Jump to content


VE, WN4, PR, and SE rating metrics compared with results


This topic has been archived. This means that you cannot reply to this topic.
256 replies to this topic

VaporGator #1 Posted Jan 09 2013 - 06:59

    First lieutenant

  • Players
  • 9520 battles
  • 809
  • [5PANZ] 5PANZ
  • Member since:
    09-19-2011
Everyone likes to see where they rank among all the other players.  There are numerous ratings to see how you stack up.  This thread is for discussing all rating metrics in one place.  Please be constructive, saying one sucks is not helpful and pointless (each has a thread you can do that in if desired)

As of now, WN and PR are the only ratings found on websites and signatures.  VE and SE are proposed metrics.  Most use a four digit rating system so it is familiar.

These ratings ARE DIFFERENT based on type and should not always be compared to each other due to context, even though results may look similar.

Overall Rating metric:  A rating system taking into consideration numerous stats and weighing them to derive an overall rating you can compare yourself to others on the server.
- WN
- PR

Focused Rating metric: A rating system that compares only one major stat to determine how players compare and makes no attempt to weight the stat, belief that raw data is most useful.
- VE
- SE


Readers, feel free to comment on problems/concerns and/or justification for each if you feel like it.  I will cut/paste as needed.

Owners of the ratings, feel free to correct anything I post below and I will update the comparisons in this post or add info as needed.  Also give a little blurb if you want to explain what you are shooting for with your rating to put it in context.

Quick Overview of the goal of each rating:

VE (aka VE+1)- The goal of VE is to compare people of similar average tiers and their effectiveness in those tiers.  The premise of VE is that if you live longer in battle, and do more damage, your win rate will go up.  How effective you are at following that simple mindset is the VE+1 score.  VE does not attempt to take into consideration any thing other than Damage and Tier played.  It is a focused metric and is not to be used to compare a new player in Tier 1 with a top player in tier 8 for instance.  The rating you receive is only compared to players within the same average tier.  The VE rating then is used in conjunction with the player's average tier rating to determine how effective they are in dealing damage in their average tier (or average class, the chart method is tbd)  See example of the Class ranking chart below the VE+1 section.

WN# - Currently in it's fourth iteration, WN has been viewed as a better version of the Efficiency rating.  WN# takes into account many variables and attempts to rate players on an overall scale.  The rating you have is comparable to every other player on the server.

PR - From the Noobmeter site, this rating is "a calculated value assessing the player skill, based on an assessment of the tanks the player is playing, the damage dealt and the win rate." As with WN#, the rating received is comparable to every player on the server.

SE - Undergoing some changes at moment, will update this later.


For disclosure, below is a chart showing each Author, there ratings among the different metrics. If there is any benefit from their own metric, it would be seen below relative to each other metric.

Posted Image

As of right now, these are my opinions based on collected comments or thoughts I have seen.  Feel free to add to them for a fair discussion.

FYI, all stats and figures listed below have been taken from the wotlabs.com signature generator on 01/09/13  (things can change!)

VE rating: (this is original, only left here to show why VE+1 came about)
Link: Not currently on a website, use simple formula listed below chart to calculate.   Avg Tier can be seen if you go to wotlabs.net and enter your name.

Purpose:
- To follow the mantra of "live longer, do more damage".  VE follows the idea that the average damage you do based on your average tier played shows your overall battle effectiveness.  Effectiveness in battle is affect by your skill of course, but also the tanks you choose to play.  Play above your skill level and your VE rating goes down.  Rush into battle and die quick without contributing to the team's effort and your rating will go down.  Live long, do lots of damage and your VE rating goes up.  This allows your VE rating to go up in those battles where you do a huge amount of damage but the rest of your team failed to do much of anything, so you are not penalized for a lost battle, or rewarded for a win that you did not contribute to.

