Jump to content


Bugged or Intentionally deceptive penetration? WG, please respond.

Penetration

  • Please log in to reply
24 replies to this topic

Lesser_Spotted_Panzer #1 Posted Mar 15 2018 - 13:50

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

4tankers released a video yesterday that demonstrated that the published penetration numbers are completely bogus. There wasn't much discussion on the thread perhaps because like myself, people did not have time to watch the video at that time. Also, the title of that thread did not specifically highlight the egregious problem that the research 4tankers found. Therefore, I am creating a new thread to specifically discuss these findings. I really hope WG will respond to this, offering an explanation.


 

Here is the original topic:

http://forum.worldoftanks.com/index.php?/topic/572701-tanking-w-science-pentesting/page__p__11550951#entry11550951


 

The video set out to prove that the penetration of light tanks over distance is worse than other tanks. This it did prove. However, there was some strange data in the first set of tests, so further testing was done.


 

The further tests then showed that light tanks also get worse penetration (for the same published penetration number) over short distances too. However, even this data was inconsistent. So further testing was done.


 

The further tests showed that although light tanks do get worse penetration (for the same published penetration number), every tank appears to have a different hidden penetration modifier. So two heavy tanks for example, with the same penetration value and same ammo type, will in reality have very different actual penetration values.


 

This makes the numbers published by WG total BS. Now whether that is by accident or by design, we don't know. If it is by accident, then WG should thank 4tankers and fix the problem. If it is by design, then you know they will not respond, nor change anything. Or possibly (but unlikely), the testing methodology was wrong some how.


 

I'm sure most people will accept any stat that a tank has, so long as the number is real. The actual numbers don't change how a tank feels to play or how it performs, however it is a number we look at when making decisions in this game, or purchasing premium tanks. So why not just be honest? Why try to deceive your customers this way?


 

I'm also wondering if these hidden modifiers change over time, either automatically or manually? Perhaps if a tank is performing too well or badly, they will tweak this number.


 


 


 


 


 



t8z5h3 #2 Posted Mar 15 2018 - 14:01

    Captain

  • -Players-
  • 38512 battles
  • 1,097
  • [FUSON] FUSON
  • Member since:
    04-23-2013

you know the published ones are for 100M and is still the 25% RNG for Pen and there is also a pen drop off for long shots (i don't know when that happens (300M? maybe) 

it's working as designed 



Lesser_Spotted_Panzer #3 Posted Mar 15 2018 - 14:05

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

View Postt8z5h3, on Mar 15 2018 - 08:01, said:

you know the published ones are for 100M and is still the 25% RNG for Pen and there is also a pen drop off for long shots (i don't know when that happens (300M? maybe)

it's working as designed

 

Yes of course. RNG was factored out of the testing by repeating the tests 100s of times. The testing was done both at 450m and 50m. Please watch the video and you will see.

24_inch_pythons #4 Posted Mar 15 2018 - 14:37

    Major

  • Players
  • 25046 battles
  • 4,845
  • [-P-] -P-
  • Member since:
    03-28-2012
Way too many variables to try and isolate just the pen. You have dispersion and RNG - and no 100 shots will not remove RNG. 100,000 I might think about it. But simply dispersion will completely negate this test. To test pen you need to hit the EXACT same spot to get the same armor value.

Lesser_Spotted_Panzer #5 Posted Mar 15 2018 - 14:55

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

View Post24_inch_pythons, on Mar 15 2018 - 08:37, said:

Way too many variables to try and isolate just the pen. You have dispersion and RNG - and no 100 shots will not remove RNG. 100,000 I might think about it. But simply dispersion will completely negate this test. To test pen you need to hit the EXACT same spot to get the same armor value.

 

I don't think so.


 

Shooting the side of a Japanese heavy from 50m is unlikely affected by dispersion to any significant degree.


 

The more you repeat the tests, the more you reduce the effect of RNG. After 200 tests, you have probably reduced the effect of it to a few percent. If you run the test 100,000 times, you have reduced it to a fraction of a percent. However, the test results showed that the penetration difference between some tanks were 10's of percent different, way more than RNG alone.


 


 



CapnNoahJoJo03 #6 Posted Mar 15 2018 - 15:12

    First lieutenant

  • -Players-
  • 16026 battles
  • 839
  • [HSOLO] HSOLO
  • Member since:
    04-08-2016
Sir, would you like to buy my patented tin foil hat design? Only 19.95! Get em while supplies last!

the_dude_76 #7 Posted Mar 15 2018 - 15:22

    Major

  • Players
  • 32529 battles
  • 3,328
  • [GSRM] GSRM
  • Member since:
    12-27-2011

View PostLesser_Spotted_Panzer, on Mar 15 2018 - 13:55, said:

 

The more you repeat the tests, the more you reduce the effect of RNG. After 200 tests, you have probably reduced the effect of it to a few percent. 

