Fixing (not only) seasonal Elo - Part 2: Considering ranks when calculating Elo

To prevent what is commonly called Elo-farmers (i.e. highranks who gained most of their Elo by dueling lowranks) from distorting the leaderboard, there should be a slight modification to the original Elo-system. The idea is to consider the rank difference of players in certain cases, without impairing the benefits of the time-proven Elo system.

The modification only affects highranks who farm unexperienced lowranks with little to no experience at dueling.

Abstract: If one of the duelists has a rank lower than 10 and at the same time an Elo-rating of lower than 1150, the rank difference between the players gets multiplied by 100 and added to the higher rank's Elo for the process of calculating the gain/loss of Elo-points at the end of the duel.


Example: Consider players Alpha and Beta, where Alpha is a rank 7 with an Elo-rating of 1200 and Beta a rank 5 with an Elo-rating of 1000. The rank difference of two ranks will account for 200 Elo-points in my model, which are added to the 1200 Elo-points of player Alpha for the calculating process. So we have a token-Alpha with a rating of 1400, which together with the 1000 Elo-points of player Beta will be used instead of the original 1200 to calculate the gain/loss for each player at the end of the game, using the original Elo formula.


This would only apply when one of the players has a rank lower than 10 and a rating of lower than 1150, since higher ranks have all the upgrades anyway, and everyone with an Elo rating of 1150 or higher should be expected to have understand enough about the system to know that he has an upgrade-disadvantage.

Credit to Goblin and Desu.

No

Tophats would be unpleased, sir...

Escrito por Eagle, 09.04.2015 at 08:58

No

Why not. In Kord's thread you complained about him playing mostly lowranks, so what exactly are you for then.

Escrito por Tirpitz406, 09.04.2015 at 09:00

Tophats would be unpleased, sir...

I doubt it. We almost always had the same opinion on everything.

Escrito por Guest, 09.04.2015 at 09:06

Escrito por Eagle, 09.04.2015 at 08:58

No

Why not. In Kord's thread you complained about him playing mostly lowranks, so what exactly are you for then.

Escrito por Tirpitz406, 09.04.2015 at 09:00

Tophats would be unpleased, sir...

I doubt it. We almost always had the same opinion on everything.

well, i think it would suck for him, rank 13 vs rank 11 would gain him a huge disadvantage against goblin

Escrito por Tirpitz406, 09.04.2015 at 09:13

well, i think it would suck for him, rank 13 vs rank 11 would gain him a huge disadvantage against goblin

Escrito por Guest, 09.04.2015 at 08:18

This would only apply when one of the players has a rank lower than 10, since higher ranks have all the upgrades anyway. So the Elo formula for duels of ranks higher than 9 would stay the same as now.

Why don't you read my post in its entirety.

Escrito por Guest, 09.04.2015 at 09:06

Escrito por Eagle, 09.04.2015 at 08:58

No

Why not. In Kord's thread you complained about him playing mostly lowranks, so what exactly are you for then.

Escrito por Tirpitz406, 09.04.2015 at 09:00

Tophats would be unpleased, sir...

I doubt it. We almost always had the same opinion on everything.

I mostly complained about him kicking people while taking Wales

Make rank 10 cap for everyone rank 10 or higher since there's no difference in term of amount of upgrade you have

Escrito por Eagle, 09.04.2015 at 09:24

I mostly complained about him kicking people while taking Wales

Ok understood. Can you name me any downsides of my suggestion.

Escrito por Htin, 09.04.2015 at 09:28

Make rank 10 cap for everyone rank 10 or higher since there's no difference in term of amount of upgrade you have

This is exactly what I wrote in my opening post.

No. This is an arbitrary bias against higher ranks.

Your average rank 4 is less skillful than your average rank 13. Thus, the average rank 4 will lose more duels and thus have a lower elo score. That's how the elo system works.

I fail to see why, exactly, less skilled players losing elo to more skilled players is a problem. That is the entire point!

No support at all.

Escrito por International, 09.04.2015 at 10:06

No. This is an arbitrary bias against higher ranks.
Your average rank 4 is less skillful than your average rank 13. Thus, the average rank 4 will lose more duels and thus have a lower elo score. That's how the elo system works.
I fail to see why, exactly, less skilled players losing elo to more skilled players is a problem. That is the entire point!

