Please update ELO tables for Hanabi

[ElinaGranger]

While I can't understand completely the suggested table, I would like to show my support and say I wholeheardedly agree that Elo tables need to change.

The current system seems unfair to me in many cases. There are games where you get a perfect score and your Elo stays the same, or even decreases! This needs a change.
If we want to play with an inexperienced player and help them learn, we get punished severely Elo-wise.
I would even suggest that getting 23/25 - 28/30 as a score should not decrease our Elo.

[Jellby]

The current system seems unfair to me in many cases. There are games where you get a perfect score and your Elo stays the same, or even decreases!

I would accept that it should not decrease after a perfect score. But I don't think it should always increase. If you have a high elo and play an easy setting, it makes a whole lot of sense that you elo does not increase even if you get perfect score.

[Travis Hall]

I would accept that it should not decrease after a perfect score. But I don't think it should always increase. If you have a high elo and play an easy setting, it makes a whole lot of sense that you elo does not increase even if you get perfect score.

I don’t recall whether a perfect score is considered a win against the highest value on the ELO table, or a draw. If a win, then a perfect score will never decrease the team members’ ratings, but if the team’s average rating is high enough the increase might be small enough to round down to 0. I’m not sure that anybody has a rating high enough to do that, but maybe someone does (especially if playing with easy settings). Remember that ratings are calculated and stored with two decimal places, so there can be increases that are two small to see in some of the few places we can still see Hanabi ratings.

I would suggest that if the system is revised, rating change calculations should be done as wins and losses, never draws, to ensure that perfect scores never decrease ratings. In effect, “bot” ratings should be on half points; the top bot ratings is for a score of 29.5, not 30, so that the team either beats the bot, or is beaten by the bot, and never draws with the bot.

[Travis Hall]

If 23/28 does not decrease ratings, no matter how high the rating, we would definitely be in the region of ever-increasing ratings. Ratings would then become useless as a measure of skill, which means we cannot meaningfully restrict entry to our games by skill level.

[Jellby]

I don’t recall whether a perfect score is considered a win against the highest value on the ELO table, or a draw.

Last time I checked, it was always a draw. I don't recall specifically losing elo after a perfect win, but the elo change was consistent with a draw calculation.

[Travis Hall]

I don’t recall whether a perfect score is considered a win against the highest value on the ELO table, or a draw.

Last time I checked, it was always a draw. I don't recall specifically losing elo after a perfect win, but the elo change was consistent with a draw calculation.

That’s a problem. If a team of 4 players with an average rating of over 950 play together, their ratings will decrease no matter what. That shouldn’t be so.

[ufm]

I would accept that it should not decrease after a perfect score. But I don't think it should always increase. If you have a high elo and play an easy setting, it makes a whole lot of sense that you elo does not increase even if you get perfect score.

I don’t recall whether a perfect score is considered a win against the highest value on the ELO table, or a draw. If a win, then a perfect score will never decrease the team members’ ratings, but if the team’s average rating is high enough the increase might be small enough to round down to 0. I’m not sure that anybody has a rating high enough to do that, but maybe someone does (especially if playing with easy settings). Remember that ratings are calculated and stored with two decimal places, so there can be increases that are two small to see in some of the few places we can still see Hanabi ratings.

I would suggest that if the system is revised, rating change calculations should be done as wins and losses, never draws, to ensure that perfect scores never decrease ratings. In effect, “bot” ratings should be on half points; the top bot ratings is for a score of 29.5, not 30, so that the team either beats the bot, or is beaten by the bot, and never draws with the bot.

Unfortunately, it doesn't work like that.
Also, the system upgrade won't happen: the artificial ELO simulation itself fell out of favor and is now disused except Hanabi and Pandemic.

[Wreckage]

...the artificial ELO simulation itself fell out of favor and is now disused except Hanabi and Pandemic.

That's unfortunate. Because the Hanabi ELO system, even broken, is the best on the site for co-op games. I know there is an argument that they get fewer complaints now. My answer is of course, fewer people are going to talk about the ELO systems that are completely non-interesting.

The current model is pretty good, it just needs re-adjusted for flams, bp, and the bonus/penalty for player count has always been wrong.

[Plants2024]

While I can't understand completely the suggested table, I would like to show my support and say I wholeheardedly agree that Elo tables need to change.

The current system seems unfair to me in many cases. There are games where you get a perfect score and your Elo stays the same, or even decreases! This needs a change.
If we want to play with an inexperienced player and help them learn, we get punished severely Elo-wise.
I would even suggest that getting 23/25 - 28/30 as a score should not decrease our Elo.

Have you tried playing a friendly match?