Problems or Concerns:  
- Seal clubbers staying in tier 1 average.  Since the formula uses average tier as the divisor, a tier 1 only player will have a high rating if they are efficient seal clubbers.  
- Arty does more damage than it's tier
- Damage does not scale with tiers
- Basic formula

Response to concerns:
- Arty is being nerfed in 8.3 supposedly, so this may not even be a factor.  I don't have arty only players in the list, so it is hard to tell how any of the ratings handle them.  Also, not all arty players are arty only, nor are they all top players who devestate everyone with huge amounts of damage and many kills per game.  Arty also is susceptible to being killed early by scouts or counter arty.

- Damage scaling with tiers I think averages out based on your skill.  Just because you play some tier 9 or 10 tanks doesn't mean you will get a ton of damage.  You are likely to be killed quick by better players and as such you may hurt your VE score by raising your average tier divisor.

- Being a basic formula doesn't mean it is wrong.  Look at the results and compare them with other metrics.  2+2=4 is no less correct than ((4*9) - (18+10)) / 2 = 4 after all.  VE uses raw data with no modifiers.  Or for a more scientific view, consider the Empircal definition of Occam's Razor:  "In the related concept of overfitting, excessively complex models are affected by statistical noise (a problem also known as the bias-variance trade-off), whereas simpler models may capture the underlying structure better and may thus have better predictive performance."

(Charts may show some changes over days as people's ratings go up and down, every couple days I will verify each person's settings)

The chart showing how the players are rated:

Posted Image


Type of Metric:
Focused

Main Factors:
Average damage and Average Tier

Formula:
(Avg Damage) / (Avg Tier)*10  (please note, the *10 is used to adjust VE value to match the 4 digit values of WN and PR for easy comparisons)


VE+1 rating
Link: Not currently on a website, use simple formula listed below chart to calculate.   Avg Tier can be seen if you go to wotlabs.net and enter your name.

Purpose:
VE+1 is the same rating as VE, except it adds +1 to each players average tier.  What is the point of this?  To balance out the seal clubbers who average tier is 1.0   The rankings actually show the seal clubbers drop in the rating while the good players pretty much remain the same overall.  It was suggested for VE to add +1 to arty to balance their damage out, but with the pending arty nerf, it seems this +1 applied to everyone works pretty well and lowers the seal clubbers rating as a result.

Problems or Concerns:
- None at the moment, but I just thought of this and ran the numbers.  Let me know...

The chart showing how the players are rated:

Posted Image

Type of Metric:
Focused

Main Factors:
Average damage and Average Tier

Formula:
(Avg Damage) / (Avg Tier  + 1)*10 (please note, the *10 is added to put calculation into the more familiar 4 digit rating levels for easier comparison)

From the VE+1 rating, the user then would be compared to a chart of either individual tiers, or by class (tbd).  Take the VE rating and compare it with the Average Tier of the player (the original VE chart, or subtract one from the VE+1 listing above)

The first draft of the Class chart is listed below with all current players in this post listed where they would fall.  The first draft of the individual Tier charts are on page 9 of this thread.

http://i1278.photobu...zps53fdeca0.png


Chart for comparison with all other ratings, sorted by VE+1 ratings:

Posted Image


WN4 rating
WN4 listing can be seen on wotlabs.net or noobmeter.com

Purpose:
- Author states "I guess WN goal is to achieve a formula which measures skill as accurately as possible, with a simple formula (well relatively simple, lol), and which does it independently of tier played. We are constantly changing and improving it in the WN thread. Feedback appreciated. I consider this a group effort and not my own invention."

Problems or Concerns:
- Kills are factored in along with Win Rate and various factors and multipliers.  Is a 100% Kill treated the same as someone stealing the last 1 HP and getting the kill?
- Win Rate is a factor.  Win Rate is affected by the team, not just the individual.
- Seal clubbers generally rank higher than the accepted "best" players.
- Requires 2,000+ battles to be accurate (per author)

Response to Concerns:
- Calculations are being modified to address seal clubbers, at which point this will become WN5.

