Implement the Elo ranking system

[pancakemix]

Concise description:

• Implement the Elo ranking system in place of the current scoring system

Specifics/Details:

• Compaints about farming and point dumping abound on this site, mostly because there is a scoring system in place which allows this to happen. If the Elo ranking system (the system used internationally in chess leagues) were used instead, this would prove beneficial to all the site's patrons. The formula and an explanatory example can be found here.
• New recruits can either be set automatically at 1000 (the standard beginner rank in chess leagues) or undergo a "placement period" in which they are ranked 1000 for calculation purposes only.
• For team games, the total of the teammates' scores can be taken.

How this will benefit the site and/or other comments:

• Would virtually eliminate farming of new recruits
• I was actually surprised to find that this had NOT been formally suggested (brought up at least once, but not in its own thread in the proper format).

[Metsfanmax]

Implementing the Elo system would almost certainly require a wipe of all scores, or at the very least a re-scaling of the current ones, which I'm not convinced people would like very much.

One thing to keep in mind with Elo is that in chess, there is generally an intentional separation of rankings within divisions, so that your IMs or GMs are not playing your Class C players. On Conquer Club, there is no such enforced separation in rated games which abound in casual and tournament play. The Elo system discourages playing against people who are much lower rated than you, because you can lose a lot of points but not win very many. The atmosphere on this site is that higher ranked players are encouraged to play against lower ranked players; penalizing them more for doing so will make this go away. Yes, it will also help against farming, but then it would start to ruin the casual atmosphere of the site.

Finally, the underlying theory of the Elo system is that it is a true measure of a player's skill; it makes no sense as a statistical system if it is not based on some measure of a player's ability. But let's face it, having a high rank on CC does not correlate strongly with facility at the game. It correlates strongly with how many team games you play on maps you know well.

[greenoaks]

how can i put this - N O

[jpcloet]

The site inherently uses ELO, however, you will need to define better the K factor which could be different by game type. When you talk about farmers on the site, they usually are at a point where they only pick up 1 to 6 points in a game and risk 50 or so to do so. That's why you see them in games that are 90% or higher win rates for them.

[IcePack]

Sorry as others pointed out this discourages playing others in different class.
It's already hard enough to play games with my RL friends on CC, cuz they've sunk to privates cooks etc and have to limit the # of games I play with them cuz I'll lose to many points.

Make penalizing a higher player vs low ranker even more, and it'd be impossible to play with my friends

IcePack

[SirSebstar]

teach your friends to play better? or make new friends..
or maybe do not care so much about your rank... would that help at all???

also the OP, how is elo different from the now used system, and what do you think would be its chances.
If i.e. i was a freestyle farmer in oasis with quads on cooks, how would this new formula change things?

[mc05025]

I agree that the point system is really bad at this site.
It is encourage farming, encourage team games with teams of players with high difference at their rating, the top players uses tricks and have ridiculous high rating and so the competition is very low.

But the chess rating system is not good for cc because in chess all the games are 1vs1

[SirSebstar]

just a quick reminder, elo as the OP states should be

For team games, the total of the teammates' scores can be taken.

???
but thats already how it is..

[IcePack]

teach your friends to play better? or make new friends..
or maybe do not care so much about your rank... would that help at all???

also the OP, how is elo different from the now used system, and what do you think would be its chances.
If i.e. i was a freestyle farmer in oasis with quads on cooks, how would this new formula change things?

My friends play fine, we play 8 person standards so competition is always huge. I don't care as much about rank as some, but I enjoy feeling rewarded for hard work and play and don't want to be punished even more for playing friends or random people.

IcePack

[Evil Semp]

Sorry as others pointed out this discourages playing others in different class.
It's already hard enough to play games with my RL friends on CC, cuz they've sunk to privates cooks etc and have to limit the # of games I play with them cuz I'll lose to many points.

My friends play fine, we play 8 person standards so competition is always huge. I don't care as much about rank as some, but I enjoy feeling rewarded for hard work and play and don't want to be punished even more for playing friends or random people.

