🌿ELO System Updates

[Stroom]

Elo Decay - I know players being afraid of losing their elo is a thing. I have invited a few ppl to games before and they have told me this. But........
I personally do not stop playing games to keep my elo. In stone age I am expert status (but ranked like 300th) and haven't played in probably a year. According to this I would lose over 50 elo every 3 months? I am currently a strong player in 20 games and have been a strong player in an additional 15. So if I don't participate in each of around 30+ games in a 3 month period I lose anywhere from 30-60 elo each 3 month period? If I am understanding correctly this is drastic and should be changed.

If you don't play, why would your elo even matter to you? If you start playing agian, I'm sure you would quickly get back to your old elo score... unless your skill is not the same it was... so losing 10% elo for not proving that you still have the skill is fine. All you have to do is keep playing... or just don't care about a pointless number.

Elo ranking is meant primarily for the active players. If someone climbs to the top and refuses to play anymore, he might have gotten that elo from previous kinds of systems where cheating elo was easier or where you gained more points faster so newer players just don't have a meaningful way to get that high. Plenty of Hanabi players exist like this.

Over time, the system would inflate the highest possible elo anyway. New players come, play a bit, lose elo to stronger players and leave. The players who keep playing, essentially leech elo from them and the pool of elo among active players increases. So it becomes more possible to get a higher elo. You can add a new player with 1000-1200 elo, he can play a bit and often stop playing at ~900 elo.

[Stroom]

I think ELO decay is defeating the purpose, that you said ELO does on this platform:

...

But decaying ELO is really against this idea of what ELO shows. It's also blurring the difference to the Arena-Ranking - which is already a ranking for recent events. The normal ELO doesn't need to depend on any recent events - it only needs to show skill level.

Another problem: When everyone starts at 0 ELO, you needed ways to feed points into the system. That's so far achieved by favorable K-factor for new players and the down-cap at 100. This way seemed quite balanced as it ensured that new players joining the community don't draw their points out of the community, dragging everyone down. So far so good - without ELO-decay. This system didn't need a mechanic to erase points from the system repeatedly and I think this is bound to drag everyone down - not just the players it's supposed to hit. People will hate to play with players whose ELO got artificially deflated.

I do realize that - as many other players - I've just jumped right at the change that I think is bad. So I'll also say this: I really appreciate most of the other changes you made - especially fixing the evaluation within teams!

The elo system is inflating the available points in the system over time anyway. Someone having 1000 elo 5 years ago vs having 1000 elo now would be different. Easier to achieve today. So there's no point comparing players by their elos. Similarly you don't compare a millionaire today to a millionare in the 80s. Having a system remove elo from the system is not necessarily a bad thing. I think it's OK to give it a chance.

As long as you keep playing, you would still be able to keep your elo. And with a few tweaks, it would be possible to keep the system healthy. Maybe you need to monitor what the active player's total elo pool is and to see if it within normal levels.

http://www.glicko.net/glicko.html - this system can theoretically work. The elo calculation itself is almost the same. But the score that is displayed, will decay over time. Essentially, the more inactive you are, the more uncertainty there is about your skill level. But if you play more, the uncertainty is smaller so your score is more closer to your elo score, therefore higher. And the elo change after the game depends both on your elo and your skill uncertainty. What decay actually is, is an automatic calibration of your skill level.

[Ranior]

Power leveling--I understand the concern about the people abusing the system, but I think this "fix" hurts more than it solves. There's a lot of people that just play a bunch of games in a row with friends sometimes, this "fix" isn't good for that. For lots of games with smaller player pools, if you play real time games around the same time most days, you wind up facing the same opponents fairly regularly. Especially for higher player count games. I do not like this change.

Actually, this is perfect for this purpose. Players shouldn't be basing their ratings on their performance against just their friends. The elo pool bases player skill relative to their opposition, playing a minimal number of players messes with that. Nothing wrong with playing a lot of games with friends, but that's more of a casual thing that can be done in unrated as well. If you were trying to gauge your skill level though, you should be playing a much more diverse field than that.

You didn’t address the portion of his post not regarding friends. I’ve played many consecutive games w/ ppl I don’t know. I played 8 straight matches against the back to back arena champ in dig up recently. Won 5. I don’t know this person. Why should these wins count less? Also how do you make friends here? By bumping into ppl you share common games with.

The change they've made in this regard is horrible. I like to play Saint Petersburg live. There's another person that is often on during the same hours I am. We're not directly friends, just sort of online friends as we run in to each other. There's not a ton of people playing 4p St Pete games, so I play them fairly often (although still sometimes don't).

As of today, I am now heavily encouraged to avoid playing them. I cannot gain rating from beating them, but there's not that many other games to find. How has this improved the rating system--made it so my skill level is better reflected by elo? We both play a lot, play plenty of different opponents in those games, and don't always play each other....yet if I choose to play at their table with two others, I'm at a sizeable disadvantage since beating them is meaningless to elos eyes.

