Jump to content


How player skill affects the WR?

statistics work numbers WR skill

  • Please log in to reply
94 replies to this topic

Excavatus #1 Posted 19 April 2017 - 10:03 AM

    Second Lieutenant

  • Player
  • 17614 battles
  • 1,327
  • [T-D-U] T-D-U
  • Member since:
    07-28-2013

TL/DR: This is a statistical simulation ( a very simple one ) done by me in my free time..

If you want to learn what is it, just read below the original message down there.
In short, it is basically creating random skilled teams 15vs14 and adding a constant player in each of the matches and look that If that player can change the outcome.

and doing this for players with different skill levels. If you dont want to read what has happened at the start and what is the way of the simulation working,

Just follow the link below and go for the latest update and see how skill affects the outcome of a battle and how many battles cannot be affected by skill..

All the comments, additions, negative or positive critism, pointing my faults, errors etc etc.. are wellcome.. in case they are written in a civil manner..

Otherwise please GTFO! Thanks..

 

THE NUMBERS ARE CONSTANTLY UPDATED! PLEASE FOLLOW THE THREAD FROM THE LATEST NUMBERS FROM HERE

http://forum.worldof...3#entry14146403


Here is the Original thread starter!


Most of the guys who know me here knows also that I love playing with numbers, data mining, working on them etc etc..

A great aspect of the forum goers suggest that because of the random nature of the MM and not including the skill levels of the players,

the result is depending on your team, not your skill and the games decided from the beginning.

I am not talking about the rigged MM threads, they are non-sense but talking about the random MM works for you.

 

In short, it is said that, your individual skill level doesnt affect the outcome of the battle.

I want to try this in some random numbers.. and did a little numbers experiment. I will give the results here, not gonna comment on them.

Commenting part is on the community.

 

I can share my detailed work with anyone who wants to check the numbers and calculations, but it is basically very simple.

 

What did I do..

first I find an onlie die roller..

http://www.brockjone...roller/dice.htm

 

Then I throw 15xD100 and sum all the results (Automatic in the site)

then I Throw 14xD100 and sum all the results...

 

These are the skill levels of a 15vs14 matchup.

and after that, I added different skills into the sum of the 14, and look at the result.

Basically bigger number wins the battle..

 

These are the skill levels I used, The numbers is a combination of skill and the tank and the grind level etc. I am not taking them individual skill levels. I am taking them their affect on the battle.

 

50: a total average player.. carries his/her weight half of the time

60: a little competent player, who can perform good in certain tanks but who struggles with the stock grinds

70: an above average player, who plays good on certain tanks, better on some, worse on some. But mostly carry his own weight or do better than that.

80: A ver good player, who knows what to do.. carry mostly.. do better than most of the players in almost every tank he/she drives

90/100 unicums and super unicums.. mostly try/can carry all the matches they get in. they overperfom %95 and %98 of the community. Basically wot gods.

 

I threw dice for 200 times..

The average of the red teams skill after 200 battles is : 759,37

The average of the green teams skill after 200 battles (with 14 players) is : 698,94

and respectively the averages after the addition of the last players skill are; (as you can imagine)

 

(A)50: 748,94

(B)60: 758,94

©70: 768,94

(D)80: 778,94

(E)90: 788,94

(F)100: 798,94

 

The red team won, 127 of 200 matches against a 14 player green team. a %37 WR

After that I wanted to look how many matches that an individual player can turn byhimself.

 

 

 

After adding the player (A) into the roster, he turned 17 losses into the wins. - 90/110

Player (B) turned, 20 - 93/107

Player © turned, 29 - 102/98

Player (D) turned, 42 - 115/85

Player (E) turned, 43 - 116/94

Player (F) turned, 53 losses into wins. 126/74

 

Another note and I think it is important, the best player even cant help the team win 74 matches, and the red team lost 73 matches when it is even 15vs14

So basically %36,5 of the wins for our player doesnt entirely depend on him but the team. He may do nothing and stil win those matches.

and similarly, %36 of the losses for our player, doesnt depend/caused by him, because even he was a ubersuperultra unicum, he couldnt prevent the loss..