It sounds to me that you care more about points than you want other to think. The system doesn't punish you for playing with your friends, you punish yourself by not playing with your friends because you lose to many points.

[pancakemix]

also the OP, how is elo different from the now used system, and what do you think would be its chances.
If i.e. i was a freestyle farmer in oasis with quads on cooks, how would this new formula change things?

It's very hard to give an example here as the highest scores are numerically far beyond what is normal in Elo (the highest rating in Chess was in the 2800s). Bear with me here.

Let's say I played Blitzaholic 1v1. He would get 4-5 points for beating me under the current system. In the Elo system against me (going by our current scores, and again his score is astronomical in the Elo system) he would have an expected win value of .89 (89% win chance). So let's again say he wins. His win counts as a 1 for the calculation, so his score after beating me would be calculated like this:

(Blitz's current score)+(The K factor multiplier. Using the FIDE values, this would be 10)((Wins in the series. Since it's only one game, this equals 1)-(Total expected win value. Again, since it's just the one game, this is .89)=(Score after calculation)

Leading to:

5548+10(1-.89)= New score
5548+10(.11)= New score
5548+1.1=5549.1

So Blitz would only gain 1 point from beating me. As you can see, the curve is much steeper in this system. If totals are taken for both teams in a farming game, this 1 point outcome would still apply, meaning if the farming team won, they would have 1 point to divide between them. Compare to the 5-9 each they get now.

just a quick reminder, elo as the OP states should be

For team games, the total of the teammates' scores can be taken.

???
but thats already how it is..

I was attempting to differentiate between 1v1s and team games. Remember, Elo uses much lower numbers, so I felt this distinction was necessary, but I guess I didn't have to make it.

But the chess rating system is not good for cc because in chess all the games are 1vs1

I would say take the average of the ratings of all players in the game and calculate the winning score that way. There is an Elo calculator for multiplayer games, but I'm not entirely sure how that works.

[IcePack]

Sorry as others pointed out this discourages playing others in different class.
It's already hard enough to play games with my RL friends on CC, cuz they've sunk to privates cooks etc and have to limit the # of games I play with them cuz I'll lose to many points.

My friends play fine, we play 8 person standards so competition is always huge. I don't care as much about rank as some, but I enjoy feeling rewarded for hard work and play and don't want to be punished even more for playing friends or random people.

It sounds to me that you care more about points than you want other to think. The system doesn't punish you for playing with your friends, you punish yourself by not playing with your friends because you lose to many points.

I simply stated facts. 8 player standards (group of 8 friends) when ur a captain cuz u play premium and tonnes of games with others, but the group are privates and cooks makes it painful points wise. I still play them, and other games with lower ranks.

But I don't go playing 100 - 8 player standards with them either bcuz I'd end up handing points to others after spending time playing and winning my team and other games.

Everyone cares about them to a point, I'm saying I'm not obsessed with them but I definitely don't want any system where I'm going to be penalized MORE points then I already am for picking up casual / friendly games with lower ranked players & friends on a casual gaming site. This isn't a chess league or sports league, it's a casual gaming site made to encourage casual games with others.

The OP and everyone else posting here clearly care about points as well frying to give feedback on how points are achieved etc. I don't understand why anyone needs to long out I care about them? Clearly anyone spending time on this topic trying to debate any points system is going to care enough to make post(s)?

IcePack

[jrh_cardinal]

Everyone here is right, this will just discourage people who care a lot about their rank from playing low ranked players. Also, it will increase the (team) "farming" problem, this statement is incorrect:

If totals are taken for both teams in a farming game, this 1 point outcome would still apply, meaning if the farming team won, they would have 1 point to divide between them. Compare to the 5-9 each they get now.

the highest ranked team farmers play with low ranks on their team, so for a quads game against very average players:
5000 + 1000 + 1000 + 1000 = 8000
1500 + 1500 + 1500 + 1500 = 6000
not a huge margin with numbers so high. The team farmers may not keep as high of scores, but people who play singles games (5000 losing to 1500) lose a whole lot more, they'll just end up even farther ahead comparatively

