Jump to content


winrate is not a reflection of player skill


  • Please log in to reply
801 replies to this topic

Copacetic #221 Posted Apr 21 2019 - 22:25

    Captain

  • -Players-
  • 47250 battles
  • 1,666
  • [ZEUS] ZEUS
  • Member since:
    02-04-2014

View PostSquishySupreme, on Apr 21 2019 - 21:54, said:

 

Neither does statistics, apparently.  But I suspect few people here have actually taken Statistics at a college level.

No, really.  WR is missing two critical pieces of data that keep it from being usable for any sort of statistical analysis.

1: Group performance is in no way indicative of personal performance.  It is impossible to extrapolate data on what the individual players did or did not do from a single aggregate score.
eg: You take a group of dice.  2D4,(new players)  2D6,(better), 3D8 (normal), 2D10, and 1D12(elite player).  10 dice of varying values. Roll them together.   That single value cannot tell what individual die rolled what value.  Those two D4s could have rolled a 3 and a 4, that D10 could have rolled a 2 or a 10.     People also mistakenly think that the previous roll or even 100 rolls will tell you what will happen on any future single roll of the die.  

Which brings us to point #2
2:A single death per match/a single set of values added together also doesn't even generate a bell curve.  The issue is that #1 might work if it was the same people over multiple matches and enough data could be collected on that team.  But the MM makes that all random. When you take 15 random players of random skill levels and toss them in a room together, then have them all roll their dice ONCE, you in no way have a means to do any meaningful long-term analysis.  There simply is not enough data to make the math work.  Giving everyone a skill rating on what the group rolled as a total is invalid unless the exact same group of players is used several times and multiple sets of data are generated to run a comparison on. 

So the WR might be useful in clan wars or something with fixed teams, but for random battles it's useless.   

This is why all other games from the beginning of multiplayer gaming have focused on personal kills and accuracy.  That can be tracked and is a valid way to determine individual skill over time.

 

No actually a lot of us have. Hence why we call idiots out on this kind of crap.

Tolos #222 Posted Apr 21 2019 - 22:27

    Major

  • Beta Testers
  • 25997 battles
  • 18,534
  • [GDF] GDF
  • Member since:
    09-22-2010

View PostSquishySupreme, on Apr 21 2019 - 20:54, said:

 

Neither does statistics, apparently.  But I suspect few people here have actually taken Statistics at a college level.

No, really.  WR is missing two critical pieces of data that keep it from being usable for any sort of statistical analysis.

1: Group performance is in no way indicative of personal performance.  It is impossible to extrapolate data on what the individual players did or did not do from a single aggregate score.
eg: You take a group of dice.  2D4,(new players)  2D6,(better), 3D8 (normal), 2D10, and 1D12(elite player).  10 dice of varying values. Roll them together.   That single value cannot tell what individual die rolled what value.  Those two D4s could have rolled a 3 and a 4, that D10 could have rolled a 2 or a 10.     People also mistakenly think that the previous roll or even 100 rolls will tell you what will happen on any future single roll of the die.  

Which brings us to point #2
2:A single death per match/a single set of values added together also doesn't even generate a bell curve.  The issue is that #1 might work if it was the same people over multiple matches and enough data could be collected on that team.  But the MM makes that all random. When you take 15 random players of random skill levels and toss them in a room together, then have them all roll their dice ONCE, you in no way have a means to do any meaningful long-term analysis.  There simply is not enough data to make the math work.  Giving everyone a skill rating on what the group rolled as a total is invalid unless the exact same group of players is used several times and multiple sets of data are generated to run a comparison on. 

So the WR might be useful in clan wars or something with fixed teams, but for random battles it's useless.   

This is why all other games from the beginning of multiplayer gaming have focused on personal kills and accuracy.  That can be tracked and is a valid way to determine individual skill over time.

 

 

And poor players like yourself always have those ' vital ' statistics low, which interestingly enough would tie in with WR, funny how that works, isn't it. It's almost like, if you play better by  doing more damage, hitting more, tanking more, you win more....who knew !!!!



Copacetic #223 Posted Apr 21 2019 - 22:32

    Captain

  • -Players-
  • 47250 battles
  • 1,666
  • [ZEUS] ZEUS
  • Member since:
    02-04-2014

View PostTolos, on Apr 21 2019 - 22:27, said:

 

 

And poor players like yourself always have those ' vital ' statistics low, which interestingly enough would tie in with WR, funny how that works, isn't it. It's almost like, if you play better by  doing more damage, hitting more, tanking more, you win more....who knew !!!!

