Jump to content


win8 correlates inversely with winrate?

win8 winrate

  • Please log in to reply
17 replies to this topic

Geoffrey_Ironfist #1 Posted 02 October 2019 - 04:39 PM

    Staff Sergeant

  • Player
  • 27728 battles
  • 398
  • [YBB] YBB
  • Member since:
    06-23-2018

Here are my five best tanks by win8, almost perfectly correlating inversely with winrate. Has anyone else noticed this with their tanks?

 

 

Attached Files

  • Attached File   inverse_correlation.jpg   91.28K

Edited by Geoffrey_Ironfist, 02 October 2019 - 04:48 PM.


Miepie #2 Posted 02 October 2019 - 04:52 PM

    Second Lieutenant

  • Player
  • 3976 battles
  • 1,030
  • [WJDE] WJDE
  • Member since:
    05-19-2018
20 battles is not a sample size! 163 battles in the M3 Lee is bound to get you forcibly institutionalised in several countries! :izmena:

shikaka9 #3 Posted 02 October 2019 - 04:52 PM

    Warrant Officer

  • Player
  • 70538 battles
  • 930
  • [-SJA-] -SJA-
  • Member since:
    02-27-2013
not until today

Echotun #4 Posted 02 October 2019 - 05:00 PM

    Staff Sergeant

  • Player
  • 21943 battles
  • 496
  • [FA-1] FA-1
  • Member since:
    03-15-2015

It is extremely easy even for a mediocre player to boost wn8 at the expense of win rate. Which is one of several reason why statistics should be judged in a proper context. 
 

ofc, your sample size is so small that the numbers are irrelevant for a couple of the tanks. 


Edited by Echotun, 02 October 2019 - 05:01 PM.


Geoffrey_Ironfist #5 Posted 02 October 2019 - 05:07 PM

    Staff Sergeant

  • Player
  • 27728 battles
  • 398
  • [YBB] YBB
  • Member since:
    06-23-2018

View PostEchotun, on 02 October 2019 - 05:00 PM, said:

It is extremely easy even for a mediocre player to boost wn8 at the expense of win rate. Which is one of several reason why statistics should be judged in a proper context. 
 

ofc, your sample size is so small that the numbers are irrelevant for a couple of the tanks. 

 

Except that I wasn't trying to boost my win8, I was trying to win games.

 

Twenty battles is not much but the inverse correlation with win8 in that tank is so big, that's exactly what tells me that there's something fishy about the win8. But even if there was no correlation between winrate and win8, wouldn't that be a bit strange? I mean the win8 is supposed to be a measure of something that's good, rather than bad or irrelevant.


Edited by Geoffrey_Ironfist, 02 October 2019 - 05:28 PM.


LordMuffin #6 Posted 02 October 2019 - 05:49 PM

    Field Marshal

  • Player
  • 52608 battles
  • 13,297
  • [-GLO-] -GLO-
  • Member since:
    06-21-2011

View PostGeoffrey_Ironfist, on 02 October 2019 - 05:07 PM, said:

Twenty battles is not much but the inverse correlation with win8 in that tank is so big, that's exactly what tells me that there's something fishy about the win8. But even if there was no correlation between winrate and win8, wouldn't that be a bit strange? I mean the win8 is supposed to be a measure of something that's good, rather than bad or irrelevant.

No, this tells me you have pretty much no clue about how small of a sample 20 games is.

1 changed result would take you to 50% WR. 2 changed results to 55% which would be the expected value.

 

 

There is actually a very strong correlation between WR and wn8 as can be seen on pretty much every tank I have, and on my account as a whole, on your account and pretty much every other players account aswell. 


Edited by LordMuffin, 02 October 2019 - 05:50 PM.


Echotun #7 Posted 02 October 2019 - 05:50 PM

    Staff Sergeant

  • Player
  • 21943 battles
  • 496
  • [FA-1] FA-1
  • Member since:
    03-15-2015

View PostGeoffrey_Ironfist, on 02 October 2019 - 05:07 PM, said:

 

Except that I wasn't trying to boost my win8, I was trying to win games.

 

Twenty battles is not much but the inverse correlation with win8 in that tank is so big, that's exactly what tells me that there's something fishy about the win8. But even if there was no correlation between winrate and win8, wouldn't that be a bit strange? I mean the win8 is supposed to be a measure of something that's good, rather than bad or irrelevant.


wn8 is based on you damaging enemy tanks. Win rate is also heavily tied into that, but not nearly as much as wn8. Win rate also cares about if you do the damage before or after the battle is lost, to give one example. 



LordMuffin #8 Posted 02 October 2019 - 05:51 PM

    Field Marshal

  • Player
  • 52608 battles
  • 13,297
  • [-GLO-] -GLO-
  • Member since:
    06-21-2011

View PostEchotun, on 02 October 2019 - 05:50 PM, said:


wn8 is based on you damaging enemy tanks. Win rate is also heavily tied into that, but not nearly as much as wn8. Win rate also cares about if you do the damage before or after the battle is lost, to give one example. 

Winrate doesn't care about damage, it only cares about outcome when calculated.



15JG52Brauer #9 Posted 02 October 2019 - 06:33 PM

    Warrant Officer

  • Player
  • 81524 battles
  • 731
  • [WHO] WHO
  • Member since:
    06-29-2011

Do not cofusecausation with corelation ;-)  For me my highest WN8 tanks tend to have good win rates as well  - in these cases I think it was because I did damage early in the game, tilting the balance in favour of my team.

 

Screenshot_2019-10-02 15JG52Brauer - WoT-Life com - World of Tanks Statistics.png

 


Edited by 15JG52Brauer, 02 October 2019 - 06:37 PM.


