Jump to content


Why WTR is a terrible metric


  • This topic is locked This topic is locked
110 replies to this topic

Ezz #101 Posted 08 December 2019 - 11:56 AM

    How many flipping posts do I need to get past Major?

  • Beta-Tester
  • 69877 battles
  • 36,309
  • [PBKAC] PBKAC
  • Member since:
    07-17-2012

Maybe to help vindi get a better understanding of the various metrics and their pros and cons...

 

WR - summarises ones ability to do well in the game. The game goal being to win after all. Lacks solo vs platoon (tho this is an issue for all these metrics). Easily available both in game and via 3rd party.

 

Derivative metrics.

WN8 - focuses on damage done in a vehicle relative to expected values for said vehicle. 3rd party metric. Available in both overall and 'recent' 1k battle sample sizes. Can be padded thru negative play.

Efficiency - barely worth exploring but generally speaking similar in concept to wn8 yet refined in different ways. 3rd party.

WN7 - precursor and easily padded. 3rd party.

PR - uses combo of in game stats. Includes stuff like assisted as well as WR. Very little sample size selectivity.

WTR - uses combo of in game stats including assisted and relative to expected for a vehicle / class. Only available in overall and monthly sizes. Unclear on validity of WG's methodology for expected values. Has glaring issues in small samples. Allows filtering of tier 8+. Appears to have a strong tier component.

 

Clearly among the derivative metrics there are advantages and disadvantages and all could be improved. Eg wn8 could get a better api that incorporates assisted and could allowing filtering out of the low tiers. While wtr could have a more useful recent component. Either way of course, neither will tell the story of who is doing well at the game better than win rate. After all, the goal of the game is to win, not to pad wn8 or wtr.

 

Further, just as any discussion on ability, the key would be to use a combination - just like a combo of WR and wn8, a combo of WR and wtr could be used. If wg realeased a meaningful 'recent' wtr (call it wtr(improved)) that could be filtered for tier 8+ WR + wtr(imp) would probably be a better combo than WR+wn8.

 


Edited by Ezz, 08 December 2019 - 12:05 PM.

Who the [edited] are you? Get Spoofed! "wouldn't be a proper WG balance change if they didn't [edited] something up after all "

>9000 cynicism brought to you by P2W, RIP Balance and the Cartoon Connection

Currently moderating your English speaking community : AALGMadibaCenturion_NZ, Elite911, Moonbase Patrol Copter 7

R. Pubbie: "why are all PBKAC players so rude, arrogant and nasty? and why do Mods favor them?"


Ezz #102 Posted 08 December 2019 - 12:54 PM

    How many flipping posts do I need to get past Major?

  • Beta-Tester
  • 69877 battles
  • 36,309
  • [PBKAC] PBKAC
  • Member since:
    07-17-2012

View PostAoyama_Blue_Mountain, on 08 December 2019 - 01:19 PM, said:

Just for some useful discussion, do you guys know about unproductive / negative-productive / vampire / unoptimal damage?

 

....

Nice write up. The key here for players to understand what sort of damage / assist they are doing is looking at their WR. If it's low, then chances are their influence on battles is in one of the meaningless categories. If their WR is good, then they are clearly having a positive influence on battles.


Who the [edited] are you? Get Spoofed! "wouldn't be a proper WG balance change if they didn't [edited] something up after all "

>9000 cynicism brought to you by P2W, RIP Balance and the Cartoon Connection

Currently moderating your English speaking community : AALGMadibaCenturion_NZ, Elite911, Moonbase Patrol Copter 7

R. Pubbie: "why are all PBKAC players so rude, arrogant and nasty? and why do Mods favor them?"


_stevethegecko_ #103 Posted 08 December 2019 - 01:24 PM

    Major

  • Beta-Tester
  • 29350 battles
  • 13,193
  • [PBKAC] PBKAC
  • Member since:
    03-13-2012

View PostAoyama_Blue_Mountain, on 08 December 2019 - 02:19 PM, said:

 

Spoiler

 

+1


kilig ako


NameWasStolenStresslevel #104 Posted 08 December 2019 - 06:57 PM

    Captain

  • Member
  • 44462 battles
  • 1,201
  • Member since:
    01-19-2015

View PostAoyama_Blue_Mountain, on 08 December 2019 - 12:19 PM, said:

 

Just for some useful discussion, do you guys know about unproductive / negative-productive / vampire / unoptimal damage?

 

Damage (dealt by u) that does not result in difference in outcome is unproductive

Negative-productive is causing damage to your team in the process

And vampire is stealing from others (and will fall under one of the above two)

Unoptimal damage is less-useful damage that even if you trade 1:1 (or higher in worse cases), you can still make the team worse off

 

Unproductive: At the end of the game and the battle is already own and the enemy has only 3 arties left. Any damage done at this point has no effect on outcome except personal reward. And then you see that Progetto at the base beeline towards enemy cap.

