Phoxtrot wrote: ↑04 October 2021, 21:52any player that is high ELO enough compared to the other players to pretty much always be 1st or second and certainly almost never last would be screwed. He is already "paying" for being high ELO in the standard elo calculation that makes him win very little ELO point for beating someone to the point that he loses ELO for being 2nd of 4 or even for being 2nd of 6 players but now that is compounded by making his first places be worth half as much as his second places. That's pushing high ranked people away from playing tables with many players and that is not good.
No, if I understood you correctly, that's completely wrong.
Each person will have I will call it a 'score' which would work like chess elo, so where the sum of each 'score' on every game is always 0.
Then, the K-factor arrive, and multiply by 2 the elo gained/loss by the player who got few games.
So don't consider is elo is worth half as much, consider the person who got few games gain twice their elo.
There is some problem about that, when a player which got few games gain lots of games in a row (so btw, being able to reduce their K-factor to 20 when they reach top 20 would be a great idea), but form what I have seen, that work well otherwise.
The point of a multi-player factor, as I see it, is that because there are many 1 to 1 comparisons in a multi-player game, you need to scale it down so that those multi-payer games don't end up being so much more important than 2 players games. How much you scale the ELO variation down for say a 6-players game might logically take into account how long a 6-players game of that particular game takes in average compared to a 2-players game and it might take into account how much more (or less) "chaotic/based on luck" such a game. If you make the scaling depending on the position at the end of the game then you introduce a bias that is correlated to how "good/bad" each player is which is already taken into account in the basic ELO computation.
Again, the scaling does not depend of position. It depend of the number of games played by each person. You can see on my post 'If the two first players have >50 games, and the three last <40 games, that can be coherent, let me check the profile:' this was because these three players got <40games. And that caused their gain of elo of the three last players were halved, and not because they were in the second half of the game.
For the rest of your post, yeah this system isn't good, but the elo system need to not be too complicated if possible, and be able to classify player in a lot of range of differents games, and differents numbers of players. (Creating distinct ranking risk to split the community.)
