EOL Ranking Number help?

[devin]

Hello, this has probably been answered before but do ranking points go up automatically? Because i check how many i earned when i play a balle with alot of people and i been stuck at 9100-9150 for about the past 10 battles. some earned up to 80 points. why is my number not going up accordingly? Any insight? thanks mans!

[Lousku]

It freezes when you look it in the eyes.

[Kopaka]

In short it's something like it takes 100 best results per half year, so not all battles will count.

[gimp]

In short it's something like it takes 100 best results per half year, so not all battles will count.

which is why playing as much as possible is the best way to get higher in rank. people who aren't necessarily as good will be ahead if they play all the time. but for the most part the ranking system does the best it can. i personally pay more attention to relative rank, probably because my normal rank sucks

[Zweq]

27. Markku 2159.37
33. Xiphias [NK] 2112.93
34. nobody 2107.75
38. YEAHS 2057.79
39. Pab [SV] 2032.90
40. LazY [TAAF] 2027.02
46. Madness [SPEED] 1982.64
49. Kazan 1962.12
56. talli [EPO] 1901.74

GO GO RELATIVE RANKING

[Kopaka]

Yeah relative rank is more responsive you might say. Of course that means you can quicly loose points aswell if you make some 0 apple results or such.

[Lukazz]

i'm not even in the top 100 of the relative ranking

[Grace]

I was top10 relative ranking once.

Why is it that the relative rankings perpetually increase then? 1 year ago the top 5 maybe had over 2000 relative, now it's like 40 people.

Also, for example, players like VT have not played in the last year and have gone from ~1700 relative ranking to 2200+. Doesn't make sense.

[Zweq]

I don't really understand the need of obviously broken relative ranking. If ranking doesn't reflect skill it's just an arbitrary list. Anyway, that has been known for a long time already and no need to discuss it. However, what about ranking bat... actually never mind, cba to write any further.

[devin]

hmm so basically your best 100 "rankings" (you mean total points earned) will be the only that count towards your overall ranking score? And 6 months later your start over again building your top 100 rankings? It kinda makes sense how i went over bout 2000 points in my first 80 battles and my graph has like equaled out.

But i think it would be cool if there was ranking system relative to who you play in a battle related to how many points earned. prob not the first to give thought tho.

[Grace]

hmm so basically your best 100 "rankings" (you mean total points earned) will be the only that count towards your overall ranking score? And 6 months later your start over again building your top 100 rankings? It kinda makes sense how i went over bout 2000 points in my first 80 battles and my graph has like equaled out.

But i think it would be cool if there was ranking system relative to who you play in a battle related to how many points earned. prob not the first to give thought tho.

As i attempted to explain today before being distracted, no, this isn't how it works.

It's a constant thing. It doesn't reset every 6 months.

If your best ever performance was 6 months ago tomorrow, tomorrow evening that battle falls out of the possible battles for your ranking and your 101st best battle will fill the spot.

It's the 100 best battles in the last 6 months (from NOW.)

[Kopaka]

Why is it that the relative rankings perpetually increase then? 1 year ago the top 5 maybe had over 2000 relative, now it's like 40 people.

More people in the system means more points to earn from other people, so I guess that's the reason.

--

Almost right haru

...

To explain more fully the system: For each player you beat in battle you get an amount of points, these depends on his and yours ranking, so you get more for beating better players. The points you get from each beated player is added together as the amount you get for that battle. These points are saved for every battle you have ever played. Then to calculate your overall ranking it does this calculation: (100 best last 6 months) + (100 best 6-12 months ago * 0.5) + (100 best 12-18 months ago * 0.3) + (100 best 18-24 months ago * 0.2)

[abruzzi]

kopaka kopaka guwno kurwa sraka

/offtopic

[Stini]

It should be quite easy to make for example an ELO based rating system to determine the actual battling skills of kuskis more accurately than the current ranking system. In chess this works very well, but of course battles are somewhat different, so the system should be adjusted a bit. It could be fun nevertheless to calculate some sort of unofficial ratings based on all the battling data we have.

[Zweq]

Stini the math prodigy should be assigned to design a new ranking system !

[SveinR]

I agree we should at least try some sort of ELO system. I remember when ranking systems were discussed several years ago people didn't want to lose points if the were beaten by someone, but maybe attitudes have changed on this.