The Elo system is still in place with this change, the only difference is that it includes rank in the equation. Basically it encourages high ranks to duel each other instead of 'farming' low ranks. The original Elo system comes from chess, where skill is the only component influencing performance, while in atWar you have upgrades which distort the system and create a bias in favor of high ranks, which the original Elo doesn't have to consider. What you are saying is right, low ranks most of the time are less skilled than high ranks, and my modification won't change the fact that less skilled players lose more often, it just reduces the amount of Elo gained by a high rank when all he does is playing low ranks.

Given the rather low duel-activity in atWar, the highest rated players still stagger around in the 1500-tier Elo, which is simply too low to work accurately in showing players' skill. The higher the ratings the more accurate the system gets, this is an inherent mathematical feature of the Elo-formula. There are still people in the top 20 who wouldn't be considered skilled players, but got there by playing very low ranks for hundreds of games. The Elo-system was implemented to rate players by their skill, not who can beat the most unexperienced players who lack even the basic upgrades.

I do not randomly suggest this, I watched the development of the ratings for two years and can with certainty say that the activity isn't high enough for the Elo-system to work in its original form. I wanted it to do so more than anybody else, but it doesn't. The suggestion I made simply includes rank in the equation, it doesn't affect the way Elo works.

I think it´s not bad idea.

If i have understood right, the only difference finally is that the high rank could loose more points in the case he would be defeated and the low rank would earn more points if he would win the duel, right?

In that case it looks positive.

Escrito por hormigaatomica, 09.04.2015 at 10:49

If i have understood right, the only difference finally is that the high rank could loose more points in the case he would be defeated and the low rank would earn more points if he would win the duel, right?

Exactly. Ranks to a certain extent would be treated like Elo, but only for the process of calculating the gain/loss at the end of each game and only for duels between high ranks and low ranks.

In that case it could be positive.

The only difference is that if before he could win 4 points now would win 2 for example.

But it won´t stop the player to go on dueling low ranks if he wants, cause, as international said before, a rank 4 is still a rank 4 and the % of posibilities of winning or loosing are the same.

As cons, i think more high ranks would make alt acounts.

I´m up too for some kind of restriction like 2 or 3 groups of ranks as we wrote before, like leagues, in the way a rank 10, 11 or 12 could never duel a rank 3 or 4 or 5 for example.

Escrito por Guest, 09.04.2015 at 08:18

Since some people disagreed with rank restrictions for dueling, and it was pointed out to me that a reset would produce a new problem, I tried to come up with another solution, inspired by Goblin.

The idea is to consider the rank difference of players, without impairing the benefits of the time-proven Elo system.
For this we would modify the way Elo-points are calculated at the end of a duel as in the following example: Consider players Alpha and Beta, where Alpha is a rank 7 with an Elo-rating of 1200 and Beta a rank 5 with an Elo-rating of 1000. The rank difference of two ranks will account for 200 Elo-points in my model, which are added to the 1200 Elo-points of player Alpha for the calculating process. So we have a token-Alpha with a rating of 1400, which together with the 1000 Elo-points of player Beta will be used instead of the original 1200 to calculate the gain/loss for each player at the end of the game, using the original Elo formula.
This would only apply when one of the players has a rank lower than 10, since higher ranks have all the upgrades anyway. So the Elo formula for duels of ranks higher than 9 would stay the same as now.

Abstract: If one of the duelists has a rank lower than 10, the rank difference between the players gets multiplied by 100 and added to the higher rank's Elo for the process of calculating the gain/loss of Elo-points at the end of the duel. This way we still have the original Elo-formula in use for duels of highranks, while rewarding duels with players of similar rank.

i still don't get the idea, why do you need to modify elo?, if there is a duel between a rank 11 and a rank 5, i would bet by rank 11, cause he is probably better player, and that is exactly for what elo is.

Escrito por Zeeroo, 09.04.2015 at 11:12

i still don't get the idea, why do you need to modify elo?, if there is a duel between a rank 11 and a rank 5, i would bet by rank 11, cause he is probably better player, and that is exactly for what elo is for.

Because a rank 11 can reach 1700 Elo by playing only players with 1000 Elo. If he does so by playing rank 4s, which have an obvious disadvantage because of a lack of upgrades, then he could stay #1 in the Elo-ranking for another one or two years without ever playing someone else with a rating higher than 1000. But he obviously isn't the best duel-player, he only plays the most lowranks.
Other skilled players who only play frequently dueling highranks, which are also ranked high but not as high as our first player, will for many more months or years fluctuate at a rating of around 1500 or 1600, even though they are objectively more skilled than the first player and could easily beat him in duel.
So the point is that the current Elo-ranking doesn't show skill, but that's what it was implemented for.
And this why we need to modify the system.

