ELO SYSTEM

[salseba]

Actually i have 2376 elo at Caylus.

If i win some expert (1800+ elo) i win only 2 elo points.
If i play someone with 1300 elo and win i still win 1 elo (which is ridiculous imho, it should be 0).

So, if my goal is to be top ranked what is the point taking risks playing good players when i can play super bad players and win almost same elo ?
The current elo system is weak and encourage people to play weaker players and run away when someone stronger wants to play with you.

Completetely agree with the statement.

I sometimes use the replay funtion to watch games between masters to improve my game. Actually the number one ranked player in Tzolkin is hiddendraggon. I had selected his profile and searched games with strong opponents (one of the eight top players). No games found (except one no ELO game) but also (it seems) no games at all against expert or even good players....!?
It seems that this player has built his number one position only by winning one ELO pt each time and again versus poor opponents (under 1500 ELO for the majority). In a game like Tzolkin when you play 2 players mode if you ar strong and your opponent is weak you are sure to win as he probably don't go for a ressource-building strategy, even probably let you the uxm012 first move if he is first player and even if he plays a ressource-building strategy, he probably wouldn't manage to win with it (but a 1700 elo player could have beat a master).

To make a comparison,ELO ranking is as bad as if in tennis Federer could have become number one ATP ranking player never playing nadal, djokovic or others strong players, never winning one slam, but only by winning ATP challengers againt over 200 ranking players.

[Doodads]

Yes this guy (hiddendragon) is doing exactly the same at caylus; he plays only very bad players with sometimes 1300 elo.

Please BGA fix the elo system, lot of people are playing for stats and rank, the game level is poor, it would be far more interesting, even for them, to have some elo system that encourage people to grow up and improve.

[nautilusmp]

Well, selective pairing (i.e choosing opponents that are likely to give optimal rating benefits) had always been a problem in games using Elo rating system. Perhaps we should change the K-factor (note this is not my original suggestion). Your K-factor should be less when you win against a new entrant (i.e. those who have played 30 or less games). In fact, I do think the "usual" K-factor of 30 is a bit too large already. That results in the possibility of losing/gaining plenty of points within a few games, and that the ratings would only reflect the results from the past few games. The recommended value of the K-factor will be always subject to debate and will require a bit of a research, so I wouldn't make suggestions as to what K-factor should be used.

Also, we could introduce a "rating ceiling" for each player, which is the maximum possible rating a player could attain at each particular time. The rating ceiling for each player would be 400+(the highest rating of a player (s)he had beaten in the past 60 days), or equal to the highest rating of a player (s)he had drawn in the past 60 days, whichever is larger. We would need to discuss whether this should apply to one's actual rating or only to the "displayed" rating, whether there are better caps (I throw it around simply because the number 400 often comes out in Elo rating calculations), and how to combat the potential problem of ratings deflation.

These are just ideas, be it from me or someone else, that I put forward, and I am open to discussions.

[Maranos]

To me the problem seems to be the 600-points rule: When two players have an Elo-difference of more than 600 points, the system computes the point gain/loss as if the difference was 600 points. As a result of this rule you will always win at least 1.2 Elo points no matter whether your opponent is 600 points below you or 1000 points. Without this rule winning against someone with 1000 points less would give only 0.13 points. Players would need 10 times as many games to boost their Elo by playing such weak players.

Additionally, new players have a K-factor of 60 (compared to the usual 40). I might be wrong, but I believe that in a game with a new player, the higher K-factor is applied to all players. This means that beating a new player 600 or more Elo points below you will actually give you 1.8 points, making the strategy 50% more effective.

I don't know the reason why the 600-points rule exists. A similar rule (400 points rule) exists in Chess, but a brief web search didn't come up with a good reason for that rule either (on the contrary: http://en.chessbase.com/post/the-elo-ra ... ncy-tables )


Here are the formulas, btw: http://en.doc.boardgamearena.com/Rating

[qwertasdfg]

The higher factor is only applied to the new players

[Doodads]

Could it be possible to have some staff feedback ?

Is the elo system going to be fixed ?

BigFake didn't play caylus for long time and was out of classment then.
He played a no elo game and lost but got back into top rank players.
In my opinion, it seems obvious that games with no elo should not allow you to get back in elo classment.

Some players like hiddendragon are playing bad players only to have safe win and point by point be top ranked.
He's almost 2000 elo and fight people with 1400 elo lol.

Does this way of "playing" reflects what bga want to communicate ?

