Top Players List

[conavar]

While looking at the top player lists, i noticed with have got a score chart for everything from Att to Engineering and now Pk points/fights etc.

 

I think it would be nice if there was a table listing top overall players. (overall not in OA level since there is a table for that but for overall character) this would be based on a score of all there skills (excluding OA ).

 

A simple system used to work this out could be.

 

all a players skills added together and divided by the amount of skills there is.to give and average score

 

example using my own (low) skills

 

char skills: 82 + 82 + 68 + 72 + 36 + 19 + 8 + 26 + 31 + 14 = skill total 438. divided by the number of skills (10) to give an average overall char score of 43.8

 

thoughts/suggestions ?

[molime]

It was still discussed here: http://www.eternal-lands.com/forum/index.php?showtopic=33461 and yes, I'd love to see a top like that too myself. Have been considering to give the other topic a kick again after the PK points got in

[conavar]

I did wonder while looking at the list why it hadnt been suggested before, ( my bad for using wrong search key words)

 

please feel free to lock this thread

[agis29]

It would be very interesting!

[Blodoks]

78

Edited July 20, 2007 by Blodoks

[LilBear]

41.4 (I think, I did this from memory with game closed) and definitely not a top player that way but working on it

 

lilbear

[trollson]

A simple average doesn't reflect an all-rounder skill set, as a character could have one huge skill and the rest completely undeveloped, and still get a high ranking.

 

To reflect all-roundness, you need to be more interested in a character's lowest skills.

 

One formula that could do may be familer from resistors in parallel:

R = n / sumi ( 1 / Si )

Unmodified all-rounder score

Harmonic Mean

Where there are n skills (10), with values S0 ... Sn, and R is the "all rounder" score. R is guarenteed to increase if any skill increases; but will increase more when a character's lowest skill is improved.

 

Of course, if any skill is zero (0), then we get divide-by-zero, but can defined R to be zero in these cases. We could play tricks to avoid this, such as adding n to each skill, and substracting n from the result (reusing n arbitarily):

 

R = ( n / sumi ( 1 / ( Si + n) ) ) - n

Modified all-rounder score

Ideally, the addition of a new skill (such as Engineering) would not change the scores. However, I doubt this is possible without breaking the concept of "all-rounder".

 

With this formula the OP's skills give a "all rounder" score of 22.8 due to the few low values:

R = 10 / ( (1/82) + (1/82) + ... + (1/31) + (1/14)) = 22.8

Or 30.6 with the modified method:

 

R = (10 / ( (1/(10+82)) + (1/(10+82)) + ... + (1/(10+31)) + (1/(10+14)))) - 10 = 30.6

Edited July 9, 2007 by trollson

[masterpiter]

I checked this with my stats:

139+147+87+86+77+72+20+20+20+0=286(a+d) +282(other skills)=568/10= 56,8

 

As u can see, even tho my a/d gives me a lot, im still way below Blodoks in allrounder ranking

 

I plan to make 20 engi in the future( coz 20 looks pr0 ), which will give me +2 allrounder skill

 

mp

[Ermabwed]

We possibly might want to consider not just using the harmonic mean as trollson suggested but a weighted harmonic mean to reflect game mechanisms, like attack/defense skill being tied together, harvesting being a rather secondary skill etc. But on second thought this might create more strife than it's worth dealing with (proud harvesters )

[Darc0]

...

286(a+d) +282(other skills)=568/10= 56,8

...

 

286+282=668/10=66,8

 

still blows my 35,6 out of the water... lol

 

(36+35+37+39+34+37+44+35+36+23=356/10=35,6)

 

edit: wow...just wow... your right...long day at work is killing my brain I guess

 

edit2: ahah! I found it! 87+86+77+72+20+20+20+0=382, not 282, so its still 66,8

Edited July 10, 2007 by Darc0

[masterpiter]

...

286(a+d) +282(other skills)=568/10= 56,8

...

 

286+282=668/10=66,8

 

still blows my 35,6 out of the water... lol

 

(36+35+37+39+34+37+44+35+36+23=356/10=35,6)

Darc0, r u 100% sure that 2 numbers below 300 can give 600+?

[Vanyel]

Yeah, why not? Math sucks, let's be creative

 

For me it's 50.8, doesn't mean much, tho:) Long time since I've grinded for levels.

[hades]

har skills: 82 + 82 + 68 + 72 + 36 + 19 + 8 + 26 + 31 + 14 = skill total 438. divided by the number of skills (10) to give an average overall char score of 43.8

 

thoughts/suggestions ?

 

 

There is no need to complicate much but using the median is much more accurate and still pretty easy to understand then the mean because the goal here is to be an all around.

 

8 14 19 26 31 36 68 72 82 82

 

median = 33.5

[Ermabwed]

I wouldn't having 4 skills at zero call well rounded, but median would allow that.

A player with all skills at 50 would be ranked at the same position a player with (0, 0, 0, 0, 50, 50, 50, 50, 50, 50)

 

edit: played around with Spreadsheet application and harmonic mean (what trollson suggested), from the following examples (took conavar's, eMPi's and Darc0's numbers) one can see that it severly punishes low levels (as trollson mentioned, this is intended)

mean\name	  conavar  masterpiter   Darc0
harmonic		24.97	   n.d.		 34.72
modified		 29.62	  28.48		 34.96

It seems that not well rounded characters benefit (significantly) more from the modified harmonic mean (more numbers would be needed perhaps or somebody ought to have a look at the formula ), it might be better to just set the harmonic mean to zero if one value is zero (as that's the limit).

 

I noticed the harmonic mean my spreadsheet application calculated for conavar is different from the one trollson came up with for the same numbers, this is probably due to the application using a statistical version to correct multiply by *9/10, which then yields

mean\name	  conavar masterpiter  Darc0
harmonic		22.47	   n.d.			  31.25
modified		 26.66	  25.95			 31.46

 

How to calculate this for yourself: the function in openoffice calc is called harmean(), so you'll have to enter your skill levels let's say in the cells A1 to A10 (add 10 to each if you have one at zero), enter =HARMEAN(A1:A10)*9/10 (or if you had a skill at zero: =(HARMEAN(A1:A10)-10)*9/10 ) in another cell and hit enter I'm sure pretty much every spreadsheet application has this function, possibly under a different name, check the function list for the harmonic mean

 

edit: can't really get a nice table going there but it's good enough to see it I'd say

Edited July 10, 2007 by Ermabwed

[Cyprom]

Bah, just below Anshar 50,7

 

Since we have exactly 10 skills, there's not really a need to divide it by 10.. Easier to just sum up all levels

[trollson]

I was just using a calculator to work out the scores, so its more likely I mis-typed, giving the difference Erm saw.

 

For weighted harmonic mean, consider adding up all the skill sets of all the (active) characters in the game, as a source of the weights (or their reciprocals).

 

Though I don't think this really corrects for bias in non-orthogonal skills (the assumption is that all skills are independent - orthogonal - which is not the case for A & D, and to a lesser extent some of the others).

 

Principle components analysis could be used to establish a new basis for the stats, but this still assumes that the bases are orthogonal (and the weighting suggested above come from this method).

 

I think that independent component analysis may be what we want here, but I am not really familiar with this method.

 

We do want the score to be stable when new skills are added to the game, and to produce "sensible" results for characters with zero level skills. Ideally, adding a new skill to the game should not change peoples scores, until a population begins to develop that skill. This is where the weighted harmonic mean, with weights derived dynamically from the population, works well.