This change therefore does not achieve a good result. There are lots of players that in no way are colluding or causing an issue, yet now are encouarged to not play each other. It will hurt smaller communities and player pools.

I get that it's annoying some people abuse the rating system to boost their elo, but this change is not how to combat that. It does not produce positive outcomes or effects in aggregate, far more honest players are being hurt than the few abusers of the system were causing issues.

[Lumin_S]

While I'm not excited about Elo decay as a solution (please consider TrueSkill-type rating systems ) that is not a big deal compared to a possible change to red thumb effectiveness.

As others have already said, there NEEDS to be a way that we can prevent being matched with quitters, or suspicious new accounts which are likely made by past quitters who are likely to quit again. The Arena matchmaking has a tragically low Karma threshold already, so if there are no changes to that while we can no longer prevent playing against accounts which are likely to quit before completion (or bad mannered trolls), I would play MUCH less often.

Over a year ago I made these suggestions to improve the Arena matchmaking experience, and they have a good number of votes. I hope Admins are reading this:
Higher Karma penalties - https://boardgamearena.com/bug?id=66050
Raising Arena Karma threshold - https://boardgamearena.com/bug?id=66049

I get that there is a problem with high rated players red thumbing their competition. But please don't make the mistake of fixing that problem while greatly exacerbating other existing problems.

[orgle]

RE: Hanabi

Please return visible ELO to Hanabi games. Your changes, while good intentioned, are actually harmful to the spirit of co-operation. Players use ELO to understand which clues are appropriate for the current table. There is a large gulf in knowledge and experience between a Master with 700 ELO and one with 900+ ELO. I wouldn't expect a Master with 700-750 ELO to understand layered or clandestine, but I certainly expect a Master with 900+ to be comfortable with all complex conventions & nuance.

Also, some high ranking Masters prefer to play with 1-2 lower ranked Masters in order to help coach (and get better ELO if successful). Others only wish to play with other high ranking Masters (regardless of the paltry ELO result) because they enjoy the challenge of complex, efficient clues with other high ranked players. These same players get frustrated when their beautiful 7 card layered clue is destroyed by a player who simply doesn't have the experience & knowledge to understand the convention. This leads to a negative experience for all players, which could likely have been avoided by simply knowing the player only had 703 ELO.

Alternatively, you could help us better understand each player's experience by showing the number of games each player has completed successfully with the expected table player count.

Either way, please separate Master into 2 ranks 700-899 and 900+ so we can have a better sense of the experience level of the players at the table. Previously, expectations were much higher for players with 900+ ELO, but now we can't see what level of "Master" a player is. The result is that we have to a) dumb down our clues, b) increase our expectations for all Masters, or c) only play with the players we already know can handle our prefered level of play. These are all very bad options.

While I'm here, it would be very helpful to have a distinction between 2 player and 3+ player games. The conventions used in 3+ player game are often diametrically opposed to those used in 2p games (eg. early saves). A 2p-only Master would be completely lost in a high level 5p Master game & vice versa. Separate ELO ranking for 2p games would amazing. For instance, I'm a Master who generally plays 5p games, but I would probably be a Novice in a 2p Hanabi system, and that accurately reflects my understanding of & experience with 2p conventions. I shouldn't be playing 2p games with 2p Masters, but I'm often restricted from lower ranked 2p games because of my experience with 5p Hanabi. This makes it incredibly challenging to practice 2p Hanabi, as 2p Masters obviously have much higher expectations due to my Master rank.

Oh, and it would be great if the "Presentation" & "Restriction" fields remembered your settings, so you don't have to keep typing in the expected conventions when creating games.

In conclusion, please

1. Return visible ELO to Hanabi
2. Alternatively, show number of successful games played with the expected table player count
3. Separate Master Hanabi ranks into 2 ranks at 900+ ELO
4. Calculate ELO for 2 player Hanabi separately from 3+ player Hanabi
5. Save "Presentation" & "Restriction" fields

Thank you for such a great site, and for listening to player feedback.

[JackHH]

No more Elo ratings for Hardback. Sad times.

When the game had competitive mode enabled there were problems with cheating, however, the decision to remove it from arena over a year ago successfully got rid of the cheats and I had been enjoying the game much more ever since, having general fun playing, but also taking a little bit of pride in my attempt to get my ELO over 1000. I believe the step to hide ELO completely is a step too far, and takes away my motivation to use bga

I am not sure why taking the Elo rating away from Hardback really offers any benefit based off of my experience. Having obscured ratings hidden behind a wide ranging rank is just a really frustrating way to hide from users how close they are from levelling up.

I presume the decision was taken at a blanket level based on the game classifications, so I would like to ask if its possible to change this on a game by game basis, and if so, to reinstate Hardback please

Some comments regarding overall BGA rating system and this update:

ELO ranking allows to classify players in 7 distinct skills levels: Beginner (0 point), Apprentice (1-99 points), Average (100-199), Good (200-299), Strong (300-499), Expert (500-699), Master (700+).