[jdietz]

I am a good player in Race for the Galaxy. I want to be one of the best players.
I feel the ELO system is broken, because I often beat expert players, maybe 40% of the time. I beat master players 20% of the time - and I should almost never beat them because they have at least 300 more ELO points than me. There is no way I should be winning that often against them. The ELO rating of expert players is too high, to say nothing of master players.
Race for the Galaxy is too random a game for anyone to have an ELO above expert.

I propose a system where expert and higher players lose ELO points every (day) they don't play at least (X) games. And for master players, they must play even more games than experts to keep their master rating. Maybe it could be tied into number of tables everyone plays that week. Since there's at least 1000 tables of Race for the Galaxy a week, expert and master players can be expected to play more tables per week than expert players of a less popular game (or just a game that takes longer to play). The time period, number of games required, and ELO points lost can be tweaked to BGA's liking. I need not say that games played not for ELO won't count toward the quota.

Make good players prove they are the best. If they are unwilling, they're not really best and don't deserve their player ratings.

Muse is the top ranked player in Race for the Galaxy and hasn't played a game for ELO in over 300 days. This player doesn't deserve this ranking, because they are unwilling (or possibly unable) to play other BGA users for ELO.

The ELO system should be designed to reward competing players, not players who don't compete.

[Liallan]

I propose a system where expert and higher players lose ELO points every (day) they don't play at least (X) games. And for master players, they must play even more games than experts to keep their master rating. Maybe it could be tied into number of tables everyone plays that week. Since there's at least 1000 tables of Race for the Galaxy a week, expert and master players can be expected to play more tables per week than expert players of a less popular game (or just a game that takes longer to play). The time period, number of games required, and ELO points lost can be tweaked to BGA's liking. I need not say that games played not for ELO won't count toward the quota.

This doesn't work all that fabulously for people doing turn-based games cause it takes too long to finish them. Not to mention that just because someone is really good at a game doesn't mean they have to play it every day or even every week. If you have nothing better to do and can play your games every day, great, but not everyone is like that. And people don't necessarily need to play that often in order to remain good at it. I go to another site where I lose skill points (whatever they call them there) for every day I don't play. I hate it. I don't have time to be on there every day, nor to play every game I know every single day. I think it's ludicrous. The only redeeming value is that once I do play, I start back where I was before that decline.

Muse is the top ranked player in Race for the Galaxy and hasn't played a game for ELO in over 300 days. This player doesn't deserve this ranking, because they are unwilling (or possibly unable) to play other BGA users for ELO.

I don't know this game, nor that player, so I'm just asking... but why have you decided this person is "unwilling" to play this? Sometimes people get burnt out on things.

[Doodads]

hiddendragon is almost in Caylus top 10 players, almost "master" and he never played any master or expert or even good player.
He uses to play 1400 elo players.

For example, his last games :


http://fr.boardgamearena.com/#!table?table=12160704 1977 1496
http://fr.boardgamearena.com/#!table?table=12163800 1979 1441
http://fr.boardgamearena.com/#!table?table=12170951 1981 1436
http://fr.boardgamearena.com/#!table?table=12194843 1982 1394
http://fr.boardgamearena.com/#!table?table=12211063 1984 1469
http://fr.boardgamearena.com/#!table?table=12212234 1985 1415
http://fr.boardgamearena.com/#!table?table=12226472 1987 1378
http://fr.boardgamearena.com/#!table?table=12242933 1988 1460
http://fr.boardgamearena.com/#!table?table=12284166 1990 1472
http://fr.boardgamearena.com/#!table?table=12321626 1992 1299
http://fr.boardgamearena.com/#!table?table=12349609 1995 1485 1473
http://fr.boardgamearena.com/#!table?table=12356674 1996 1411
http://fr.boardgamearena.com/#!table?table=12360158 1998 1443

Is it good idea to allow people with 1997 elo win 1 elo when beating a 1444 (!) elo guy ?
In my opinion it should be 0. Then no fake masters or experts in top 10.

Should players like him be top ranked ?

There are a lot of fake players like hiddendragon on bga, all abusing this unbalanced elo system.

I don t get the points rewarding cowards and bad players.
Please someone explain to me.

[jdietz]

I don't know this game, nor that player, so I'm just asking... but why have you decided this person is "unwilling" to play this? Sometimes people get burnt out on things.

Give other people the chance to be top ranked by de-ranking players who won't or don't play (regardless of reason).

You could make it weekly or monthly instead of daily. Any system would be better.
You see master level players always play for no ELO. Why? They are afraid of losing. They have a high chance to win, but it's not 100%, so no ELO. You have to encourage them to play for ELO. Otherwise, they never will.
You're right. Such a system is harsh on turn-based players.