 September 2019 - 09:55 PM

    Brigadier

  • Player
  • 53231 battles
  • 4,987
  • Member since:
    04-02-2016

View PostVela_ZG, on 11 September 2019 - 11:16 PM, said:

I understand, but the size of the reticle SHOULD change with the distance of the target. Because our shells hit inside that circle! Do you not understand that? If the circle doesn't change in size, our accuracy is completely off! There could be any number of reasons why the circle is not in sync with the stats, but this only shows how completely pointless any of the stats in this game are.

 

I understand perspective. I own a company that does 3D prototyping. I create 3D models daily.

 

Thank you for taking the time to do this test though. I certainly learned some things from that. :)



Vela_ZG #36 Posted 11 September 2019 - 09:59 PM

    Warrant Officer

  • Player
  • 27859 battles
  • 735
  • [CBS87] CBS87
  • Member since:
    10-09-2013
I'll try to do some more tests with different tanks, maybe we'll come up with something. We'll see.

Karasu_Hidesuke #37 Posted 11 September 2019 - 10:03 PM

    Brigadier

  • Player
  • 16912 battles
  • 4,428
  • [I-S-L] I-S-L
  • Member since:
    12-03-2013

View PostVela_ZG, on 11 September 2019 - 08:18 PM, said:

OK, I know, I accept that. But the circle should be much smaller. Let them fly all around but IN that (blue) circle.

 

When using the sniper mode, it seems to me as if the aiming/dispersion circle changes size inconsistently. Going from the penultimate setting to the maximum setting does not seem to give any additional benefit.


Edited by Karasu_Hidesuke, 11 September 2019 - 10:04 PM.


CmdRatScabies #38 Posted 11 September 2019 - 10:04 PM

    Major General

  • Player
  • 38527 battles
  • 5,098
  • [-MM] -MM
  • Member since:
    10-12-2015

View PostVela_ZG, on 11 September 2019 - 09:59 PM, said:

I'll try to do some more tests with different tanks, maybe we'll come up with something. We'll see.

I think  WG published some data on accuracy & how shots land within the circle.  If I recall correctly there is a deadish spot in the centre that was introduced a few years ago to reduce the accuracy of snap shots.  Old thread on it somewhere with the the adjusted normal curve shown - I think Baldrick posted on it and might remember.



Pansenmann #39 Posted 11 September 2019 - 10:10 PM

    Field Marshal

  • Player
  • 36091 battles
  • 13,687
  • [WJDE] WJDE
  • Member since:
    08-17-2012

you mean this thread?

 

http://forum.worldoftanks.eu/index.php?/topic/652684-how-dispersion-and-aimtime-actually-work-maths-inside/

 

or this

 

http://forum.worldoftanks.eu/index.php?/topic/646723-the-rng-distribution-calculated-dmg-accuracy-updated/


Edited by Pansenmann, 11 September 2019 - 10:14 PM.


Vela_ZG #40 Posted 11 September 2019 - 10:38 PM

    Warrant Officer

  • Player
  • 27859 battles
  • 735
  • [CBS87] CBS87
  • Member since:
    10-09-2013

View PostCmdRatScabies, on 11 September 2019 - 10:04 PM, said:

I think  WG published some data on accuracy & how shots land within the circle.  If I recall correctly there is a deadish spot in the centre that was introduced a few years ago to reduce the accuracy of snap shots.  Old thread on it somewhere with the the adjusted normal curve shown - I think Baldrick posted on it and might remember.

This is all good but I'm interested in relation between the size of the aiming circle and the distance to the target. We all know how dispersion within the circle works, about aim time and what not, I'm interested in ratio of given accuracy and the size of the surface the dispersion is applied to. If this is off, everything changes, the gun stats mean nothing. And this is before RNG comes into play.







Also tagged with Proof

2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users