Experimental Guild Rankings

[masterpiter]

Nice idea, will motivate ppl to log in and play, its good for the game, i really like it

[Ermabwed]

As I see it, the logarithmic relation between exp and levels has been designed in the game for a reason - it is not something you want to factor away, quite on the contrary!.

I don't want to factor it away, I just don't think that a simple average as it is now says anything at all

Ermab, in your example above (two level 50 alcher's vs. 1 level 99 and 1 level 1), you compare one level 50 and one level 99: what has this to do with guild rankings? The math should be as follows:

 

guild of 2 alchers, lvl 50 & lvl 50; 100 hydro bar each: yeld=166 (they loose 17 each)

guild of 2 alchers, lvl 99 & lvl 1; 100 hydro bar each: yeld=99 (one loose 1, the other 100)

 

So, in terms of performances, I could advocate that the 50+50 should rank much better than the 99+1!

The top-rankers already have the spotlight in the individual rankings (with all due respect for their outstanding achievements!); there is no point in having a guild ranking which is so skewed as to mimic the individual rankings.

Well, that only works if the "99+1 alchers" are idiots of course It's not an assumption I usually work with, though EL forums provides enough evidence to the contrary (not talking about this thread in particular). When comparing guilds I think it's fair to assume that people will be using their skills to the best of the guild..

 

As a side note, from a purely statistical point the current system computes avg(log(exp_i)) with i over the recently logged, non-bot, non-0 level players in the guild. Blodoks' proposal would have log(avg(exp_i)). Unfortunately, the sum of experiences would exhibit a Gaussian distribution, and applying the pseudo-logarithmic conversion from exp to level would "squash" most guilds in a very narrow level zone, so we would end up with a huge number of ties -- not a good ranking system imho.

I think Blodoks meant without log, then there'd be no squashing and to make it simpler you could just count the total exp of the current level. However, the general problem here is as you say, that high level members are weighted too much...

Perhaps using the harmonic mean of the ranks would be an option to be considered here...?

[Usl]

As I see it, the logarithmic relation between exp and levels has been designed in the game for a reason - it is not something you want to factor away, quite on the contrary!.

I don't want to factor it away, I just don't think that a simple average as it is now says anything at all

Ermab, in your example above (two level 50 alcher's vs. 1 level 99 and 1 level 1), you compare one level 50 and one level 99: what has this to do with guild rankings? The math should be as follows:

 

guild of 2 alchers, lvl 50 & lvl 50; 100 hydro bar each: yeld=166 (they loose 17 each)

guild of 2 alchers, lvl 99 & lvl 1; 100 hydro bar each: yeld=99 (one loose 1, the other 100)

 

So, in terms of performances, I could advocate that the 50+50 should rank much better than the 99+1!

The top-rankers already have the spotlight in the individual rankings (with all due respect for their outstanding achievements!); there is no point in having a guild ranking which is so skewed as to mimic the individual rankings.

Well, that only works if the "99+1 alchers" are idiots of course It's not an assumption I usually work with, though EL forums provides enough evidence to the contrary (not talking about this thread in particular). When comparing guilds I think it's fair to assume that people will be using their skills to the best of the guild..

Yep, I know what you mean: in fact, some time ago I had suggested using the max value of a skill in guild, under the assumption that one would go to the best alch'ers available in guild to make hydro bars, to the best manu'er to make an armour, and so on. That kind of chart would favour guilds with several top-ranking members, but would not tell much about the guild as a whole - only about their best members.

As usual, the problem is not in the data: it is in defining data in one way, and then trying to place a different interpretation on them. The current guild rankings do not try to be descriptive of the best abilities of a guild, but rather of the average level of their members. It is a clear definition, easy to state and easy for onlookers to understand.

 

As a side note, from a purely statistical point the current system computes avg(log(exp_i)) with i over the recently logged, non-bot, non-0 level players in the guild. Blodoks' proposal would have log(avg(exp_i)). Unfortunately, the sum of experiences would exhibit a Gaussian distribution, and applying the pseudo-logarithmic conversion from exp to level would "squash" most guilds in a very narrow level zone, so we would end up with a huge number of ties -- not a good ranking system imho.