[pancakemix]

Everyone here is right, this will just discourage people who care a lot about their rank from playing low ranked players. Also, it will increase the (team) "farming" problem, this statement is incorrect:

If totals are taken for both teams in a farming game, this 1 point outcome would still apply, meaning if the farming team won, they would have 1 point to divide between them. Compare to the 5-9 each they get now.

the highest ranked team farmers play with low ranks on their team, so for a quads game against very average players:
5000 + 1000 + 1000 + 1000 = 8000
1500 + 1500 + 1500 + 1500 = 6000
not a huge margin with numbers so high. The team farmers may not keep as high of scores, but people who play singles games (5000 losing to 1500) lose a whole lot more, they'll just end up even farther ahead comparatively

Not true. That's a misrepresentation of who is actually getting farmed and who is doing the farming. Take Game 8633027 Game 8633031 and Game 8636621 as examples (and yes I realize there's a cook on Blitz's team in the third one).

Just browse over Blitz's finished games. It's a sea of Clandemoniums against at best Corporal 1st Classes, minus the tourney games (which are to Blitz's credit) and the occasional game against officers who try to take down the Conqueror.

Secondly, you didn't take the formula into account with those numbers. 8000 vs. 6000 still only garners 1 point for the winning team. Even if you take the averages (2000 vs. 1500) the same result occurs.

As far as discouraging playing against lower ranks, if that's the end result then this doesn't have the casual atmosphere it claims to have. And the people who are that serious about rank? They're farming.

BTW, I'm using this site to get my figures. Plug them in for yourself if you wish.

[Darwins_Bane]

The problem with your proposed system, is that it only takes into account certain types of games offered on this site. How do you compare win percentages when it is a free for all game with 3 or more players? If someone plays mainly 6 or 8 man games, or basically any games where is isn't side 1 vs side 2 and thats it, your win percentage is immediately lower than someone who plays 1v1, even though your skill and points are about the same.

[pancakemix]

The problem with your proposed system, is that it only takes into account certain types of games offered on this site. How do you compare win percentages when it is a free for all game with 3 or more players? If someone plays mainly 6 or 8 man games, or basically any games where is isn't side 1 vs side 2 and thats it, your win percentage is immediately lower than someone who plays 1v1, even though your skill and points are about the same.

I would say take the average of the ratings of all losing players in the game and calculate the winning score that way. There is an Elo calculator for multiplayer games, but I'm not entirely sure how that works.

EDIT: see bold.

[pancakemix]

Oh, and you'd have to make adjustments to the K factor regarding the expected value as it's going to be much lower due to the larger number of players. Again, there is a multiplayer calculator which I have not used. I will check it out later.

[Darwins_Bane]

Ok, that works mostly, what about if a player who mainly plays, as an example, 1v1. If they have a high win % and join with a bunch of ppl with low win percentages, how does that work? I'm not saying the system shouldn't or doesn't need to change, I'm just trying to point out that a lot of thought went into the current point system to make it as fair as possible. Is it perfect, no. But I don't think this solution as is would work to fix the problems that are currently there.

[Metsfanmax]

I'm just trying to point out that a lot of thought went into the current point system to make it as fair as possible.

Is that actually true? I don't have a reason to distrust you, but it seems like (loser's score / winner's score) * 20 is not something that it took a lot of time to come up with. I would expect that a system based on a rigorous or thoughtful analysis of what the point system ought to be, would not have such an arbitrary coefficient and be based on such a simple formula.

[SirSebstar]

also the OP, how is elo different from the now used system, and what do you think would be its chances.
If i.e. i was a freestyle farmer in oasis with quads on cooks, how would this new formula change things?

It's very hard to give an example here as the highest scores are numerically far beyond what is normal in Elo (the highest rating in Chess was in the 2800s). Bear with me here.
.........[edited out example.............]

So Blitz would only gain 1 point from beating me. As you can see, the curve is much steeper in this system. If totals are taken for both teams in a farming game, this 1 point outcome would still apply, meaning if the farming team won, they would have 1 point to divide between them. Compare to the 5-9 each they get now.

