AOE-II Your ELO in perspective (ELO Distribution RM 1v1)

[Scogo_]

Hey peoples, spontaneouously came up with the idea to do this.

hf feeling like a pro / nub / average player

(made with MS paint xd)

 

[GmanStreams]

Does this also take 0 game accounts into account?

 

[Scogo_]

Does this also take 0 game accounts into account?

nope. I should mention the weaknesses of this: 0 game accounts as well as multiple accounts and non-2000-elo retired accounts are not taken into account (i.e. not taken out of consideration as they should).

That would take me at least 10 times more time to do : p

I do not believe tho that it would make much of a difference especially with respect to the percentile numbers. So Overall this is still very accurate.

 

[King_Marv]

noob get 19xx


joke well done - interesting figure

 

[Moonspring]

Woot. I beat 60% of the players and thus I am super average.

 

[robo]

Does this include NPL ladder?

 

[misafeco]

Does this include NPL ladder?

No, it says RM 1v1 ladder.

 

[Scogo_]

noob get 19xx


joke well done - interesting figure

haha I believe i'll be there soon hopefully (after exams : p)

thanks

 

[Poxo]

How did you get all this data?

Nice info btw, makes me feel better

 

[Scogo_]

How did you get all this data?

Nice info btw, makes me feel better

The only data I needed was the elo list from rm 1v1 ladder, the rest is just maths xd

 

[drmrboss]

Strange! Mathematically, the number of people below 1600 and above 1600 should be equal,as 1600 is the medium (initial elo for everyone).

 

[robo]

There is a rating boost for the first 10-20 games played, where you gain/lose more points than normal to help players get to their "correct" rating faster.

 

[nimanoe]

On top of that there is decay, so players lose some ELO without other people gaining any

 

[JoshuaR]

Strange! Mathematically, the number of people below 1600 and above 1600 should be equal,as 1600 is the medium (initial elo for everyone).

I can think of a simple case where that is not true: If there are three players in the system, and two of them (A and B) always beat C and only ever play vs C, then both A and B will be > 1600, and C will be < 1600.

 

[Slaiter]

Strange! Mathematically, the number of people below 1600 and above 1600 should be equal,as 1600 is the medium (initial elo for everyone).

That's not true at all, the fact that everyone starts at 1600 ELO doesn't mean it's the medium. The medium it's the value that has 50% of players below and ther other 50% avobe it , in ELO in this case. Following OP's image it would be something from 1500 to 1600, closer to 1600 so I'd say something around 1570-1580

 

[Poxo]

There is a rating boost for the first 10-20 games played, where you gain/lose more points than normal to help players get to their "correct" rating faster.

Not on 1vs1, that's only on RM TG.

 

[drmrboss]

That's not true at all, the fact that everyone starts at 1600 ELO doesn't mean it's the medium. The medium it's the value that has 50% of players below and ther other 50% avobe it , in ELO in this case. Following OP's image it would be something from 1500 to 1600, closer to 1600 so I'd say something around 1570-1580

What? Unless people smurf and quit playing, the elo should be like that. In sample of 100 people, 50 people play each other, 50 win 50 lose. 50 % get 1616 rating meanwhile another 50% get 1584 rating. If the winner keep playing each other and the loser keep player each other, the total number of distribution will be the same on both positive and negative sides.(medium will still be 1600). And again when the loser 1584 win 1616, both will get 1600, so the medium will be the same. The scenario of unbalanced distribution happens when the better players make multiple smurf accounts, such as MBL.

 

[drmrboss]

On top of that there is decay, so players lose some ELO without other people gaining any

I agreed that, it must be the main reason behind it.

 

[Slaiter]

What? Unless people smurf and quit playing, the elo should be like that. In sample of 100 people, 50 people play each other, 50 win 50 lose. 50 % get 1616 rating meanwhile another 50% get 1584 rating. If the winner keep playing each other and the loser keep player each other, the total number of distribution will be the same on both positive and negative sides.(medium will still be 1600). And again when the loser 1584 win 1616, both will get 1600, so the medium will be the same. The scenario of unbalanced distribution happens when the better players make multiple smurf accounts, such as MBL.

Not only the best players make smurf accounts, I think more or less everyone at some point might have 1 or even couple smurf accounts. Your calculation for the medium of the distribution being 1600 could be right if you don't take into consideration that after certain ELO there's points decay if you don't play an amount of games per period of time, meaning this is not a zero-sum system (everything gained by someone is lost by someone else), so the 50% of ELO would be lower than 1600, not much because of what you explained but still a bit

 

[drmrboss]

Hey, I got those stats from chess.com website from data of 5 million chess players. The two figures are my chess rating at 3 mins blitz (2046 )and 1 min bullet(2006) compared with others. The initial elo for everyone in that site is 1200, however, average rating (medium) is 1036.31 and 1041.36. So average elo (medium) is below the initial elo.

 

[drmrboss]

Btw, there is no decay system on rating on chess.com website.

 

[LordTrolldemort]

The fact that the median isn't equal to 1600 has nothing to do with smurf accounts or rating decay. The MEAN of the distribution would always be 1600 without rating decay since the net change in points for both players combined in any 1v1 game is zero. but the median can be significantly different due to outliers on the high ELO end skewing the mean higher than the median.

Easy example: 4 players start at 1600. Player 1 plays all of the other players once starting with Player 2. Player 1 takes 16 points from Player 2 then 15 points from Player 3, then 14 points from Player 4 (or something like that) so the end ELOs are:

1584
1585
1586
1645

Which makes the median 1585.5 and the mean 1600

 

[Poxo]

You guys are not considering moderation

 

[Tarsiz]

Median value and average are different things altogether.
Average elo should be 1600 indeed since the amount of points gained is equal the number of points lost after every game.
Median elo is the elo that separates the set of players in two sets of equal size, there is no reason to assume it should be exactly 1600.

 

[Scogo_]

"Average elo should be 1600 indeed since the amount of points gained is equal the number of points lost after every game."

An example to the opposite has been posted above tho, see the chess dot com statistics. The initial elo for every player over there is 1200, yet the average is around 1000.

Btw. the average elo here is roughly 1592.