I think Blodoks meant without log, then there'd be no squashing and to make it simpler you could just count the total exp of the current level. However, the general problem here is as you say, that high level members are weighted too much...

Perhaps using the harmonic mean of the ranks would be an option to be considered here...?

 

I might have misunderstood; but I based my interpretation on

Average exp (4 players) = 11,439,950 = lvl 77

If the 5th player want to start the skill and get 1, average exp will drop at 9,151,988 = lvl 74 (-3.89%)

where, in the end, the result was again based on levels (emphasis mine).

 

Maybe you are right and an harmonic mean, as suggested by Trollson in a related thread, could be better... My suggestion would be to let the data stabilize, as Learner noted above, and then review the issue in a couple of months.

[Aislinn]

Maybe I'm missing something, but isn't the point of a guild ranking to rate the entire guild? Not just the top players, or be a "top 3 in your guild" ranking. If you have a large guild with low level players, you have a large guild with low level players. A few high levelers should not skew the rankings higher than they should be. That is why we have the individual rankings.

If anything, it should encourage those lower level players to level up. I think from all the options and suggestions I've seen so far, the current way is still the most accurate. (And I say that having had my guild kicked out of it entirely due to there only being 2 members).

 

You need to pick your priority. If being listed high up on a list is more important to you than the enjoyment you get from your players in your guild, then by all means boot them out and remake your guild (or leave yours and join a guild with only high leveled players). But if you get enjoyment out of helping newer players and feel their contribution and playing is important and want them in your guild, then that is more important than being a number on a list. People are taking this way too seriously.

[Vanyel]

It really looks like some people have too big ego and can't stand the fact their guild ain't first in sum... I mean one of rankings.

From my point of view the way Learner solved the issue is a good one. From my point of view this ranking doesn't mean much, either. I keep thinking that the best guild ain't the one with top players, but one that fits my way of playing. Hence, my best guild won't be Learner's or Aislinn's best guild. I may be wrong, though.

[Learner]

Maybe I'm missing something, but isn't the point of a guild ranking to rate the entire guild? Not just the top players, or be a "top 3 in your guild" ranking. If you have a large guild with low level players, you have a large guild with low level players. A few high levelers should not skew the rankings higher than they should be. That is why we have the individual rankings.

If anything, it should encourage those lower level players to level up. I think from all the options and suggestions I've seen so far, the current way is still the most accurate. (And I say that having had my guild kicked out of it entirely due to there only being 2 members).

And in the long run, I might have to increase the minimum number of players detected in a guild before reporting on them. What some people don't realize is also the number of guilds out there. Since I started tracking the detail level, I've spotted 218 different guilds! I've spotted 67 with one member, and 36 with only two so far.

[Lorlen]

It will be ages before this gives accurate readings. =Hc= is not really number 2 in attack and defence, if you look it says we have 7 members, not the 30 something we actually have.

 

Few weeks or months it will even out and give a much better overall picture ;>

Edited March 24, 2008 by Lorlen

[Learner]

It will be ages before this gives accurate readings. =Hc= is not really number 2 in attack and defence, if you look it says we have 7 members, not the 30 something we actually have.

 

Few weeks or months it will even out and give a much better overall picture ;>

Those Guilds interested in improving the accuracy of the readings will probably make sure their guildies and friends have their guild tag recorded by one of the Trade bots. Remember it's not just the 3 I own, but any trade bot I run as a service.

[LevinMage]

Do you have to trade with the bot or just have to be in its view?

[Learner]

Do you have to trade with the bot or just have to be in its view?

It only has to see you for it to record the guild tag for this.

[Slavia]

sorry for my ignorance but which 3 trade bots do you own?

[Vanyel]

Lemme guess, Trader, C2_trader and Tailor (in order of appearance)

[Infamous]

w00t! =Hc= second in a/d \o/

 

now we only need Scarr, Valkrie and the rest to log in again to be #1 ^^

Edited March 25, 2008 by Hardcore

[Slavia]

Lemme guess, Trader, C2_trader and Tailor (in order of appearance)

 

 

thanks very much

[Schmurk]

Please note that I do ignore very small guilds and players that 0 for that skill. The Rankings for a guild is based on the average skill of it's members, not the total or top members. The data about guild membership is still being gathered, so rankings will be changing a lot as players login and their guild gets recorded.