 

These are the preliminary results, I will add more results.

200 is not a huge number but it is ok I think to get some results and fire some civil discussion.

 

I am really asking for "Rigged MM" guys to stay away with their rigged mm arguments because that is not the issue here.

The sole purporse of this thread and work is to try to shed some light on the question "HOW DOES PLAYER SKILL AFFECTS THE RESULTS WHEN THE MM DOESNT INCLUDE SKILL AND BE TOTALLY RANDOM?"

 

Any data you want me to add, please ask freely and I will try,

If you have any questions about the work itself or suggestions please feel free to contribute,

But do that all in a civil way, the arguments like "YOU KNOW NOTHING NOOB! YOUR WORK IS WRONG" will not be taken seriously.

You may think I am doing something or everything wrong, please tell me WHAT IS WRONG that you think..

 

Thank you,

Please discuss..

 

PS: Here is another statistical work done by me to answer another question asked by the community numerous times.

 

http://forum.worldof...statistic-work/


Edited by Excavatus, 20 April 2017 - 01:08 PM.


jabster #2 Posted 19 April 2017 - 10:15 AM

    Field Marshal

  • Beta Tester
  • 11936 battles
  • 17,602
  • [WSAT] WSAT
  • Member since:
    12-30-2010

I did something similar a few years ago just to try and understand what sort of results you might expect from the format. The two biggest questions I was interested in was how much performance above average translates to improved win-rate and also just how many people can get unlucky with their teams. The answers were, I think, an average performance twice above average gives you about a 60% win-rate and the unlucky ones were not a lot of bugger all.

 

Edit: Why use a online RNG by the way and not a local RNG function?

 

Edit: Assuming that the player skill is 1.. 100 then 50 isn't an average player but very slightly below it.


Edited by jabster, 19 April 2017 - 11:36 AM.


Laatikkomafia #3 Posted 19 April 2017 - 10:24 AM

    Brigadier

  • Beta Tester
  • 20846 battles
  • 4,077
  • [ELC-P] ELC-P
  • Member since:
    12-27-2010

Player skill has a positive effect on WR, lack of it has a negative effect.

 

The higher the player skill, the higher the achievable WR becomes, up a point.

No matter how good one is, it is pretty much impossible for most players to retain 70%+ WR.

 

Same goes the other way around.

No matter how bad one is, it is pretty much impossible for most players to lose 70% of the played games.

 

The way you move the Maus, which keys you press and how often you watch the minimap isn't a matter of RNG.

EDIT: J, in the long run, a random player will get as much good teams, as they get "unlucky" with them.


Edited by Laatikkomafia, 19 April 2017 - 10:25 AM.


Ullakkomorko #4 Posted 19 April 2017 - 11:02 AM

    Staff Sergeant

  • Player
  • 11487 battles
  • 359
  • [HKI] HKI
  • Member since:
    04-12-2013

Your approach seems to make sense and suggests that player skill can have a clear impact on win rate.

 

Were your skill numbers (0-100) evenly spread? Would it be possible to run the simulation with some kind, maybe normal distribution, of skills across the 0-100?



Excavatus #5 Posted 19 April 2017 - 11:17 AM

    Second Lieutenant

  • Player
  • 17614 battles
  • 1,327
  • [T-D-U] T-D-U
  • Member since:
    07-28-2013

View PostUllakkomorko, on 19 April 2017 - 01:02 PM, said:

Your approach seems to make sense and suggests that player skill can have a clear impact on win rate.

 

Were your skill numbers (0-100) evenly spread? Would it be possible to run the simulation with some kind, maybe normal distribution, of skills across the 0-100?

 

skills for the team are pure random, and for the skill numbers, I used 50, 60, 70, 80, 90, 100..

For example, for 1 battle.

 

Red Team skill total,

688 (from die throws)

 

Green team skill total (14 players)

622 (from dies throws..

 

According to this, Red team wins 15v14

Player A with 50 skill level, added to the green team,

688vs672, reds win again

Player B with 60 skill level, added to the green team instead player A

688vs682, reds win again,

then the player C added,

688vs692,

Now greens win, Player C changed the loss into a win, by his/her skill..