 

^this duh

NeatoMan #224 Posted Apr 21 2019 - 22:59

    Major

  • Players
  • 28180 battles
  • 20,552
  • Member since:
    06-28-2011

View PostSquishySupreme, on Apr 21 2019 - 15:54, said:

Neither does statistics, apparently.  But I suspect few people here have actually taken Statistics at a college level.


No, really.  WR is missing two critical pieces of data that keep it from being usable for any sort of statistical analysis.

1: Group performance is in no way indicative of personal performance.  It is impossible to extrapolate data on what the individual players did or did not do from a single aggregate score.
eg: You take a group of dice.  2D4,(new players)  2D6,(better), 3D8 (normal), 2D10, and 1D12(elite player).  10 dice of varying values. Roll them together.   That single value cannot tell what individual die rolled what value.  Those two D4s could have rolled a 3 and a 4, that D10 could have rolled a 2 or a 10.     People also mistakenly think that the previous roll or even 100 rolls will tell you what will happen on any future single roll of the die. 

 

So you are trying to tell us that you can't use statistics to figure out the probability of success of:

 

3d4 + 4d6 + 5d8 + 2d10 + 1d12 

vs  

5d4 + 5d6 + 4d8 + 1d10

 

you have got to be kidding.

 

That's all this is.  Do the same thing over thousands of games and you will generate a curve based on all the various combinations.  You are trying to use the gambler's fallacy to disprove long term statistics.   It doesn't work that way.  Is it exact?  of course not, it's probabilities, but that doesn't mean it doesn't apply



Rocketmandb #225 Posted Apr 21 2019 - 23:28

    Major

  • Players
  • 53741 battles
  • 2,140
  • [STRNG] STRNG
  • Member since:
    12-29-2011

View PostSquishySupreme, on Apr 21 2019 - 20:54, said:

 

: Group performance is in no way indicative of personal performance.  It is impossible to extrapolate data on what the individual players did or did not do from a single aggregate score.
 

 

For a single event, yes. For n events no. 



Viruzzz #226 Posted Apr 22 2019 - 06:15

    First lieutenant

  • Players
  • 16791 battles
  • 900
  • Member since:
    09-26-2012

View PostSquishySupreme, on Apr 21 2019 - 20:54, said:

 

Neither does statistics, apparently.  But I suspect few people here have actually taken Statistics at a college level.

No, really.  WR is missing two critical pieces of data that keep it from being usable for any sort of statistical analysis.

1: Group performance is in no way indicative of personal performance.  It is impossible to extrapolate data on what the individual players did or did not do from a single aggregate score.
eg: You take a group of dice.  2D4,(new players)  2D6,(better), 3D8 (normal), 2D10, and 1D12(elite player).  10 dice of varying values. Roll them together.   That single value cannot tell what individual die rolled what value.  Those two D4s could have rolled a 3 and a 4, that D10 could have rolled a 2 or a 10.     People also mistakenly think that the previous roll or even 100 rolls will tell you what will happen on any future single roll of the die.  

Which brings us to point #2
2:A single death per match/a single set of values added together also doesn't even generate a bell curve.  The issue is that #1 might work if it was the same people over multiple matches and enough data could be collected on that team.  But the MM makes that all random. When you take 15 random players of random skill levels and toss them in a room together, then have them all roll their dice ONCE, you in no way have a means to do any meaningful long-term analysis.  There simply is not enough data to make the math work.  Giving everyone a skill rating on what the group rolled as a total is invalid unless the exact same group of players is used several times and multiple sets of data are generated to run a comparison on. 

So the WR might be useful in clan wars or something with fixed teams, but for random battles it's useless.   

This is why all other games from the beginning of multiplayer gaming have focused on personal kills and accuracy.  That can be tracked and is a valid way to determine individual skill over time.

 

Can you take a single result from a group and extract one players performance? No, you cannot.

 

Can you take a thousand results from randomly selected groups where that player was present and extract his average impact (i.e. skill) on the groups? You absolutely can.



PuddleSplasher #227 Posted Apr 26 2019 - 12:00

    Major

  • Players
  • 39173 battles
  • 3,152
  • Member since:
    07-11-2011

View PostcKy_, on Apr 17 2019 - 14:31, said:

 

Buddy, it is better to be thought of as a fool than open your mouth and remove all doubt.

 

I so love a good troll so bring it on, SON!!!

