ELO calculation change

[dschingis27]

Seems this issue is still ongoing: https://boardgamearena.com/bug?id=29568

Thanks, ufm, for posting this!

So, to conclude, the factors are:
(1) N / (2*(N-1)) = (1/2) x [N/(N-1)] , if N is equal or less than recommended number of players
(2) (recommeded N)/(2 * (N-1)), if N is more than recommended number of players
Source: https://boardgamearena.com/bug?id=29568 and viewtopic.php?p=35177#p35177

Note that (1) and (2) are equal if N is equal to recommended N. It seems to check out. For 7Wonders with recommended player number 4, the formula (2) from above simplifies to 2/(N-1), the one that I supposed earlier.

I checked recent 6Nimmt! games with 7, 8, and 10 players. For all of them, the factors follow the formula (1) from above. Maybe they have set the recommended number of players for 6Nimmt! to 10 at some point. In any case, the factors sourced from my link posted earlier are wrong as of today:

So, finally, I found the very relevant thread again.
https://boardgamearena.com/forum/viewtopic.php?p=60122

So, like it or not, the recommended player number does play a role and explains the different calculations between games. How to know what is recommended player number? Sometimes it's written during table creation, but I think it's usually not documented.

[ufm]

I checked recent 6Nimmt! games with 7, 8, and 10 players. For all of them, the factors follow the formula (1) from above. Maybe they have set the recommended number of players for 6Nimmt! to 10 at some point.

I can answer that as well. Currently, recommended number of players doesn't exist in 6 nimmt!:

// Suggest players to play with this number of players. Must be null if there is no such advice, or if there is only one possible player configuration.
// NB: the automatic lobby will try first the lowest number of players if this is not specified. So you _have to_ specify this parameter if the lowest player number is not compatible with the default options.
'suggest_player_number' => null,

// Discourage players to play with these numbers of players. Must be null if there is no such advice.
'not_recommend_player_number' => [2, 3, 4],

So, like it or not, the recommended player number does play a role and explains the different calculations between games. How to know what is recommended player number? Sometimes it's written during table creation, but I think it's usually not documented.

If possible, checking the source code gives you the exact answer.

[Jellby]

And what if there is no recommended number of players? Is it 0 or infinity?

[dschingis27]

And what if there is no recommended number of players? Is it 0 or infinity?

Funny enough, it's kind of both "null" and infinity. (actually it is infinity)

In the code, if there is no recommended number, then 'suggest_player_number' => null which means "nothing specified" which in this case is treated like infinity. Because there is no recommended number specified, the calculation will always stick to formula (1), and never "jump" to formula (2).

At least we know it works that way for 6Nimmt! and I cannot imagine it to be different for other games. That would be very weird and stupid coding.

[FrankJones]

What a needlessly abstruse Elo system!

(Perhaps the word "abstruse" is itself needlessly abstruse?)