 

What kind of results you want from a simulation, can you give me more details?

 



Enforcer1975 #6 Posted 19 April 2017 - 11:21 AM

    General

  • Player
  • 12964 battles
  • 8,318
  • [WJDE] WJDE
  • Member since:
    05-04-2014
I got carried all the way yesterday so i can't really tell tbh...

Anthony_1972 #7 Posted 19 April 2017 - 11:29 AM

    Lance-corporal

  • Player
  • 4968 battles
  • 95
  • Member since:
    03-29-2016

I don't understand how this can be an issue.   Of course skill has an influence in win rate.   It makes a different is your AFK, or actively playing.

 

What people DON'T understand is how long it takes for random chance to average out.   Lets assume a random player, playing against other random players.   Obviously the win-change is 50%.  (excluding draws)

 

From statistical theory, we actually know what the spread in results will be !   This is a Poison distribution.   Meaning that the spread from the average will be the square root of the number of games that have been played.   For a 100 games, that means a spread of 10.   That 10%!   So forget about unicum illusions when you've got a good win rate after 100 games in a certain tank.  It means absolutely nothing.  Even a 1000 games, still give a 3% error.    You have to have played 10.000 games to bring the error down to 1% !

 

And that's were the problem lies.   Obviously win rate depends heavily on skill.   But the nature of random chance means that you need A LOT of games to get an accurate win rate.     In a day of playing, or even a week, you won't get enough games to average out the randomness, and thus you will mostly just feel the randomness of the game.   If you look over a far longer time, you will notice the influence of your own skill.



shishx_the_animal #8 Posted 19 April 2017 - 11:40 AM

    Lieutenant Сolonel

  • WGL PRO Player
  • 28492 battles
  • 3,298
  • [FAME] FAME
  • Member since:
    04-06-2013
So we needed a third topic to discover that fire is hot? :hiding:

jabster #9 Posted 19 April 2017 - 11:41 AM

    Field Marshal

  • Beta Tester
  • 11936 battles
  • 17,602
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View PostUllakkomorko, on 19 April 2017 - 10:02 AM, said:

Your approach seems to make sense and suggests that player skill can have a clear impact on win rate.

 

Were your skill numbers (0-100) evenly spread? Would it be possible to run the simulation with some kind, maybe normal distribution, of skills across the 0-100?

 

I did think about that when I wrote my abstract model but I thought that the 'skill' of a player in a single game is really how they perform in a game so you then get to how to try and model variations in that performance and whether that's consistent across skill levels. In the end I decided that the added complexity wasn't worth if for what it would add so didn't bother.

jabster #10 Posted 19 April 2017 - 12:01 PM

    Field Marshal

  • Beta Tester
  • 11936 battles
  • 17,602
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View PostAnthony_1972, on 19 April 2017 - 10:29 AM, said:

I don't understand how this can be an issue.   Of course skill has an influence in win rate.   It makes a different is your AFK, or actively playing.

 

What people DON'T understand is how long it takes for random chance to average out.   Lets assume a random player, playing against other random players.   Obviously the win-change is 50%.  (excluding draws)

 

From statistical theory, we actually know what the spread in results will be !   This is a Poison distribution.   Meaning that the spread from the average will be the square root of the number of games that have been played.   For a 100 games, that means a spread of 10.   That 10%!   So forget about unicum illusions when you've got a good win rate after 100 games in a certain tank.  It means absolutely nothing.  Even a 1000 games, still give a 3% error.    You have to have played 10.000 games to bring the error down to 1% !

 

And that's were the problem lies.   Obviously win rate depends heavily on skill.   But the nature of random chance means that you need A LOT of games to get an accurate win rate.     In a day of playing, or even a week, you won't get enough games to average out the randomness, and thus you will mostly just feel the randomness of the game.   If you look over a far longer time, you will notice the influence of your own skill.

 

I always think that's a somewhat misleading way of how being lucky changes your win-rate. To me the real question is how likely it is that variation is down to bad luck compared to that's what you expect. Every player has the possibility (when using what is essential an arbitrary definition of accurate) of being +/-3% is very different from every player does.