Good example, Thank you for your time. I do get it now. I believe your basic premise would be to have an higher/ steeper curve..
mmm, i am neither against nor for it. Although i have to agree with the other writers that i do not see how this stops farming. Also, it brings the total amount of points down so the conqueror will have a much lower number and playing low rankers will be even more unprofitable for anybody but a farmer..
still good luck, at least i understand what you are trying to do, and to tell you the truth, i can see reason for change, but not the right yet just yet

[Qwert]

anyone say elo ranking?
Im implement some kind of these in clan ranking.

[SirSebstar]

anyone say elo ranking?
Im implement some kind of these in clan ranking.

rather then the ladder ranking? I concur, that would be good. Just make sure every team understands the calculations and can make them themselves

[pancakemix]

Ok, that works mostly, what about if a player who mainly plays, as an example, 1v1. If they have a high win % and join with a bunch of ppl with low win percentages, how does that work? I'm not saying the system shouldn't or doesn't need to change, I'm just trying to point out that a lot of thought went into the current point system to make it as fair as possible. Is it perfect, no. But I don't think this solution as is would work to fix the problems that are currently there.

I'm not sure what your example is. Are you saying a person with, say, .6 EV (which will henceforth mean expected value) from plating 1v1s joins a 5-player ffa (or similar game) against players with around .3 expected value? If so, the total of EVs in Elo is always equal to 1. The EVs would have to be adjusted to total 1, and to do this you could lower all expected values to a certain degree. So in our 5-player scenario we have:

.6
.3
.3
.3
.3

The total is 1.8, which is too high. We need to reduce them all by an equal ratio for fairness. Solving this is a simple equation:

1.8/n=1
(1/1.8 )x1.8/n=1x(1/1.8 )
n=1/1.8
n=.5555555...

Now we multiply each of our expected values by our reduction coefficient:

.6x.555555...=.3333333...
.3x.555555...=.1666666...
.1666666...x4=.6666666...*

.6666666...+.3333333...=1.

*- Four players have the value .166666..., so this part has been added to save time. Assuming they'd all be different, we'd actually figure them all out, but this saves time.

Actually, now thinking about it, you probably wouldn't have to adjust the K factor at all. K is determined by rank and experience anyway.

[Darwins_Bane]

I see two problems with the way that you do that. Firstly, how do you scale them all equally when one of the has a higher EV and therefore wins more, which means he should be less rewarded for the win. But at the same time, he may have a horrible record in 5-man and have much better records in 6 or 8 mans. Should he get less or more points if he wins then?

[pancakemix]

To that I'd say that games are what they are. Strategy doesn't change fundamentally between a 5-player and a 6-player. Obviously, your approach will be slightly different, but I'd think different maps would be a far greater factor in determining the chances of winning. Maybe I have a pretty good track record at 5-player Classic games, but that's very different from a 5-player World 2.1 game or a 5-player Doodle Earth game or a 5-Player Clandemonium game. It would make far more sense to adjust someone's EV based on the map instead of number of players.

But that's not practical. I'm not sure on the exact number of maps, but it's getting close to 200, I believe (If anyone has this figure, it'd be nice to know. Thanks). That's a lot of maps to determine an EV for based on each person. Especially when you consider the scoreboard gives over 20000 hits. I've never even played half these maps to begin with. How would you expect to calculate my EV for a map like Oasis that I've never played on?

There are other things I'd take into account before number of players, too: spoils type, fog, fortifications, and play order all have a big impact on strategy, arguably more so than player number because they effect the way you actually play the game rather than where and how you start out which could be just as bad even with less players and the way you play a game of Risk is just the way you play a game of Risk. It's still holding continents, trading in for troops, etc. I will agree, EVs should take more into account than win %, but I will say that:

1. Win % should be the bulk of EV determination
2. I'm not sure how much player numbers should be taken into account for EVs, but beyond the scaling I doubt it should be much if anything at all because player number's effect is really a function of what map you're playing on more than anything.