1171
Tävlingar / Förändring av WPPR-rankingen
« skrivet: september 04, 2008, 06:21:08 »
Allright, Alvar! Nu har jag boxat lite på dina idéer, och kokat ihop följande (på engelska, orkar inte översätta nu):
Allright, here's a place based proposal for tournament ranking:
Each player gets a value (call it X) calculated using the formula C/(C+r^c) where:
C = A constant that determine how big numbers the system should use. Setting it to 1 gives PAPA 16 points, setting it to 10 gives 136 points and setting it to 100 gives 1096 points. It doesn't change the relation between the tournaments.
r = Each players ranking position.
c = A constant that determines how steep the curve should be. Decreasing it means that the number of players is of bigger significance, increasing it means the top ranked players is of bigger significance.
Then I added up all the X values of players #2 - #2598 and gave to #1. And then the X values of player #3 - #2598 and gave to #2. And so on, all the way down to player 2597. I.e. all players get the sum of the players below them. Lets call this new value Y.
The curve is quite steep for the first 50 players, going from 7.4 to 1.5 points for the Y value. Then there's a soft turn from #50 to #300 who gets 0.4 points. After that, it is linear towards the end. Tweaking the c constant is an easy way to alter the formula later on, if we want to change this curve.
The reason for the Y value is to include the number of players in the system into the formula. Only using the X value would give #1 (and #2, and #3, etc) the same amount, no matter if there was 3 or 3000 players in the system.
Then I added up all the Y values of all participants in a tournament, and using C=10 and c=1.6 this gave:
PAPA 11: 136,2 - 100%
EPC 07: 117,9 - 87%
SO 08: 64,4 - 47%
PMO 08: 45,0 - 33%
GPO 08: 42,0 - 31%
BPT 08: 17,5 - 13%
PAPA B: 37,2 - 27%
PAPA C: 4,37 - 3,1%
Monthly: 27,9 - 20%
Best 24: 89,0 - 65%
Last 24: 0,02 - 0,014%
This is including the four opt-outs.
GPO is still a bit underestimated I think, but I believe that when we recalculate the standings retroactively, the germans will get more points than they currently have, raising the level of GPO. We will know after more extensive simulations with more data.
What do you think about this one?
Allright, here's a place based proposal for tournament ranking:
Each player gets a value (call it X) calculated using the formula C/(C+r^c) where:
C = A constant that determine how big numbers the system should use. Setting it to 1 gives PAPA 16 points, setting it to 10 gives 136 points and setting it to 100 gives 1096 points. It doesn't change the relation between the tournaments.
r = Each players ranking position.
c = A constant that determines how steep the curve should be. Decreasing it means that the number of players is of bigger significance, increasing it means the top ranked players is of bigger significance.
Then I added up all the X values of players #2 - #2598 and gave to #1. And then the X values of player #3 - #2598 and gave to #2. And so on, all the way down to player 2597. I.e. all players get the sum of the players below them. Lets call this new value Y.
The curve is quite steep for the first 50 players, going from 7.4 to 1.5 points for the Y value. Then there's a soft turn from #50 to #300 who gets 0.4 points. After that, it is linear towards the end. Tweaking the c constant is an easy way to alter the formula later on, if we want to change this curve.
The reason for the Y value is to include the number of players in the system into the formula. Only using the X value would give #1 (and #2, and #3, etc) the same amount, no matter if there was 3 or 3000 players in the system.
Then I added up all the Y values of all participants in a tournament, and using C=10 and c=1.6 this gave:
PAPA 11: 136,2 - 100%
EPC 07: 117,9 - 87%
SO 08: 64,4 - 47%
PMO 08: 45,0 - 33%
GPO 08: 42,0 - 31%
BPT 08: 17,5 - 13%
PAPA B: 37,2 - 27%
PAPA C: 4,37 - 3,1%
Monthly: 27,9 - 20%
Best 24: 89,0 - 65%
Last 24: 0,02 - 0,014%
This is including the four opt-outs.
GPO is still a bit underestimated I think, but I believe that when we recalculate the standings retroactively, the germans will get more points than they currently have, raising the level of GPO. We will know after more extensive simulations with more data.
What do you think about this one?