 

 

You have probably reduced the effect of it to a few percent. Probably... Yeah, sure sounds scientific to me

 

Of course the reality is that a set of a few hundred attempts isn't going to produce a verifiable result. It is an interesting trend but it hardly proves anything 

 

Block Quote

 Shooting the side of a Japanese heavy from 50m is unlikely affected by dispersion to any significant degree.

 

I can't watch the video right, so how has it been determined that the discrepancy is within the penetration values as opposed to the armor values??

 



Lesser_Spotted_Panzer #8 Posted Mar 15 2018 - 15:34

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

View Postthe_dude_76, on Mar 15 2018 - 09:22, said:

 

You have probably reduced the effect of it to a few percent. Probably... Yeah, sure sounds scientific to me

 

Of course the reality is that a set of a few hundred attempts isn't going to produce a verifiable result. It is an interesting trend but it hardly proves anything 

 

 

I can't watch the video right, so how has it been determined that the discrepancy is within the penetration values as opposed to the armor values??

 

 

The test was to compare different tanks with the same penetration value and the same ammo type shooting the same piece of armor. It doesn't matter what the absolute armor value is, or the absolute penetration value is, you would expect the guns to perform the same (within the margins of RNG). When one gun pens 30% more than another equal gun, something is not right.

Lesser_Spotted_Panzer #9 Posted Mar 15 2018 - 15:34

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

View PostCapnNoahJoJo03, on Mar 15 2018 - 09:12, said:

Sir, would you like to buy my patented tin foil hat design? Only 19.95! Get em while supplies last!

 

I would love to add this to my collection.

 

Can I sell you some sand, so you can bury your head in it?


Edited by Lesser_Spotted_Panzer, Mar 15 2018 - 15:37.


ArmorStorm #10 Posted Mar 15 2018 - 16:01

    Major

  • Players
  • 35631 battles
  • 7,849
  • [F__R] F__R
  • Member since:
    08-12-2011

Not to minimize all the hard work, but if I came on here stating that anything I did for 200 battles proved something, I would be called out rightfully for “small sample size”. 

 

At best there is a bit of a trend showing that would need thousands of rounds fired to confirm.  Which isn’t going to happen. 



Lesser_Spotted_Panzer #11 Posted Mar 15 2018 - 16:44

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

You naysayers have forced me to do some math. Oh, how I hate math. I hate statistics even more. Especially when they prove you wrong:(


 

So I calculated the margin of error on some of the tests performed in the video.


 

For the short range tests, I used a 99% confidence level since the chances of any unexpected outliers is small (for example missed shots due to dispersion). Here are three cases where I extracted data from the video:


 

Case 1: LTTB (78% pens) vs. BC12T(71.7% pens)  firing at E100 at 50M, 100 shots. MOE = +/-11.2%

Case 2: T100 (60% pens) vs. t26e5 (51% pens) firing at Type 5 at 50m, 100 shots, MOE = +/-12.8%

Case 3: Sheridan (72% pens) vs. amx13 105 (62% pens) firing at Type 5 at 50m, 100 shots, MOE = +/-12.1%


 

For the long range shots, I used a 95% confidence level, since there is a greater chance that dispersion would result in hitting a different part of the armor, or actually missing. I'm only guessing this is an accurate reflection of what is happening, so this may not be correct.


 

Case 1: LTG (19% pens) vs. Tiger1 (25% pens) firing at IS7 at 435m, 100 shots. MOE = +/-8.3%

Case 2: T100 (32% pens) vs. T26E5 (44% pens) firing at IS7 at 435m, 100 shots. MOE = +/-9.5%


 

So in every case, the test results are within the margin of error. Meaning that the video shows no conclusive proof of anything suspicious. As the naysayers suggest, there is insufficient testing to prove anything, although the results do seem to suggest there is something fishy. More testing is necessary.


 

My apologies for poo-pooing the naysayers.