_Tsavo_ #228 Posted Apr 26 2019 - 12:22

    Major

  • Beta Testers
  • 46204 battles
  • 19,693
  • [BRVE] BRVE
  • Member since:
    02-16-2011

View PostPuddleSplasher, on Apr 26 2019 - 06:00, said:

 

I so love a good troll so bring it on, SON!!!

 

You agreed with a troll post, dude.   And got called for it. 



cheapbooks #229 Posted May 02 2019 - 09:04

    Captain

  • Players
  • 38710 battles
  • 1,140
  • [CBKS] CBKS
  • Member since:
    08-21-2013

If you look at QuickyBaby's "free to play" account and his premium account, in almost every measure possible, he outperforms his 'free to play account' with his pay account.

 

His EU usernames are QuickFingers and PlaysForFree

 

only 1 or 2 tanks perform better on his free to play account, one being a premium tank which has the same modules  in both accounts.

 

I am thinking that player skill only affects winrate noticeably if you are one of the few remaining tanks in the game. the more allies and enemy tanks die, the more of an impact individual skill has on winrate.

 

I bet that if you take 1 skilled player vs 2 unskilled players, the unskilled players will win if they use teamwork.



NightmareMk9 #230 Posted May 02 2019 - 09:26

    Major

  • Players
  • 35258 battles
  • 2,655
  • [ZOUTH] ZOUTH
  • Member since:
    03-11-2012

I bet if you had $1 you might be able to buy a clue.

YES, spamming gold in your account with 5 skill crews and improved equipment will win more often.

But win rate is ABSOLUTELY a measurement of player skill.

If you take 2 people with 45% win rate (after say 20,000 games) and put them on voice chat, I bet a 60% player (with 20,000 games) kills both of them 8 times out of 10.



Nixeldon #231 Posted May 02 2019 - 11:52

    Major

  • Players
  • 60879 battles
  • 2,215
  • Member since:
    10-30-2011

View Postcheapbooks, on May 02 2019 - 03:04, said:

I am thinking that player skill only affects winrate noticeably if you are one of the few remaining tanks in the game. the more allies and enemy tanks die, the more of an impact individual skill has on winrate.

Your thinking is still illogical. Win rate and skill aren't about that one game but consistency over many games. 

 

Player skill is the ability of said player to affect their win rate and win rate measures this skill independently of what the player is doing. Win rate doesn't describe what the skill is only that it exists. Matches are most reliably won by eliminating enemy tanks and the skill is in doing so. The more enemy tanks are gone, the less of an opportunity for a positive impact; you can't damage enemies when there are none.

 



NeatoMan #232 Posted May 02 2019 - 12:03

    Major

  • Players
  • 28180 battles
  • 20,552
  • Member since:
    06-28-2011

View Postcheapbooks, on May 02 2019 - 03:04, said:

If you look at QuickyBaby's "free to play" account and his premium account, in almost every measure possible, he outperforms his 'free to play account' with his pay account.

Of course.   Less time in stock tanks will help a little,.as does tank selection, but the degree is less than what skill brings.  Typically, the better the player, the more they can make an OP tank work for them.  There are a few tanks that are OP for everyone, but few people play just one tank their whole account.  Those are outliers that are easily spotted.

 

 

Block Quote

I bet that if you take 1 skilled player vs 2 unskilled players, the unskilled players will win if they use teamwork.

 but they don't use teamwork.  That's why they are unskilled.  If they did use team work they wouldn't be unskilled

 



_Tsavo_ #233 Posted May 02 2019 - 12:12

    Major

  • Beta Testers
  • 46204 battles
  • 19,693
  • [BRVE] BRVE
  • Member since:
    02-16-2011

View Postcheapbooks, on May 02 2019 - 03:04, said:

If you look at QuickyBaby's "free to play" account and his premium account, in almost every measure possible, he outperforms his 'free to play account' with his pay account.

 

His EU usernames are QuickFingers and PlaysForFree

 

only 1 or 2 tanks perform better on his free to play account, one being a premium tank which has the same modules  in both accounts.

 

I am thinking that player skill only affects winrate noticeably if you are one of the few remaining tanks in the game. the more allies and enemy tanks die, the more of an impact individual skill has on winrate.

 

I bet that if you take 1 skilled player vs 2 unskilled players, the unskilled players will win if they use teamwork.

 

You're still rambling about this?   It has been established more times than can be counted but you refuse to accept it.   Probably because you think your ability is greater than it actually is. 

HorrorShow #234 Posted May 02 2019 - 13:56

    Sergeant

  • Players
  • 20124 battles
  • 227
  • Member since:
    06-26-2011

Logically speaking, winrate is NOT a reflection of player skill. It's purely a rate of wins, and nothing more. However, one's ability to win at a higher rate is a skill when weighted by the number of games played. Likewise, one's ability to lose at a higher rate can be considered a skill as well, when weighted by the number of games played.