How about adding more levels?
This is one of the most upvoted suggestions: https://boardgamearena.com/bug?id=13021

Seconded! For some low luck games, there can be players in the 900s. This is deserving of disctinion from `Master` IMO. Yes I am aware I sound like a bit of a tool here requesting my own distinction for hardback haha

[sprockitz]

Team Play:
Good change! It never made sense to consider it a tie with your teammate.
Cooperative Play/Non-Competitive
Why? Elo already gives no benefit, nice to have it accurate so you can better understand skill of players in game. The Crew games already didn't have real Elo, Elo was just games played.
Power Leveling
If winner is just reduced this is terrible, if it is K-factor then it is okay...but would still be better to take the player pool into account. Games with huge player pools could see a significant reduction in K-factor, games with tiny player pools this makes little sense. Also how consecutive is defined makes a huge difference. If it is as stated you just jump between 2 different players if that resets things and avoid the reduction. There are games with very small player bases where only 3-4 people are actively playing arena games.
ELO Decay
This is a very bad idea. It will lead to significant Elo deflation. Elo is meant to measure skill not activity. Arena measures active skill and resets every 3 months, doing the same thing to Elo is nonsensical...the meaning of Elo will be lost much moreso than it was when changing it last time with the 0-100 weirdness. And the scale is just silly, 10% is so much more than playing and coming last in a game. If I skip 3 months of playing a long game like Terra Mystica I would drop 70 points if I am at 700 Elo...or I play a 2 player game and lose to someone and lose 10 points instead, completely unequal. I have 72 games ranked strong or above, and don't always play all of them but that doesn't mean I've lost my ability to do so.
Red Thumbs
This is unfortunate too, would be better to restrict red thumb usage. The point of games is to have fun and matching wits with others, forcing you to play someone who is abusive or a known quitter or cheater takes away from the enjoyment of the game.

[ChiefPointThief]

Elo Decay - I know players being afraid of losing their elo is a thing. I have invited a few ppl to games before and they have told me this. But........
I personally do not stop playing games to keep my elo. In stone age I am expert status (but ranked like 300th) and haven't played in probably a year. According to this I would lose over 50 elo every 3 months? I am currently a strong player in 20 games and have been a strong player in an additional 15. So if I don't participate in each of around 30+ games in a 3 month period I lose anywhere from 30-60 elo each 3 month period? If I am understanding correctly this is drastic and should be changed.

If you don't play, why would your elo even matter to you? If you start playing agian, I'm sure you would quickly get back to your old elo score... unless your skill is not the same it was... so losing 10% elo for not proving that you still have the skill is fine. All you have to do is keep playing... or just don't care about a pointless number.

Elo ranking is meant primarily for the active players. If someone climbs to the top and refuses to play anymore, he might have gotten that elo from previous kinds of systems where cheating elo was easier or where you gained more points faster so newer players just don't have a meaningful way to get that high. Plenty of Hanabi players exist like this.

Over time, the system would inflate the highest possible elo anyway. New players come, play a bit, lose elo to stronger players and leave. The players who keep playing, essentially leech elo from them and the pool of elo among active players increases. So it becomes more possible to get a higher elo. You can add a new player with 1000-1200 elo, he can play a bit and often stop playing at ~900 elo.

If I don't play a game for awhile that doesn't mean it doesn't matter. I play tons of games. Bga is also constantly pumping out new games some of which I try (and like someone mentioned there is life as well) so some games will just be on the back burner from time to time. I don't lose my ability to play a game though just because I haven't played it in 3 months. You basically only play one game here so it would be no problem for you to keep your status. The purpose of the elo system is to rate how strong of a player you are. You believe the real master level for hanabi is 900+. With this current change if a 900 elo hanabi player didnt't play for 1yr their elo would drop to 590. This player has the exact same skill they had prior but now you wouldn't even allow this player to sit at a table with you. I just played a few games I haven't played in months this week. Just as good as ever. I don't want to feel like I have to play this game by this date or lose this amount of elo.

[ChiefPointThief]

→ ELO Decay

I wrote a post about this 3 years ago: viewtopic.php?t=18601
But, now I think Arena mode also needs ELO decay as parking frequently happens even in Arena mode.
Also, playing 1 or 2 games periodically should not either stop or affect rating decay, for obvious reasons.

I just wouldn't participate in arena until late in the season. That is for games with a luck factor at least. For those games it is about who had the best streak at the end of the run and not the totality so I can't blame people who play early and park. With an elo decay there is no benefit to jumping out to an early lead because losses are inevitable and the math doesn't balance out. Being able to filter opponents by elo in arena would be good.

[Cheery Dog]

Why has Decrypto received the non competitive game treatment?
That's a team versus team game.