Escrito por Guest, 09.04.2015 at 11:19

Escrito por Zeeroo, 09.04.2015 at 11:12

i still don't get the idea, why do you need to modify elo?, if there is a duel between a rank 11 and a rank 5, i would bet by rank 11, cause he is probably better player, and that is exactly for what elo is for.

Because a rank 11 can reach 1700 Elo by playing only players with 1000 Elo. If he does so by playing rank 4s, which have an obvious disadvantage because of a lack of upgrades, then he could stay #1 in the Elo-ranking for another one or two years without ever playing someone else with a rating higher than 1000. But he obviously isn't the best duel-player, he only plays the most lowranks.
Other skilled players who only play frequently dueling highranks, which are also ranked high but not as high as our first player, will for many more years fluctuate at a rating of around 1500 or 1600, even though they are objectively more skilled than the first player and could easily beat him i n duel.
That's why we need to modify.

and? you want to modify the great ELO because some players abuse of it? besides, ELO gives nothing,only a number that show if you are a good player, if you reach 1700 ELO by framing rank 4, it doesn't mean nothing

Escrito por Zeeroo, 09.04.2015 at 11:29

and? you want to modify the great ELO because some players abuse of it? besides, ELO gives nothing,only a number that show if you are a good player, if you reach 1700 ELO by framing rank 4, it doesn't mean nothing

You are contradicting yourself. It shows whether I am a good player but it means nothing?

And no, I don't want to modify the great Elo because some players abuse it, as I said the formula stays the same. I want it to take into account a reality of atWar which is different from that of Chess: Lower activity and another factor besides skill, namely the existence of upgrades.
What my modification does is that it speeds something up that would happen anyway: The skilled players rise to the top while the less able ones drop in rating.
One can only gain so many points by playing mostly lowranks, which happen to be the players who have 1000 Elo most of the time. Still it is possible to gain a considerable amount of points this way, and due to the low activity in atWar this leads to an overrepresentation of less skilled players where they shouldn't be: In the top ranks.

Escrito por Guest, 09.04.2015 at 11:51

Escrito por Zeeroo, 09.04.2015 at 11:29

and? you want to modify the great ELO because some players abuse of it? besides, ELO gives nothing,only a number that show if you are a good player, if you reach 1700 ELO by framing rank 4, it doesn't mean nothing

You are contradicting yourself. It shows whether I am a good player but it means nothing?

And no, I don't want to modify the great Elo because some players abuse it, as I said the formula stays the same. I want it to take into account a reality of atWar which is different from that of Chess: Lower activity and another factor besides skill, namely the existence of upgrades.
What my modification does is that it speeds something up that would happen anyway: The skilled players rise to the top while the less able ones drop in rating.
One can only gain so many points by playing mostly lowranks, which happen to be the players who have 1000 Elo most of the time. Still it is possible to gain a considerable amount of points this way, and due to the low activity in atWar this leads to an overrepresentation of less skilled players where they shouldn't be: In the top ranks.

im saying that the function of elo is to show if a player is good, however, a 1700 elo obtained by farming noobs doesn't means nothing, a quick view at the profile can show this

Escrito por Zeeroo, 09.04.2015 at 12:02

im saying that the function of elo is to show if a player is good, however, a 1700 elo obtained by farming noobs doesn't means nothing, a quick view at the profile can show this

Agreed. But we can still optimize the system and prevent those people from being in the top ranks in the first place.

Escrito por Guest, 09.04.2015 at 09:32

Escrito por Eagle, 09.04.2015 at 09:24

I mostly complained about him kicking people while taking Wales

Ok understood. Can you name me any downsides of my suggestion.

Escrito por Htin, 09.04.2015 at 09:28

Make rank 10 cap for everyone rank 10 or higher since there's no difference in term of amount of upgrade you have

This is exactly what I wrote in my opening post.

oh ok didn't see, Did you altered it

Escrito por Htin, 09.04.2015 at 14:31

oh ok didn't see, Did you altered it

No it was like this since the beginning, there was one typo though.