Useless_Pathetic_Stupid #11 Posted 19 April 2017 - 12:42 PM

    Warrant Officer

  • Player
  • 38450 battles
  • 588
  • Member since:
    12-19-2012
Great lab maths.

However WOT matchmaking is not 'random'.

Where are the 15 v 15 arty games?

Where are the 15 v 15 light tank games?

Where are the 15 v 15 tier10 heavy games?

The game is completely unbalanced due to the matchmaking. Most games these days are decided before the countdown is ended. Regardless of player skill.

Your random mathematics take equality of chance into consideration.

WG does not apply the same fairness to its matchmaking

End of thread

jabster #12 Posted 19 April 2017 - 12:48 PM

    Field Marshal

  • Beta Tester
  • 11936 battles
  • 17,602
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View PostStr0nkTenk, on 19 April 2017 - 11:42 AM, said:

Great lab maths.

However WOT matchmaking is not 'random'.

Where are the 15 v 15 arty games?

Where are the 15 v 15 light tank games?

Where are the 15 v 15 tier10 heavy games?

The game is completely unbalanced due to the matchmaking. Most games these days are decided before the countdown is ended. Regardless of player skill.

Your random mathematics take equality of chance into consideration.

WG does not apply the same fairness to its matchmaking

End of thread

 

So you want a model that takes into account that the MM is rigged against you personally for reasons?

Ullakkomorko #13 Posted 19 April 2017 - 01:32 PM

    Staff Sergeant

  • Player
  • 11487 battles
  • 359
  • [HKI] HKI
  • Member since:
    04-12-2013

View PostExcavatus, on 19 April 2017 - 11:17 AM, said:

What kind of results you want from a simulation, can you give me more details?

 

I'm assuming that player skills have a bell curve type distribution (I'm not too good at statistical terms) with the majority of the players concentrated near the average skill (50 in your simulation).

 

I was wondering what such a distribution would have on the results. If I understood correctly, the probability of a (simulated) players skill being either 0, 50 or 100 were equal. Whereas actually we'd be much more likely to have a player of, say, 45 than someone with skill of only 1.



jabster #14 Posted 19 April 2017 - 01:37 PM

    Field Marshal

  • Beta Tester
  • 11936 battles
  • 17,602
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View PostUllakkomorko, on 19 April 2017 - 12:32 PM, said:

 

I'm assuming that player skills have a bell curve type distribution (I'm not too good at statistical terms) with the majority of the players concentrated near the average skill (50 in your simulation).

 

I was wondering what such a distribution would have on the results. If I understood correctly, the probability of a (simulated) players skill being either 0, 50 or 100 were equal. Whereas actually we'd be much more likely to have a player of, say, 45 than someone with skill of only 1.

That's true but it quite different to how a player performs in a single game. 



Long_Range_Sniper #15 Posted 19 April 2017 - 02:38 PM

    Major General

  • Player
  • 23469 battles
  • 5,564
  • [T-D-U] T-D-U
  • Member since:
    04-04-2011

I read an interesting article a few years ago about chess, randomness and skill.

 

On the one hand chess could be the most skillful and balanced game. No teams, balanced map, balanced pieces, and the only random element being a coin toss to decide who is black or white. There is even debate as to which is better, to play first or play second, but hopefully the most ardent of tin foil WoT players wouldn't argue that a coin toss is a 50/50 random event.

 

So if you take chess as a pure example of a skill based game, what would happen if you introduced a coin toss at different part of the game. When the coin toss is at the beginning, it does not change the skill aspect of the game. The results over time would then tend to be a pure skill distribution. If it was me against a grand master over 10 games it would probably be 10/0 to them!

 

But what if the coin toss was at the end of the game? What if it didn't matter how you played, at the end of the game there's a coin toss and if the loser wins the coin toss they're declared the winner. This is the total opposite of the coin toss at the start. So no matter how well you play a coin toss decides the winner. Skill goes out the window and you get a simple 50/50 "skill" distribution.