Or just killstealing a tank which obviously is going to get killed without repercussion anyway (i.e. having to reload is not a problem) because 4 shots are already in-flight towards that tank. Arties take note.

 

Negative-productive: Keep moving in-and-out of the corner to bait an enemy shot when there are teammates sharing that corner so the moment you retreat the enemy just shoots your teammate instead. And then you move in to take the shot while blocking your teammate from doing so. So the overall trade is 1:1 for neutral productivity if you didn't screw up and miss, and most likely you will so the productivity is negative. Your teammate might have handled the encounter better if you didn't exist, at least a 1:1 trade or a better trade if he is good.

Or having two tanks sharing a bush which can only accommodate one. Even if the good player decides to let the bad player have the bush, there is still opportunity cost since you are not making the best use of the position. With two bad players, one or both players will die for nothing.

 

Vampire: Killstealing mentioned above is one. Another scenario is having two LT scout the same thing when one is enough, so the other is just stealing his assist damage after already knowing where the enemy is. (Does not apply if the LT are supporting each other e.g. spreading out attention, killing stuff)

Suiciding your full HP tank to killsteal - The act of using one whole tank for one shot itself falls under simply being bad, but the killstealing itself is unproductive if the guy who, like, did 80% of the damage would have killed easily anyway.

 

Unoptimal damage: Using high tier to trade with low tier is one. Using scout to trade against HT. Or a good player losing HP against bad players while the good player on the enemy team is full HP. Yes aiming good players is a strategy. Arties are a valid strategy.

 

Even aggressive play can boost your damage using the above points: If the team could have won with 1/1/1/1 damage shared between 4 tanks, but since you played aggressively the outcome is 2.5/0.5/0.5/0.5 damage, and may even have downsides since you lost HP - see unoptimal damage. Mopping up is when everybody is playing the most aggressively.

 

Then there is your plain usual doing damage unproductively at the redline.

 

--------------------

 

That's why above a certain threshold (enough to tell us that the player does not suck), damage per battle tells us more about playstyle. Whether that playstyle results in win, well, there's winrate. The guy who gets that Defender (or Invader) medal has lower damage than the guy who just continued farming damage, but guess who has the higher winrate.

 

Add: Want to add on regarding sniping - There is useful sniping which a lot of good players are doing even if done at the redline. Because sniping can 1) Allow more guns to hit a single target as opposed to just a 1:1 fight, 2) Have angles of attack otherwise unachievable due to terrain. In a sense they are similar to arty. And arty is good in Malinvoka.

And this is not appreciated by the game reward system.

Yea, still kind of pissed when my EBR managed to run to the south base in Serene Coast and buttsecks the entire enemy hill yet got no Chev despite doing lots of damage.

 

 

we should have a thread to discuss specifically things like this.


Having META problems?

 #Hashtag break the META

     (HAHAGHa i must have drank somethin)

             https://worldoftanks.asia/en/hall-of-fame/vehicles/A100_T49/#wot&w_ts=overall&w_nb=500

         Currently Hall of Fame T49's number 1 in WTR!!! (parameters are *battles +500* and *since 2014's* ranking)


DeL_UwU #105 Posted 08 December 2019 - 11:36 PM

    Private

  • Member
  • 1164 battles
  • 5
  • [CHLDA] CHLDA
  • Member since:
    09-19-2019

although tbh, I value Damage above everything else. eg. I am more then willing to throw a game to get more damage.

yet i still have a 65% wr. You guys talk as if a dpg [edited] cant win games. when thats not true.

The people you are talking about are redline snipers who dont actually have good dpgs/also dont have WR



Vindictus_Maximus #106 Posted 09 December 2019 - 11:52 AM

    Sergeant

  • Member
  • 6906 battles
  • 122
  • Member since:
    12-16-2017

View PostSaffelicious, on 08 December 2019 - 10:35 AM, said:

This game is about winning,

 

 want to know if a player is good?

 

 

Yes but this thread is about the WTR, which will soon replace the current PR rating system. Read for your self - winning is not factored in.

 

https://worldoftanks...ersonal-rating/

 

Some in this thread are going through the various stages of grief coming to terms with it all, acceptance will come in the end. And hopefully change the way they play, the way they play thier tanks, and which tanks they choose to play.

 

The new metric calculates how well you play a given tank, and how well you play that tank compared to others. Basically in this new era you don't choose the strongest OP tank,. instead you choose a tank that you will able to play correctly, and one you are most likely to 3-mark.(just remember - marks of excellence doesn't factor wins either) Also you are awarded multipliers the higher the tier you play, killing off low-tier stat padding.

 

So whether Slav King gets 53% winrate or 69% winrate, he still finished in the top5 because he plays those tanks better than everyone else.