[Zweq]

Maybe stini could write some php script and skicka to kopaka to testrun with the existing data ¨=D, if it's looking good (I'm 1st) then use it on the new webshiet

[Kopaka]

It should be quite easy to make for example an ELO based rating system to determine the actual battling skills of kuskis more accurately than the current ranking system. In chess this works very well, but of course battles are somewhat different, so the system should be adjusted a bit. It could be fun nevertheless to calculate some sort of unofficial ratings based on all the battling data we have.

That's pretty much the same as the relative ranking we have now. You take points from the players you beat, and the amount depends on your ranking compared to each other. The problem is that it requires players to make their best effort at every battle they play, in EOL many might join a battle late or play one run and quit because they don't like the level. Also it might discourage people to play once they have a good rating in order to maintain it, as SveinR said the attitude was that people didn't want to be able to lose points. But if you don't do these things, I think the relative ranking we have is working pretty well, thought it might be a bit too responsive, you can lose or gain a lot of points in just one battle.

--

The main ranking is an attemp to solve these problems, but I do agree that it is not that good. I've tried to consider the following criteria:
Gain more points by beating better players
Can't lose points by playing battles
Doesn't matter how many battles you play
Recent battles are given more weight

It's inspired somewhat by Tennis aswell as international soccer team rankings. In Tennis your points are calcuated from last 12 months' tournaments. So when you play one you replace that tournament's points from last year and might gain or lose points that way. In EOL it's 24 months and the earlier results are given less weight as in soccer rankings. So in EOL when an old battle reaches a 6 month mark it's replaced by the 101st best result from those 6 months etc. Some of the problems with this systems as I see it, is 1) that you play a variable amount of opponents, so beating someone good in a battle with few players might not count at all, battles with many players gives more points generally, 2) active players will be playing a lot of battles that will never count in their ranking, as only 100 best from each 6 months count. In Tennis you can only play so many tournaments each year, whereas in EOL you might play a 1000 or more in 6 months while some barely plays a 100.

You could take out the time element and just take last 100, last 100 to 200 etc. but that brings back the problem that if you play a battle you might lose points, but calculating it by time you HAVE to play new battles in order to try to maintain your ranking, you can't just stop and stay at the top, so I think this element is good to have in some way.

However it would be nice to remove the variable of number of players in a battle. For example rather than saving the total point value from each battle, save the amount you get from each individual player you beat in the battle, and then take best 2000 or so of these, so that beating a good player in a battle with few players is as good as beating a good player in a battle with many.

I like the idea of old battles expiring so that you have to play a new to try to defend your ranking, but for active players it causes a lot of unused battles in the ranking that we take X best from last X months like explained. Any ideas how to make it possible to use points from all battles while still making it theoretically possible to be best even if you play only a 100 battles in 6 months?

I'm open for some completely different ideas too of course.

Long ass post but I hope it made some sense ;)

[Stini]

A simple way would be just to use the basic ELO system (let's call it EOLELO ) as in chess and think of each battle as a tournament where everyone plays against everyone. After the battle ends, you "win" against all those below you and "lose" against all above you in the results. Then you calculate new ratings as you would for a chess tournament. This would probably be good enough for our needs, although it might not be entirely accurate. Then you probably have to ignore results with 0 apples, weird battle modes etc. Maybe there are some other special cases as well? To get a more accurate model, you'd need to use the old battle data and do some data analysis, which isn't that difficult either really and it's actually pretty close to what I study at the university, but it would require me to get access to the data and experimenting a bit with it and I'm not really sure if it's really worth the effort. I guess this could be considered if there are some obvious faults with simpler models.

[Lousku]

LOL EL EOLELO

[Hosp]

Since this is a popular topic right now, I will use this topic since I don't know where to post it :S.

http://elmaonline.net/?s=stats&p=levelpack&lp=amle
o,o

Very many levels and I just added them, played for a while and pressed F5 when I saw Orcc had made some times, then I saw that every time I had pressed F5 it had made copys of levels. Please remove all copys. x( sorry

[Stini]

That's pretty much the same as the relative ranking we have now. You take points from the players you beat, and the amount depends on your ranking compared to each other. The problem is that it requires players to make their best effort at every battle they play, in EOL many might join a battle late or play one run and quit because they don't like the level. Also it might discourage people to play once they have a good rating in order to maintain it, as SveinR said the attitude was that people didn't want to be able to lose points. But if you don't do these things, I think the relative ranking we have is working pretty well, thought it might be a bit too responsive, you can lose or gain a lot of points in just one battle.