The chart showing how the players are rated:

Posted Image

Type of Metric:
Overall Rating

Main Factors:
Everything?  See below

Formula:
465*Kills + Damage*460/(184*EXP(0.24*AvgTier)) + 125*Spots + 100*MIN(Def , 2.2) + 0.5*((185/(0.17+EXP((WR-35)*-0.134)))-500)

Chart for comparison with all other ratings, sorted by WN4 ratings:

Posted Image

WN5 rating
WN5 is not currently on a website, new proposal to update WN4

Purpose:
- Author states "I guess WN goal is to achieve a formula which measures skill as accurately as possible, with a simple formula (well relatively simple, lol), and which does it independently of tier played. We are constantly changing and improving it in the WN thread. Feedback appreciated. I consider this a group effort and not my own invention."

Problems or Concerns:
- Kills are factored in along with Win Rate and various factors and multipliers.  Is a 100% Kill treated the same as someone stealing the last 1 HP and getting the kill?
- Win Rate is a factor.  Win Rate is affected by the team, not just the individual.
- Requires 2,000+ battles to be accurate (per author)


Response to Concerns:
- Calculations are being modified to address seal clubbers, at which point this will become WN5.

The chart showing how the players are rated:

Posted Image

Type of Metric:
Overall Rating

Main Factors:
Everything?  See below

Formula:
KILLS *(1240-1040/(TIER^0.164)-(TIER^4.7)*0.001)+DAMAGE*500/(184*e^(0.24*TIER)+130)+SPOT*125+MIN(DEF,2.2)*100+((185/(0.17+e^((WINRATE-35)*-0.134)))-500)*0.45

Chart for comparison with all other ratings, sorted by WN5 ratings:

Posted Image


PR rating
PR rating can be viewed onNoobmeter.com  

Purpose:
- I don't want to speak for the authors of this, they can let me know their goals if they want.

Problems or Concerns:
- Formula is unknown and seems to be the most common complaint
- Average tier is modified

Author Response to concerns:
- Author states the above concern is only due to protecting the formula and rating system so it is not stolen by foreign websites and used as their own.

The chart showing how the players are rated:

Posted Image

Type of Metric:
Overall Rating

Main Factors:
Unknown

Formula:
Unknown

Chart for comparison with all other ratings, sorted by PR ratings:

Posted Image
SE rating:
SE rating can be viewed at theSE Calculatorsite  (* in Beta and formula being reworked, so may or may not be valid when you check)

Purpose:
- Author states in his thread "Many players agree that Win Rate + Average Tier is an accurate measurement of skill. At the end, everything a player does is reflected in his win rate in one way or another."

Problems or Concerns:
- Win Rate is affected by the team, not just the individual

Author Response to concerns:
- It is being completely reworked for a third version.
- The Win Rate is affected by the team. That is true in smaller samples, however as the number of samples (battles) grow, the average tends to shift towards the player's own performance (Law of Large Numbers)

The chart showing how the players are rated: (not currently being updated while formula changes)

Posted Image

* - These three players have no recent games within 90 days and do not show any results per Neverwish's calculator, and I am too lazy to do the calculation at the moment -Vapor

Type of Metric
Focused

Main Factors:
Win Rate and Average Tier

Formula:
((((WR-48)*3) * AT)+(AT*WR))*2


As noted at the top of the thread, please add constructive comments for concerns and such and I will copy/paste them (as long as they are not just rants or personal attacks on any rating)

Edited by VaporGator, Jan 14 2013 - 18:39.


Quesnel #2 Posted Jan 09 2013 - 07:14

    Captain

  • Players
  • 23275 battles
  • 1,152
  • Member since:
    10-06-2011
Looks interesting..........