 

And Saffe, the best way to tell how good a player is is to see how high in the WTR  rankings they get. Choose a few tanks, play for them for the month, lets see if you can get into the top 10.

 

 

 


Edited by Vindictus_Maximus, 09 December 2019 - 11:53 AM.


Vindictus_Maximus #107 Posted 09 December 2019 - 12:04 PM

    Sergeant

  • Member
  • 6906 battles
  • 122
  • Member since:
    12-16-2017

View PostEzz, on 08 December 2019 - 11:56 AM, said:

 WR - summarises ones ability to do well in the game. The game goal being to win after all. Lacks solo vs platoon (tho this is an issue for all these metrics). Easily available both in game and via 3rd party.

 

I showed you the Slav KIng example to prove how unreliable winrate is compared to WTR.

 

There is a definate correlation between OP premium tanks and winrate for example. Under this new system, maybe it will be better to play some of these less-played tanks, because you are being compared against the server average. Maybe the player that slightly improves his his W/R playing OP tanks will suddenly find his stats in decline because he plays that tank far worse than the server average.

 

There is a gold mine of rarely played tanks out  there that are easy to get marks on the barrel. Some are so easy that even average players can get a mark on the barrel from 40-odd games or something. How easily you can mark a barrel is a good example of which tanks will be good to play if you want to improve your WTR.

 

Meta tanks are no longer meta if you care about your WTR stats, it will be good to see a wider range of tanks being player tbh


Edited by Vindictus_Maximus, 09 December 2019 - 12:04 PM.


YouLoveUs #108 Posted 09 December 2019 - 12:12 PM

    Private

  • Member
  • 230 battles
  • 5
  • [REMY] REMY
  • Member since:
    10-29-2014

So you showed an account played by at least two different players as an example of what exactly?

 

Try again.

 

Clearly you have figured gun marks though with your sub par overall win rate even with a padded win rate in a Luchs. Keep looking though, one day you will find a metric to validate your sub par reroll ability, even if you have to invent VM9



_stevethegecko_ #109 Posted 09 December 2019 - 12:13 PM

    Major

  • Beta-Tester
  • 29350 battles
  • 13,193
  • [PBKAC] PBKAC
  • Member since:
    03-13-2012

View PostVindictus_Maximus, on 09 December 2019 - 02:52 PM, said:

 

Yes but this thread is about the WTR, which will soon replace the current PR rating system. Read for your self - winning is not factored in.

 

https://worldoftanks...ersonal-rating/

 

Some in this thread are going through the various stages of grief coming to terms with it all, acceptance will come in the end. And hopefully change the way they play, the way they play thier tanks, and which tanks they choose to play.

 

The new metric calculates how well you play a given tank, and how well you play that tank compared to others. Basically in this new era you don't choose the strongest OP tank,. instead you choose a tank that you will able to play correctly, and one you are most likely to 3-mark.(just remember - marks of excellence doesn't factor wins either) Also you are awarded multipliers the higher the tier you play, killing off low-tier stat padding.

 

So whether Slav King gets 53% winrate or 69% winrate, he still finished in the top5 because he plays those tanks better than everyone else.

 

And Saffe, the best way to tell how good a player is is to see how high in the WTR  rankings they get. Choose a few tanks, play for them for the month, lets see if you can get into the top 10.

 

 

 

No Vindi. The thread is about WTR being useless because it doesn’t take win% into account. Winning more makes players good. Not derived metrics. 
 

I showed you King’s stats over daily, weekly, monthly, 60 day and 1000 games and it showed consistency. You sir need to read full posts before commenting. 


Edited by FeeTFooD, 09 December 2019 - 12:19 PM.

kilig ako


Ezz #110 Posted 09 December 2019 - 12:50 PM

    How many flipping posts do I need to get past Major?

  • Beta-Tester
  • 69877 battles
  • 36,309
  • [PBKAC] PBKAC
  • Member since:
    07-17-2012

View PostVindictus_Maximus, on 09 December 2019 - 02:04 PM, said:

 .

 

Meta tanks are no longer meta if you care about your WTR stats

Suspect you hit the nail on the head there.


Edited by Ezz, 09 December 2019 - 12:50 PM.

Who the [edited] are you? Get Spoofed! "wouldn't be a proper WG balance change if they didn't [edited] something up after all "

>9000 cynicism brought to you by P2W, RIP Balance and the Cartoon Connection

Currently moderating your English speaking community : AALGMadibaCenturion_NZ, Elite911, Moonbase Patrol Copter 7

R. Pubbie: "why are all PBKAC players so rude, arrogant and nasty? and why do Mods favor them?"


Flying_Elite #111 Posted 09 December 2019 - 01:40 PM

    Always blame Hema!

  • Moderator
  • 30985 battles
  • 1,910
  • [FORT] FORT
  • Member since:
    06-13-2013
Thread closed at OP's request
~Flying_Elite

   

     

Spoiler

 





1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users