Some good points there. I guess in some ways it would encourage players to not just stop playing a battle if you can lose rating points, but I can see it could be a problem that it doesn't encourage playing if you come late. Although I doubt that most people would really care about their rating that much, but who knows.

It's inspired somewhat by Tennis aswell as international soccer team rankings. In Tennis your points are calcuated from last 12 months' tournaments. So when you play one you replace that tournament's points from last year and might gain or lose points that way. In EOL it's 24 months and the earlier results are given less weight as in soccer rankings. So in EOL when an old battle reaches a 6 month mark it's replaced by the 101st best result from those 6 months etc. Some of the problems with this systems as I see it, is 1) that you play a variable amount of opponents, so beating someone good in a battle with few players might not count at all, battles with many players gives more points generally, 2) active players will be playing a lot of battles that will never count in their ranking, as only 100 best from each 6 months count. In Tennis you can only play so many tournaments each year, whereas in EOL you might play a 1000 or more in 6 months while some barely plays a 100.

Rating should represent the skill of the kuskis. It shouldn't matter how much you play, which I think is the main problem with the current ranking. Also, beating many good players in a big battle is a better achievement than beating one good player at 4am when nobody plays. Also let's say you beat Zweq in a big battle and in a small battle. In ELO system you get as many points from beating him in both cases, but you probably get more points from the big battle since you beat many other players as well. On the other hand, you also lose some points to all the people that beat you in the big battles as well.

You could take out the time element and just take last 100, last 100 to 200 etc. but that brings back the problem that if you play a battle you might lose points, but calculating it by time you HAVE to play new battles in order to try to maintain your ranking, you can't just stop and stay at the top, so I think this element is good to have in some way.

I like the idea of old battles expiring so that you have to play a new to try to defend your ranking, but for active players it causes a lot of unused battles in the ranking that we take X best from last X months like explained. Any ideas how to make it possible to use points from all battles while still making it theoretically possible to be best even if you play only a 100 battles in 6 months?

In chess old results don't get expired and I don't see any problem with it. The rating should represent the skill of the player, so I find it artificial to just decrease the rating from inactivity. Also if you really got a good rating by playing well, you deserve it, and I doubt many players would stop playing because of achieving a good rating. Perhaps it could be possible to have separate ranking and rating lists? In the ranking list you could only list active kuskis (more than one battle a month or what ever) so you can't maintain the top spot in the ranking list by being inactive, but your rating would stay the same nevertheless.

[Lukazz]

In chess old results don't get expired and I don't see any problem with it. The rating should represent the skill of the player, so I find it artificial to just decrease the rating from inactivity.

I think the idea behind this is that new players have a chance to get a good ranking after a while. Of course if kopa would use the Elo system it wouldn't matter since the amount of battles you play doesn't interact with your rating anyway. But as long as this system is in use it would be almost impossible for player who weren't playing EOL from the beginning to reach the top 10 because they'd never be able to compensate the 100battles/6 months from before they started playing.

[Kopaka]

A simple way would be just to use the basic ELO system (let's call it EOLELO ) as in chess and think of each battle as a tournament where everyone plays against everyone. After the battle ends, you "win" against all those below you and "lose" against all above you in the results.

That is what relative ranking does, this post by mila explains it http://mopolauta.moposite.com/viewtopic ... 38#p134138 (k is 0.003 and q is 0.001)

[Ky.Jelly]

I agree we should at least try some sort of ELO system. I remember when ranking systems were discussed several years ago people didn't want to lose points if the were beaten by someone, but maybe attitudes have changed on this.

I don't people should get to decide how the ranks are worked out at all, the best way to rank us should be found by trial and error. We could simulate multiple ways to rank us based on the current data EOL already has. Then some sort jury (kinda like the GAA, or just the EOL people, Abula, px etc, whoever we think is suitable) takes all the simulated results and comes up with a decision on what we could use.
At the end of the day, people. WHY are we playing for rank, you should be playing each level at your best and enjoying elma, not battling or not battling so you can beat the rank maths.

[abruzzi]

I think it's impossible to come up with an entirely just ranking system and it's also not neccessary at all. People play for fun, some rating maybe would be extra cool one day but leaving only Bjorn caring about teh, the other. Zweq recently stated that he is readly for World Cup! Let's head all heads and their programming powers for this issue! Let's make this perfect!

[Mats]

If you don't play any battles in 6 months, but then 200 the next 6 month period, will they all count or only the best 100?

[Kopaka]

Only best 100, so you're screwed.

[Mats]

hah, guess I am.