 

 



Gunadie #12 Posted Mar 15 2018 - 16:44

    Major

  • Players
  • 41787 battles
  • 4,876
  • Member since:
    08-20-2011

View PostLesser_Spotted_Panzer, on Mar 15 2018 - 04:50, said:

4tankers released a video yesterday that demonstrated that the published penetration numbers are completely bogus. There wasn't much discussion on the thread perhaps because like myself, people did not have time to watch the video at that time. Also, the title of that thread did not specifically highlight the egregious problem that the research 4tankers found. Therefore, I am creating a new thread to specifically discuss these findings. I really hope WG will respond to this, offering an explanation.


 

Here is the original topic:

http://forum.worldoftanks.com/index.php?/topic/572701-tanking-w-science-pentesting/page__p__11550951#entry11550951


 

The video set out to prove that the penetration of light tanks over distance is worse than other tanks. This it did prove. However, there was some strange data in the first set of tests, so further testing was done.


 

The further tests then showed that light tanks also get worse penetration (for the same published penetration number) over short distances too. However, even this data was inconsistent. So further testing was done.


 

The further tests showed that although light tanks do get worse penetration (for the same published penetration number), every tank appears to have a different hidden penetration modifier. So two heavy tanks for example, with the same penetration value and same ammo type, will in reality have very different actual penetration values.


 

This makes the numbers published by WG total BS. Now whether that is by accident or by design, we don't know. If it is by accident, then WG should thank 4tankers and fix the problem. If it is by design, then you know they will not respond, nor change anything. Or possibly (but unlikely), the testing methodology was wrong some how.


 

I'm sure most people will accept any stat that a tank has, so long as the number is real. The actual numbers don't change how a tank feels to play or how it performs, however it is a number we look at when making decisions in this game, or purchasing premium tanks. So why not just be honest? Why try to deceive your customers this way?


 

I'm also wondering if these hidden modifiers change over time, either automatically or manually? Perhaps if a tank is performing too well or badly, they will tweak this number.


 


 


 


 


 

 

Ready to bet 1000 gold that anyone with any position at WG will not respond.

Hidden values that WG uses to "balance" tanks in this game are nothing new, and its likely this applies to all aspects in this game even though you would assume they,

buy design, should be fixed.

I wouldn't be at all surprised is WG has "floating" penetration values assigned based on the opponent you fighting.

Now please stop all this chatter and buy some premium what-ever !



McChaffee #13 Posted Mar 15 2018 - 16:48

    Staff sergeant

  • Players
  • 10303 battles
  • 376
  • Member since:
    06-04-2011

View PostLesser_Spotted_Panzer, on Mar 15 2018 - 12:50, said:

The video set out to prove that the penetration of light tanks over distance is worse than other tanks. This it did prove.

 

This was confirmed in a Quickybaby video when light tanks were first coming out like a year ago. He obviously didn't do his research.



the_dude_76 #14 Posted Mar 15 2018 - 16:55

    Major

  • Players
  • 32529 battles
  • 3,328
  • [GSRM] GSRM
  • Member since:
    12-27-2011

View PostLesser_Spotted_Panzer, on Mar 15 2018 - 14:34, said:

 

 When one gun pens 30% more than another equal gun, something is not right.

 

Out of a sample of a couple hundred?? As I said, an interesting trend but it proves nothing. 

 

Block Quote

 

 I'm only guessing this is an accurate reflection of what is happening, so this may not be correct.

 

Translation- "I've wasted my time and proven nothing"



Maxus_Rennix #15 Posted Mar 15 2018 - 17:33

    Sergeant

  • Players
  • 22996 battles
  • 172
  • [DOWFL] DOWFL
  • Member since:
    06-05-2011

So basically...

Low sample size (not enough vehicles to confirm anything past a weak trend)

Low shots fired in the sample size per vehicle (100 shots is too low even if you were going by real gun statistics IRL)

No mention of the differences in crew training or equipment (changes aim time/dispersion)

No mention of the angle you are shooting at the target from (changes armor values, and needs to be a degree based value for accuracy in all 3 axis)

No mention of whether barrels were lined up at the same height (changes aiming angle thus armor values)

No mention of the ammo types used in tests (as they all have differing characteristics)

No mention of the data being recorded in video or by text by you (leaving only your word as an unreliable assertion/anecdotal piece of evidence)

No mention of why extrapolating such a small sample size would be accurate at all

 

Not to be that guy, but you don't even have half of these as a bare minimum. Why then would your tests be considered in any way admissible?

If you are doing proper testing you have to account for all variables to achieve a proper result. 

However your lack of accounting for said variables means at best that your data is only good to show a general trend and at worse that you would have to start all over again to get correct values to plot your data with.



