Hanabi bots' ELO

[b3nsmall]

I really hope these aren't the final ELO numbers. Is no surprise everyone is playing the easy 6 colour version and mostly 3 or 4 players. The system discourages striving for perfect games, which is the opposite of the point of hanabi. This currently encourages to be content with just a reasonable score, and make sure you don't do anything fancy that might lose 3 lives. Way more fun to play Hanabi on Keldon at the moment.

All the 5 player scores should have higher ELO's, same with all the perfect scores, and all the multi scores too.

[qwertasdfg]

yeah, the elo value for multicolour 2-player is definitely not as high as you stated. played with 1947/1992 elo. got a perfect score and the elo change was +1.

Maybe just copy-pasted a wrong number for the bot?

[malachi]

I agree with b3nsmall that the current implementation is counterproductive for players who want to achieve a perfect score. I personally would rather strike out 50% of the time and get a perfect score 25% of the time than get 24 every game. Hanabi require reaching to get those last few points, which means sometimes it all falls apart. Rewarding conservative play results in less interesting games.

[sparr]

Playing for 25-or-bust is an official variant in the rulebook I have. Notably, that variant allows you to keep playing until you run out of cards.

If you're stopping one round after exhausting the deck, then by the rules you're playing for points, not win-lose, in which case 24 is ~96% as good as 25, and should be handled as such.

[malachi]

I think I read that variant differently than you do. It doesn't say (paraphrased), "You can play for 25 or bust, but if you do you get to play cards until you run out." It says, "You can play cards until you run out, but if you do you should only count a perfect score as a victory."

In my mind, the objective is to do as well as you possibly can, knowing that there is risk. Due to the difficulty of making 25, I think it should be valued more highly and that strikeouts should, perhaps, be less harshly penalized.

[sparr]

I don't see a difference in the two wordings you just used, as applied to this situation.

In the base game, you stop playing cards at a certain point, and you get a score from 0 to 25.

In the variant game, you play cards until you run out, and you get a score of 0 or 25.

[malachi]

To me it suggests that one was added as a result of the other.

To me, one phrasing means, "You can make the game easier with a variant where you play out all the cards, but if you do you should only really count a perfect score becasue it lowers the bar quite a bit." The other suggests, "You can play an all-or-nothing variant, but if you do you should play out all the cards to make it a little bit easier to get a perfect score by taking some of the deck risk out of the game."

It does not really matter, though. My original point is not that the game should be played as all-or-nothing, but that the rating system currently does not seem to encourage players to play for the best score they can because of the penalty for getting three strikes. I think you can get a pretty good rating by playing a lot of really conservative games with decent but not great scores.

[Lit]

first of all observation:
- 2p games are much harder than 3 4 or 5p games , but ELO table consider them as easiest.
- maintain good rating by playing 6 colour mode "tricky" is easier then by playing 5 colour mode "normal".

Maybe for each score , in each of modes , "an opponent" should be created. So , when bunch of players score
22 / 25 for example , they match in powers correspondenting opponent . Their rating changes AND
opponent's rating changes.
4 players with average rating 1800 scored 22/25 (1740). Outcome:
- player's rating go down.
- "22/25" opponents rating go up.

[ollyfish2002]

Hi
First, I would like to say that using the mean of the players is not fair.
I have played level "very difficult" with masters, end at 28 and we are losing ELO points, because of the means. I have lower ELO than masters but I manage to play with them and end at 28.
Can it be another way to count ?
IMHO, using the lowest ELO would increase masters playing with beginners, and teaching them how to play. it is a cooperative game.
Second, I think a ELO bonus should be given to games ending with no bombs. We cannot do perfect games in difficult decks, but no bombs means a good cooperation. So players in thes games should be rewarded.
Third, I don't know if ELO is active in "non-random" deck. But these decks allow perfect games. Playing with the same players with the same ordered deck can be easy with order of cards draw on a paper.
Just my 2 pennies.
Bye

[Zoorp]

Second, I think a ELO bonus should be given to games ending with no bombs. We cannot do perfect games in difficult decks, but no bombs means a good cooperation.

I don't care about who earn ELO or no. But I answer to this particular point who seems me wrong.

No bomb is maybe good, but dosen't necessarily mean good cooperation. It could mean no risk taken. And it's easy. It often means no great end result, but it's easy to do. Does an easy thing to perform earn a reward? I have no opinion on this (like I said I don't give a shit about what is fair or unfair with elo.) I just talk about facts.