 

For the third scenario, what if the coin toss was after each move of a piece. The coin toss winner then gets to choose who wins/loses the move, even if they didn't make it. So you could have a queen take a pawn, toss a coin, and the pawn wins! Nobody would want to move their queen would they? The chess equivalent of camping. But you'd risk moving your pawn maybe.....

 

With this approach the coin toss doesn't tell the player where to move, or which piece to move. This is still skill. What it does do is determine who wins the move. What this does is take the edge off the skill of the better player, and give some help to the poorer player. But over time, the chances are the better player will still make the right moves, move the right pieces, and with a 50/50 coin toss they'll come up on top in the long run.

 

This is a prefect example of how skill distributes itself in WoT. There are random elements, but unless there was skill involved you'd be left with the 50/50 distribution, just like the coin toss at the end of a chess game. Equally if there was nothing random in the game you'd have the 10/0 whitewash scenario.

 

I remember this article, and it reminds me why skill and winrate correlate in WoT, whilst still having random elements in the game.



Ullakkomorko #16 Posted 19 April 2017 - 02:49 PM

    Staff Sergeant

  • Player
  • 11487 battles
  • 359
  • [HKI] HKI
  • Member since:
    04-12-2013

View PostLong_Range_Sniper, on 19 April 2017 - 02:38 PM, said:

chess

 

You could have a team chess game where every player takes turns to move a piece without having the chance to communicate with each other apart from being able to call the players on your own team noobs and worse! Imagine the frustration of the unicum grandmaster in a team of reds beginners.



250swb #17 Posted 19 April 2017 - 02:50 PM

    Lieutenant Сolonel

  • Player
  • 16292 battles
  • 3,293
  • [WJDE] WJDE
  • Member since:
    09-23-2015
It's a convincing calculation, but then I don't need to be convinced that player skill affects WR. The players that need convincing are those with below average WR and the calculation only deals with average and above average players, not those that negatively affect their teams most of the time. If they could see how their lack of skill affects WR it would be a bigger spur to do better than maintaining the illusion it is purely MM.

Excavatus #18 Posted 19 April 2017 - 03:10 PM

    Second Lieutenant

  • Player
  • 17614 battles
  • 1,327
  • [T-D-U] T-D-U
  • Member since:
    07-28-2013

View Postshishx_the_animal, on 19 April 2017 - 01:40 PM, said:

So we needed a third topic to discover that fire is hot? :hiding:

 

View Post250swb, on 19 April 2017 - 04:50 PM, said:

It's a convincing calculation, but then I don't need to be convinced that player skill affects WR. The players that need convincing are those with below average WR and the calculation only deals with average and above average players, not those that negatively affect their teams most of the time. If they could see how their lack of skill affects WR it would be a bigger spur to do better than maintaining the illusion it is purely MM.

 

Clearly as you can see people still think that the battles decided from the start, and skill doesnt contribute at all..

Yes all my efforts ay be futile to put some sense and proof into the sensless and idiotic minds..

But you can say I simply love to play with numbers and dont have something to do today...

 

On the other hand, for the answer of the skill distribution,

I didnt put individual players skills into account. I didnt put them in writing.

I only put down the sum.. so I cant say anything about the variance of that..

But may be If I have enough time, I can do more detailed work towards that point..

 



rustysprite #19 Posted 19 April 2017 - 03:26 PM

    Captain

  • Player
  • 20792 battles
  • 2,099
  • [WJDE] WJDE
  • Member since:
    03-18-2012

What is worrying me is, how did Exca ninja sneak into the Dingers with this massive brain that he has and managed to keep it a secret from us?!!  ......

 

He's more cleverer than he looks

 

Must change his name to Rainman

 



Conor_Notorious_McGregor #20 Posted 19 April 2017 - 03:40 PM

    Lieutenant General

  • Player
  • 21581 battles
  • 7,557
  • [NERVA] NERVA
  • Member since:
    09-09-2014

I offer praise for your efforts OP however I feel that this may well be very much like the flat earthists and thier denial of reality.

 

Not even science and maths seems to be able to penetrate the fog surrounding some s-humans.

 

We can hope though..:great:

 

 







Also tagged with statistics, work, numbers, WR, skill

1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users