Lesser_Spotted_Panzer #16 Posted Mar 15 2018 - 17:54

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

View PostMaxus_Rennix, on Mar 15 2018 - 11:33, said:

So basically...

Low sample size (not enough vehicles to confirm anything past a weak trend)

Low shots fired in the sample size per vehicle (100 shots is too low even if you were going by real gun statistics IRL)

No mention of the differences in crew training or equipment (changes aim time/dispersion)

No mention of the angle you are shooting at the target from (changes armor values, and needs to be a degree based value for accuracy in all 3 axis)

No mention of whether barrels were lined up at the same height (changes aiming angle thus armor values)

No mention of the ammo types used in tests (as they all have differing characteristics)

No mention of the data being recorded in video or by text by you (leaving only your word as an unreliable assertion/anecdotal piece of evidence)

No mention of why extrapolating such a small sample size would be accurate at all

 

Not to be that guy, but you don't even have half of these as a bare minimum. Why then would your tests be considered in any way admissible?

If you are doing proper testing you have to account for all variables to achieve a proper result.

However your lack of accounting for said variables means at best that your data is only good to show a general trend and at worse that you would have to start all over again to get correct values to plot your data with.

 

Some of your points are valid, but others are not. I did put a link where you can view the video yourself and see the answers to some of your points.

n00bfarmer #17 Posted Mar 15 2018 - 17:56

    First lieutenant

  • -Players-
  • 21 battles
  • 860
  • Member since:
    12-14-2017
If you mouse over the ammo in garage it will tell you the drop off in penetration values for close range, over 200 meters and over 500 meters away. Different tanks can have different drop off values. The 57 mm gun is a good example and premium apcr ammo has some of the highest loss of penetration over range than any other type of ammo. That 180 pen becomes around 120-130 at 500 meters and that's hardly better than using the standard AP ammo at that range. Also, if a tank uses apcr as it's standard ammo, the drop off in pen over range is less than a tank that fires apcr as it's premium ammo for that ammo type.

Maxus_Rennix #18 Posted Mar 15 2018 - 18:09

    Sergeant

  • Players
  • 22996 battles
  • 172
  • [DOWFL] DOWFL
  • Member since:
    06-05-2011

View PostLesser_Spotted_Panzer, on Mar 15 2018 - 10:54, said:

 

Some of your points are valid, but others are not. I did put a link where you can view the video yourself and see the answers to some of your points.

 

Every single point I made was valid, and as far as the video/text thing goes the point was that it wasn't properly documented, nor was the video provided of your testing. It's your credibility being put into question, as it is your "testing".

Edited by Maxus_Rennix, Mar 15 2018 - 18:10.


Lesser_Spotted_Panzer #19 Posted Mar 15 2018 - 18:20

    Major

  • Players
  • 77792 battles
  • 3,258
  • Member since:
    02-11-2012

View PostMaxus_Rennix, on Mar 15 2018 - 12:09, said:

 

Every single point I made was valid, and as far as the video/text thing goes the point was that it wasn't properly documented, nor was the video provided of your testing. It's your credibility being put into question, as it is your "testing".

 

 

 

I put a link in the first comment of this thread. If you follow that link you will see the video to which is being referred to.


 

All of your items below were essentially addressed in the video:


 

 

 

No mention of the differences in crew training or equipment (changes aim time/dispersion)

No mention of the angle you are shooting at the target from (changes armor values, and needs to be a degree based value for accuracy in all 3 axis)

No mention of whether barrels were lined up at the same height (changes aiming angle thus armor values)

No mention of the ammo types used in tests (as they all have differing characteristics)



the_Deadly_Bulb #20 Posted Mar 15 2018 - 18:32

    Major

  • Players
  • 23090 battles
  • 4,701
  • [WCTNT] WCTNT
  • Member since:
    03-11-2014

View PostLesser_Spotted_Panzer, on Mar 15 2018 - 09:20, said:

 

 

 

I put a link in the first comment of this thread. If you follow that link you will see the video to which is being referred to.


 

All of your items below were essentially addressed in the video:


 

 

 

No mention of the differences in crew training or equipment (changes aim time/dispersion)

No mention of the angle you are shooting at the target from (changes armor values, and needs to be a degree based value for accuracy in all 3 axis)

No mention of whether barrels were lined up at the same height (changes aiming angle thus armor values)

No mention of the ammo types used in tests (as they all have differing characteristics)

 

The statistically insignificant sample size is still the elephant in the room.

Viewing the video is a waste of time given the premise its built upon.







Also tagged with Penetration

1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users