New ELO formula

[acocle]

I fully agree with Yaron but would insist that whatever the system chosen, you re-apply it historically (by the way, I don't think this has been done after the system of ELO loss for not playing enough was removed...)

----------------

Completement d'accord avec Yaron mais je voudrais insister sur le fait que quelque soit le systeme choisi, vous l'appliquiez sur tout l'historique (au passage, je ne pense pas que vous l'ayez fait apres que le systeme de perte de points ELO pour ne pas jouer assez ait ete supprime...)

[zfm]

sourisdudesert: agree to both higher K and adjusting the K value to a higher number for > 2 players.

Anyway, I think you should change this K to a higher number when more players playing.

Let's say, for players that have played more than 30 times, K = 30 on 2 players, K = 45 on 3 players, K = 55 on 4 players, K = 62 on 5 players, K = 68 on 6 players, and so on... I don't add this linearly, because we want smaller "reward" or "punishment" for more players, but I want K to be higher because without that, playing 7-player saboteur will give you only 1-2 points probably...

Anyway, great idea and great way listening to our suggestions!

[sourisdudesert]

@zfm: with a 7-player saboteur game the winner will get the same amount of points than in a 3-player game, and not just 1-2 points. The bonus for "more player" idea is to encourage players to play >2-player games because it's more fun for most of the games.

[yaron]

@sourisdudesert:
1 point is certainly better than 3, but 0 is best.

To be honest, I'm not sure which problem this rule is trying to fix.

Have there been many complaints from Masters about not getting points for defeating complete beginners? I imagine most strong players would not be happy getting their points in this way. However, the ones that do, will do it all the time, and get to the top without ever playing a meaningful game...

Is the problem that Masters are unwilling to play against beginners because of the risk? In that case, I suggest having an option for making a game unranked, so that it doesn't affect ELO at all, no matter what the result is. That way, I can play a "teaching" game against a beginner without fear. I think that's better than putting features into the ELO system itself that are easy to abuse.

Thanks for listening to our feedback!
Yaron

[cf_skeeve]

I have a couple of concerns with the new system. I am concerned about this changing a couple of things about the tone of the gameplay here. First since the leaver penalty is now the same as the lose penalty (if I am understanding correctly), in a competitive game you are now incentivized to leave rather than give your opponent ELO [this is especially crucial when two top ten players are playing one another]. Also the new "chess-like" ratings assume a facility with the game before playing the first game for ELO. Essentially, nobody plays chess the first time for ELO, but on this site many players play to learn a new game, if this is not taken into account the games where they are learning (and will almost certainly lose) are a more significant factor in their ELO than games they play after learning the game well. One of my favorite things about this site is you can play a few learning games and then start competing with only a modest penalty that you can quickly overcome. I am not opposed to a point rescaling so that players ELOs are not swinging so much, but there should be some sort of learning environment where one is not penalized for learning the games online (this would be a problem with retroactively changing ELOs as players would not take this into account even if in the future some sort of non-ranked, learning environment was implemented).

[sourisdudesert]

Hello,

For your information ELO scores has been recalculated from the start for the following games:

• Troyes
• Uncle Chestnut's table Gype

(Method used: K=60 for 30 first games, then K=30. K +50% for games with more than 2 players).

Before going further, I would like to have your feedback.

Cheers,

[tchobello]

I don't really know how the points are calculated in multiplayer games here.

I'm used to play Ticket to Ride & when you win a 4-player game, it considers you have won 3 games, one against each player.

[sourisdudesert]

Hi all,

We received a lot of feedback here, and change our minds a lot about the rating system.

Here's are a new proposal for the ELO rating system on BGA (let's hope it's the latest ):

• K=60 for the first 30 games (ie: +30/-30 pts for 2 players with the same level).
• K=40 afterwards (ie: +20/-20 pts for 2 players with the same level).
• When leaving a game, you loose 10pts + the number of points you would lost if you'd loose the game (last place).
• We would like to encourage everyone to play games with the "advised" number of players... because it's more fun to play these games like this. For this reason, starting from the 3rd player, each additional player on the table will increase the K factor by 50% until the "ideal number of player" has been reached.

Don't hesitate to post your feedback about this new proposal.

Cheers,

[zfm]

Hello,

For your information ELO scores has been recalculated from the start for the following games:

• Troyes
• Uncle Chestnut's table Gype

(Method used: K=60 for 30 first games, then K=30. K +50% for games with more than 2 players).

Before going further, I would like to have your feedback.

Cheers,

Since I played this both games very rarely, I cannot comment a lot... Maybe better I save the data of myself before the recalculating and if the recalculation is done, I know how much the difference is (also with my rank and so).

[latexrhombus]

As far as 2-player ratings are concerned, as long as the system is consistent (new or old, or some new hybrid of the future) it makes no difference to me.

However, when 3 or more players are involved in a game, I think the system needs a big change. I've seen someone tied for 2nd in a 7 player Saboteur game and receive -2 (yes, negative 2 points) while players ranked below him have gained points. Along similar lines, in another Saboteur game, I noticed that two players tied for 3rd and the HIGHER ranked player received +2, while the LOWER ranked player received +0.

I realize I only referenced Saboteur, but this inconsistency seems to apply to games involving more than 2 players. While I agree that winning a game with 7 players should not count as 6 independent wins for scoring purposes, there does need to be a change, because as is, something is off with the 3+ player rating system. It does not seem to fairly award placing well in big games (at least in my experience and in the experiences of other people I have talked to) and seems to make it impossible to actually increase one's Saboteur rating beyond a certain level even with consistent high placements.