Echotun #10 Posted 02 October 2019 - 07:50 PM

    Staff Sergeant

  • Player
  • 21943 battles
  • 496
  • [FA-1] FA-1
  • Member since:
    03-15-2015

View PostLordMuffin, on 02 October 2019 - 05:51 PM, said:

Winrate doesn't care about damage, it only cares about outcome when calculated.


The most common way to win is to do more damage to the other team than your own team recieves. 



Dorander #11 Posted 02 October 2019 - 09:08 PM

    Major General

  • Player
  • 21025 battles
  • 5,986
  • Member since:
    05-07-2012

View PostGeoffrey_Ironfist, on 02 October 2019 - 03:39 PM, said:

Here are my five best tanks by win8, almost perfectly correlating inversely with winrate. Has anyone else noticed this with their tanks?

 

 

 

Mostly with tanks for which win contribution isn't measured properly by WN8, which is primarily damage/kill based, such as superheavies that I use to block a lot of damage or lights that I do a lot of scouting with. WN8 seems most consistently applicable to heaviums and mediums.

 

That said, I'm not sure why you think those 5 tanks have an inverse correlation. You could say that your winrate goes up while your WN8 goes down but that sequence is broken halfway and 5 tanks is way too short a list to ignore an exception. Your ARL has only 20 battles so your sample size is too low to make a proper judgement about the correlation, disregarding the fact that it's a TD which is a class that tends to see larger discrepancies due to playstyle.

 

The M3 Lee, the JgPzIV and the DW2 all have close matching values. I do hope you're not looking at the colours, as looking at those values you're simply going past a colour shift norm. Dark green - light blue correlation is normal. The only truly deviating value in this list is your M10, you'd expect those values to be closer together, but why that is can't be said without looking at your playstyle.



LordMuffin #12 Posted 02 October 2019 - 10:26 PM

    Field Marshal

  • Player
  • 52608 battles
  • 13,297
  • [-GLO-] -GLO-
  • Member since:
    06-21-2011

View PostEchotun, on 02 October 2019 - 07:50 PM, said:


The most common way to win is to do more damage to the other team than your own team recieves. 

Yes, but that doesn't mean damage is included in WR formula. 

 

For example: it is in general good to shoot shells if you want to do damage.

Yet amount of shots fired is not included in wn8.



Baldrickk #13 Posted 03 October 2019 - 01:58 AM

    Field Marshal

  • Player
  • 32165 battles
  • 16,690
  • [-MM] -MM
  • Member since:
    03-03-2013
In lost games,  it's possible to do lots of damage because your team sucks, but not quite hard enough to hold the enemy up enough that you can farm some decent damage.
This can lead to a good WN8 with low WR, at least with small battle counts.
With more battles, the average case will dominate.

Well done on acing the ARL within 20 battles though.  It joined the now fairly long list of tanks I aced on the very last battle I played in it.

Schepel #14 Posted 03 October 2019 - 02:05 AM

    Colonel

  • Player
  • 65859 battles
  • 3,720
  • Member since:
    05-13-2013

View PostBaldrickk, on 03 October 2019 - 01:58 AM, said:

In lost games, it's possible to do lots of damage because your team sucks, but not quite hard enough to hold the enemy up enough that you can farm some decent damage.
This can lead to a good WN8 with low WR, at least with small battle counts.
With more battles, the average case will dominate.

Well done on acing the ARL within 20 battles though. It joined the now fairly long list of tanks I aced on the very last battle I played in it.

 

This. If you are the only (semi) competent player in a team, you are likely to get more damage done, but you are also likely to lose the game. If you are in a team with more (semi) competent players, they are likely to do some damage of their own, thus leaving you with less. You will, however, be more likely to win when part of such a team.

 

So, yes, in the short term, high WN8 games are also likely to be losses. Long term, doing well translates into winning games. Or it should, anyway. Not feeling much of that myself, as of late.



Jauhesammutin #15 Posted 03 October 2019 - 06:45 AM

    Warrant Officer

  • Player
  • 23608 battles
  • 963
  • [KANKI] KANKI
  • Member since:
    11-05-2013

View PostGeoffrey_Ironfist, on 02 October 2019 - 03:39 PM, said:

Here are my five best tanks by win8, almost perfectly correlating inversely with winrate. Has anyone else noticed this with their tanks?

 

 

 

I haven't noticed it.

 

 



TankkiPoju #16 Posted 03 October 2019 - 07:01 AM

    Lieutenant General

  • Player
  • 23588 battles
  • 7,233
  • Member since:
    05-20-2011

I don't see the problem here, OP clearly has higher both WN8 and winrate during last 1000 games or so.

 

Higher WN8 correlates with higher winrate, because games are won by shooting all enemies dead. Pretty simple stuff, really.



SuperOlsson #17 Posted 03 October 2019 - 12:15 PM

    Second Lieutenant

  • Player
  • 23919 battles
  • 1,202
  • [E-5OM] E-5OM
  • Member since:
    08-07-2012
Nope. What is win8 by the way?

TankkiPoju #18 Posted 03 October 2019 - 12:57 PM

    Lieutenant General

  • Player
  • 23588 battles
  • 7,233
  • Member since:
    05-20-2011

View PostSuperOlsson, on 03 October 2019 - 12:15 PM, said:

Nope. What is win8 by the way?

 

Well it's like win7, but this one goes to 8.

 

Like if you are at win7, but want to push it a bit more, yeah? Where are you going to go? That's right, win8.

 







Also tagged with win8, winrate

1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users