I don't see why players with a skill at 0 should be ignored? IMO they should be counted like others...

BTW, the players are counted as soon as they logged and cross one of your bot. So far it'll get rid of inactive players but in the future, do you plan to also get rid of new inactive players (i.e. players that didn't log for a long time)?

 

Anyway, very nice work so far!

Edited March 25, 2008 by Schmurk

[Learner]

Please note that I do ignore very small guilds and players that 0 for that skill. The Rankings for a guild is based on the average skill of it's members, not the total or top members. The data about guild membership is still being gathered, so rankings will be changing a lot as players login and their guild gets recorded.

I don't see why players with a skill at 0 should be ignored? IMO they should be counted like others...

BTW, the players are counted as soon as they logged and cross one of your bot. So far it'll get rid of inactive players but in the future, do you plan to also get rid of new inactive players (i.e. players that didn't log for a long time)?

 

Anyway, very nice work so far!

There is a setting on screen that defaults to 365, if you change it and submit, it will look at different ranges of activity, So, people can control how recent the activity needs to be.

[trollson]

How you calculate the "Ranking" for a Guild depends on what you want to be measuring. But I'd suggest that the formula should have a few simple properties:

1. An increase in level within a population should result in a increase in score.
   

Actually, just that one. I was going to suggest that increasing the population shouldn't decrease an existing score, but that depends on what you intend to measure. Certainly, guild-capability increases if you add a horde of newbies, if not your "cutting edge" capability.

 

The current formula fails (1) since members improving from skill level 0 to 1 decrease the score.

 

I was going to suggest a Lehmer Mean with p=2:

L = sum ( a[i]^p ) / sum ( a[i]^(p-1) )
L = sum ( a[i]^2 ) / sum ( a[i] )

As this absorbs the level zero characters, gives a bit more weight to cutting edge, and is independant of guild size. However, it doesn't meet (1) either, since a character going from 0 to 1 suddenly decreases the score.

Infact, a Lehmer Mean is an even

worse

model than I thought! Take a guild of all level-N's, add a newbie who starts levelling from zero. The overall guild score

decreases

for each level the newbie earns upto level N/2. Definitely not following (1).

Ah, the joy of statistical modelling. I'll be stuck thinking about this now.

 

The question I have to ask is:

So what property is guild rank intended to represent?

Update: Rephrased (1).

Edited March 25, 2008 by trollson

[Piwacket]

I found your ranking chart really interesting and posted a link in PALs forum so others could check it out also. A great idea to see how the guilds stack up against each other in different areas. Thank you for being so creative!

[Zenial]

Wow, Very nice Learner. Thank you.

 

What determines between active and non-active members? Are members who in future become inactive still have their stats counted towards guild rank?

 

Looks great

[Vanyel]

While it is interesting to have some excercise while making up new ways of counting the guild rank I'd suggest following Occam's razor principle.

[Sayre]

I prefer the KISS principle myself.

 

I really can't think of a single way to have a non-included 0 level turning into an included 1 level that doesn't lower the average (that includes all valid entities in the sample anyway).

 

On the bright side, the person shouldn't stay 1 for long, and for those guilds that give a crap, they would probably encourage the guildie to higher levels anyway.

 

S.

[LabRat]

KISS and Occam's Razor amount to exactly the same thing.

[Kaddy]

@Learner

 

Its very nice to have guild ranking. Also, some ranking changes have been done person by person as i see. But can you explain us the system for them?

 

I mean, dre has around ~120 a/d and TooMass has +130's. But dre ranked better than TooMass in CombatSkills. Kinda weird...

 

And, i m also ~140th at a/d but not top 200 in CombatSkills

 

 

Last, Can u add the allround/combatskills options to the ranking page plz?

[Tempest]

cant view my private rankings anymore guess problems with the site

Edited March 26, 2008 by Tempest

[Blodoks]

I mean, dre has around ~120 a/d and TooMass has +130's. But dre ranked better than TooMass in CombatSkills. Kinda weird...

 

And, i m also ~140th at a/d but not top 200 in CombatSkills

 

 

Last, Can u add the allround/combatskills options to the ranking page plz?

 

Your summoning?

Archery too will count in the future