 

As a skillful game winner, you:

  • Study and research tanks prior.
  • Make great strategic game decisions.
  • Communicate and strategies with your team.
  • Practice ahead of time, and develop muscle memory.
  • Purchase 100% crew training.
  • Purchase all tank upgrades.
  • Fire gold rounds.
  • Use premium tanks.
  • Use favorable tanks based on the current WoT's climate (current mechanics, rebalancing, etc.).
  • Use tanks with higher average winrates.
  • Use premium consumables.
  • Platoon only with exceptional players.
  • Use only the best hardware.
  • Use a fast ISP, and live next to the game server for low network latency.
  • Cheat. <-- This fits the list, and is not a call to cheat. Do NOT cheat!
  • ...

 

Ability to win games is a skill, weighted by number of games played.


Edited by HorrorShow, May 03 2019 - 04:17.


TankFullOfBourbon #235 Posted May 02 2019 - 14:01

    Major

  • Players
  • 30537 battles
  • 6,806
  • [DHO6] DHO6
  • Member since:
    08-10-2013
OP, what really scare me is that you might be able to vote. People like you can and are doing so much damage to your country.

HeroEnVec #236 Posted May 02 2019 - 14:36

    Captain

  • Beta Testers
  • 41299 battles
  • 1,476
  • [VILIN] VILIN
  • Member since:
    07-26-2010
You guys remember that southpark episode where they have the formula to compute actual [edited]length?  that is what this is.

_Xi #237 Posted May 02 2019 - 14:46

    First lieutenant

  • Players
  • 19761 battles
  • 711
  • Member since:
    12-11-2012
Avg Tier > WN8 > Winrate. Just how I look at it. 

H0D0R_ #238 Posted May 02 2019 - 18:44

    Captain

  • Players
  • 17376 battles
  • 1,905
  • [MAHOU] MAHOU
  • Member since:
    09-30-2013
Aye this thread is still alive. I know I have no skill people so move along.

leeuniverse #239 Posted May 02 2019 - 19:17

    Major

  • Players
  • 39072 battles
  • 8,063
  • [LAMP] LAMP
  • Member since:
    01-30-2013

The OP is both CORRECT and WRONG...

 

Skill does affect Winrate, however all the things he mentions and more ALSO affect Winrates.

In other words, it's not "just" skill that affects Winrates.

 

I've been saying this for a LONG time, and people mock completely ignoring the facts that he listed for example.

Some other factor's he didn't mention or mention as clear.

- Pre-maturely capping instead of fighting which is why you play the game artificially inflates your Winrate.

- Gaming Mouse/Computer/Screen

- Tanks that can't "flex" as well (i.e. slow) affect your winrate.

- Playing certain tanks can affect your winrate, not only whether they are slow, the armor they have, their aim-time, etc. for example even though I became good in the T-49's derp, I have a low winrate because that 8 sec actual aim-time/bloom is craphard to play with.

 

Anyway, likely forgetting some other things, it's been awhile, but the OP IS correct... lot's of things affect winrate.

The best Winrate tanks of the game are Mediums... especially armored mediums, or even also auto-loading mediums like for me, the armored med's have gotten too slow for me, the only good one left is the Obj 907 which I wish I had so I could enjoy playing an armored medium since the T-62A that I used to want is crap now.



NeatoMan #240 Posted May 02 2019 - 19:27

    Major

  • Players
  • 28180 battles
  • 20,552
  • Member since:
    06-28-2011

View Postleeuniverse, on May 02 2019 - 13:17, said:

- Pre-maturely capping instead of fighting which is why you play the game artificially inflates your Winrate.

- Gaming Mouse/Computer/Screen

- Tanks that can't "flex" as well (i.e. slow) affect your winrate.

- Playing certain tanks can affect your winrate, not only whether they are slow, the armor they have, their aim-time, etc. for example even though I became good in the T-49's derp, I have a low winrate because that 8 sec actual aim-time/bloom is craphard to play with.

 

knowing when to cap is a skill.  This is more of a thinking man's game, so if you can cap a win to "inflate" your win rate, you've just displayed better judgement (more skill) than if you fought on and lost.

 

Most people play a variety of tanks, which will tend to have some turds and more OP ones.  Unless you stick to just one tank, or one nation/class then this will also average out.

 

Knowing how to play to your tanks' strengths is also a skill.






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users