Neverwish #3 Posted Jan 09 2013 - 07:18

    Captain

  • Players
  • 6920 battles
  • 1,947
  • [LABS] LABS
  • Member since:
    06-10-2012
To address a few things about the SE rating:

- It is being completely reworked for a third version.
- The Win Rate is affected by the team. That is true in smaller samples, however as the number of samples (battles) grow, the average tends to shift towards the player's own performance (Law of Large Numbers)
- Neverwish, not NeverWinter :P

VaporGator #4 Posted Jan 09 2013 - 07:26

    First lieutenant

  • Players
  • 9520 battles
  • 809
  • [5PANZ] 5PANZ
  • Member since:
    09-19-2011
Oh, sorry about the name... just read a review of the NeverWinter pc game upcoming... guess it got stuck up in my head!  :)   I'll paste your stuff up in your section.

Edit: Fixed the picture with your proper name... oops.

Edited by VaporGator, Jan 09 2013 - 07:38.


hiipanda #5 Posted Jan 09 2013 - 07:37

    Staff sergeant

  • Players
  • 31302 battles
  • 498
  • [PARAX] PARAX
  • Member since:
    04-11-2011
i haven't found/used the VE rating before, but have used the other ones. can someone post a link to that one? I tried to google it but wasn't finding anything

Cuth #6 Posted Jan 09 2013 - 07:38

    Captain

  • Players
  • 12620 battles
  • 1,142
  • [VPG-R] VPG-R
  • Member since:
    11-03-2011

View PostVaporGator, on Jan 09 2013 - 06:59, said:

Problems or Concerns:
- Win Rate is affected by the team, not just the individual

Sorry, but denying that stats matter just invalidates your whole post. I'd suggest removing that part.

KilgorSoS #7 Posted Jan 09 2013 - 07:41

    Major

  • Players
  • 26602 battles
  • 11,792
  • Member since:
    07-23-2011
The simplest formulas seem to be the most accurate formulas...lolz


Occam's Razor at its finest...

Good Job, keep it up.

VaporGator #8 Posted Jan 09 2013 - 07:45

    First lieutenant

  • Players
  • 9520 battles
  • 809
  • [5PANZ] 5PANZ
  • Member since:
    09-19-2011

View PostCuth, on Jan 09 2013 - 07:38, said:

Sorry, but denying that stats matter just invalidates your whole post. I'd suggest removing that part.

It is a valid concern.  Doesn't mean it is right, but a concern nontheless if you are not a WR follower.  However, NeverWish already made that adjustment to his section.  

Keeping all concerns for ANY rating does not invalidate anything, it is there to be discussed.

VaporGator #9 Posted Jan 09 2013 - 07:53

    First lieutenant

  • Players
  • 9520 battles
  • 809
  • [5PANZ] 5PANZ
  • Member since:
    09-19-2011

View Posthiipanda, on Jan 09 2013 - 07:37, said:

i haven't found/used the VE rating before, but have used the other ones. can someone post a link to that one? I tried to google it but wasn't finding anything

If you mean website, it is not on any website yet  (you listening Neverwish and Noobmeter???!!!   :Smile_teethhappy: )  .  It is a rating I proposed as a raw number cruncher with no modifiers.  The formula is above, or I can add you to all the ratings lists if you want.

scharnhorst310 #10 Posted Jan 09 2013 - 08:05

    Major

  • Council of Armored Forces
  • 20905 battles
  • 3,631
  • [OTTER] OTTER
  • Member since:
    05-04-2011
I like VE+1, and win4 seems interesting as well. Could you add me to these, as i'm a good player but I'd like to see where I rank among the best.

VaporGator #11 Posted Jan 09 2013 - 08:24

    First lieutenant

  • Players
  • 9520 battles
  • 809
  • [5PANZ] 5PANZ
  • Member since:
    09-19-2011

View Postscharnhorst310, on Jan 09 2013 - 08:05, said:

I like VE+1, and win4 seems interesting as well. Could you add me to these, as i'm a good player but I'd like to see where I rank among the best.