Escrito por Guest, 09.04.2015 at 14:34

Escrito por Htin, 09.04.2015 at 14:31

oh ok didn't see, Did you altered it

No it was like this since the beginning, there was one typo though.

lawl that typo must have change that specific part hmm

Escrito por Htin, 09.04.2015 at 14:39

lawl that typo must have change that specific part hmm

Citação:

This would only apply when one of the players has a rank lower than 10

Citação:

So the Elo formula for duels of ranks higher than 9 would stay the same as now.

Citação:

If one of the duelists has a rank lower than 10

I mistyped 10 instead of 9 in the second sentence, so I wrote it incorrectly one time out of three. It says 'since higher ranks have all the upgrades anyway' though in the first sentence, so it is clear from the context what I meant.

Escrito por clovis1122, 10.04.2015 at 06:32

I expect different people to give their point of view about this suggestion.

So do I. This won't prevent me from addressing concerns and possibly changing my suggestions in a way that will satisfy most of the people.
If you have a problem with this just stop visiting my threads.

Cleaned(deleted) a few posts. Any more off topic posts will be removed. I don't have enough time to clean Cow's first topic, nor any number of other threads on atWar, but at least this thread I can cut the waste at the start.

On topic: Variable ELO gain/loss by rank does have merit, I believe it is a good thing for players of comparable rank to compete. Things should be standardized as much as possible, specially when you recognize what Cow is talking about here. Farming low ranks wouldn't be such a problem if there was thousands of duelists all competing and all had the same chance in a game(i.e. same upgrades), but on atWar we don't have that kind of activity nor do we have a level playing field.

When I brought about the Cln Wars seasonal ELO, and the administrators brought seasonal ELO to regular duels via my comments, I had thought seasons were enough time. And I am right, it's a lot of time for coalitions, but not enough to develop multiple players with 1500+ ELO from duels. It takes time to gather 3 players to play against 3 other players in a CW, and while farming is possible it's unlikely you can keep the top 4-5 without facing the other top 4-5. It's hard to farm constantly in CW's as there's not enough targets, and those that stick around become dangerous themselves. In duels, it seems you can just sit and farm r3/4's constantly as they don't know what they're contributing to when they duel an ELO farmer, and it's easier to arrange. Simply not enough dedication.

This proposition helps slow the farming of low ranks, but does have its flaws. It doesn't fix that an alt low rank can just farm actual low ranks and rise his ELO to 1300~ easily enough. When this alt goes up against a r10 of similar skill he doesn't lose much when he loses because of the rank difference in ELO this idea has created.

Perhaps this "ELO rank variable" you propose can only apply to ELO of 1100 or less. Anything higher and ELO on both sides of the duel shall be treated normally with no calculation difference between a r5 and r10 if they duel each other. This protects low ranks of 1000 ELO, doesn't give high ranks as many easy targets to farm, and lets alt accounts duel without an unfair advantage in calculation.

Escrito por Desu, 10.04.2015 at 08:12

This proposition helps slow the farming of low ranks, but does have its flaws. It doesn't fix that an alt low rank can just farm actual low ranks and rise his ELO to 1300~ easily enough. When this alt goes up against a r10 of similar skill he doesn't lose much when he loses because of the rank difference in ELO this idea has created.

Perhaps this "ELO rank variable" you propose can only apply to ELO of 1100 or less. Anything higher and ELO on both sides of the duel shall be treated normally with no calculation difference between a r5 and r10 if they duel each other. This protects low ranks of 1000 ELO, doesn't give high ranks as many easy targets to farm, and lets alt accounts duel without an unfair advantage in calculation.

You are right about the alt problem. I would maybe raise the threshold a little to a rating of 1150, to make sure that the lowrank understood enough about Elo at this point to not unhesitatingly duel any highrank he runs into.
I like the general idea behind this, taking away from the farmers their major victims. I will edit my opening post to include this.


I'm not sure if the people who disagreed understand exactly what I'm trying accomplish: My modification only affects highranks who farm unexperienced lowranks who have little experience at dueling, for everyone else nothing changes. And with the modification Desu suggested it's still possible for highranks to duel lowranks without being affected by my modification, just not the unexperienced ones.

Id support giving this a go, it would solve a lot of the elo farms, also remove the duel req button too.

I support cow and his suggestion

Escrito por Permamuted, 11.04.2015 at 11:24

Id support giving this a go, it would solve a lot of the elo farms, also remove the duel req button too.

Disagree with the second point, the not premium players sometimes want to duel too and it´s the only way.