You have been added to VE+1 and WN4 ratings lists.

RNGesus__ #12 Posted Jan 09 2013 - 08:41

    Corporal

  • Beta Testers
  • 14749 battles
  • 74
  • [RDDT] RDDT
  • Member since:
    01-22-2011
can u use me on the lists, like to see how I rank in there.

hiipanda #13 Posted Jan 09 2013 - 09:21

    Staff sergeant

  • Players
  • 31302 battles
  • 498
  • [PARAX] PARAX
  • Member since:
    04-11-2011

View PostVaporGator, on Jan 09 2013 - 07:53, said:

If you mean website, it is not on any website yet  (you listening Neverwish and Noobmeter???!!!   :Smile_teethhappy: )  .  It is a rating I proposed as a raw number cruncher with no modifiers.  The formula is above, or I can add you to all the ratings lists if you want.

i crunched the numbers after i got average tier from noobmeter since SE is still updating and said its not accurate. You can also throw me on there as well. it will be interesting since i share average tier with someone on the list.
also you messed up Camador's win% in a few graphs

scharnhorst310 #14 Posted Jan 09 2013 - 09:34

    Major

  • Council of Armored Forces
  • 20905 battles
  • 3,631
  • [OTTER] OTTER
  • Member since:
    05-04-2011
interesting that i rate so lowly on wn4... i bet its spotting...

Yibo #15 Posted Jan 09 2013 - 09:37

    First lieutenant

  • Players
  • 13739 battles
  • 662
  • [YO_OY] YO_OY
  • Member since:
    02-15-2011
PR is stupid, period.

DracoArgentum #16 Posted Jan 09 2013 - 11:12

    Major

  • Players
  • 12357 battles
  • 2,774
  • [IOC] IOC
  • Member since:
    09-25-2011
Noobmeter is clearly the best of those. It punishes sealclubbers most efectively and unlike SE its formula isn't stupid. SE it insane since Garbad's choice to freexp past tiers 1-4 actually boosts his rating. The same applies to VE btw.

Beerstein #17 Posted Jan 09 2013 - 11:14

    Major

  • Players
  • 15925 battles
  • 10,127
  • [PBKAC] PBKAC
  • Member since:
    08-18-2011

View PostDracoArgentum, on Jan 09 2013 - 11:12, said:

Noobmeter is clearly the best of those. It punishes sealclubbers most efectively and unlike SE its formula isn't stupid. SE it insane since Garbad's choice to freexp past tiers 1-4 actually boosts his rating. The same applies to VE btw.
Disagree. You have no evidence or numbers to show that either especially since you know, a formula was never released. Noobmeter is and always will be a useless trash metric I will never use unless it's made transparent.

Boom_Box #18 Posted Jan 09 2013 - 11:37

    First lieutenant

  • Beta Testers
  • 11737 battles
  • 892
  • Member since:
    11-17-2010
This is a little more accurate.

Posted Image

Reiuji #19 Posted Jan 09 2013 - 11:55

    Major

  • Players
  • 25732 battles
  • 2,246
  • Member since:
    04-04-2011
Interesting. Hmm, gonna follow this. +1

DracoArgentum #20 Posted Jan 09 2013 - 12:50

    Major

  • Players
  • 12357 battles
  • 2,774
  • [IOC] IOC
  • Member since:
    09-25-2011

View PostBeerstein, on Jan 09 2013 - 11:14, said:

Disagree. You have no evidence or numbers to show that either especially since you know, a formula was never released. Noobmeter is and always will be a useless trash metric I will never use unless it's made transparent.

The numbers are right there in what VG is trying to use to prove that his formula is any good. PR did the best out of those presented, closed formula or not. So unless he has stupid in there like using average tier and just got lucky its a decent measure. He did confirm that my guess is about right: avg damage in tank compared to server agerage and win rate in tank compared to server average in tank get munged together. So as long as those are the only inputs its at least not using stupid inputs.