Unofficial Contest Ranking
I present to you the three winners of the contest : @reCurse, @Neumann and @RoboStac !
They will be forever recognized with their names and avatars added to the puzzle’s statement.
Additional information
The final ranking procedure was described here.
As per the procedure, a total of 4500 games (900 per player) was made to provide players with an accurate ranking of their AI.
The scores obtained are placed on an arbitrary log scale. Do not be fooled: the log scale hides the utter domination by reCurse’s AI! (overall 98.5% winrate in the re-run !)
The ranking procedure allows for the calculation of a theoretical winrate between two AIs. The table below shows the difference between the theoretical and the observed winrate between two players.
It may be interpreted as indicating an AI’s over-specialization against another AI.
Note: the note in the figure is wrong and should be read as “Gloopy’s observed winrate against pb4 was 16 points above its expected value” ie: “Gloopy overspecialized against pb4”
More additional information
Raw winrates during the re-run:
Name1 Name2 Win_%
pb4 MSmits 80.00
pb4 mykeich 92.00
pb4 RoboStac 30.00
pb4 DaFish 86.00
pb4 Gloopy 72.00
pb4 LeRenard 79.00
pb4 jolindien 64.00
MSmits mykeich 54.00
MSmits DaFish 57.00
MSmits Gloopy 60.00
RoboStac MSmits 78.00
RoboStac mykeich 77.00
RoboStac DaFish 65.00
RoboStac Gloopy 82.00
RoboStac LeRenard 42.00
RoboStac jolindien 53.00
DaFish mykeich 41.00
Neumann pb4 67.00
Neumann MSmits 95.00
Neumann mykeich 97.00
Neumann RoboStac 71.00
Neumann DaFish 96.00
Neumann Gloopy 97.00
Neumann LeRenard 95.00
Neumann jolindien 85.00
Gloopy mykeich 47.00
Gloopy DaFish 45.00
LeRenard MSmits 51.00
LeRenard mykeich 77.00
LeRenard DaFish 79.00
LeRenard Gloopy 64.00
jolindien MSmits 73.00
jolindien mykeich 80.00
jolindien DaFish 70.00
jolindien Gloopy 82.00
jolindien LeRenard 60.00
reCurse pb4 97.50
reCurse MSmits 99.00
reCurse mykeich 100.00
reCurse RoboStac 98.00
reCurse DaFish 99.00
reCurse Neumann 94.00
reCurse Gloopy 100.00
reCurse LeRenard 100.00
reCurse jolindien 99.00
Note: Due to unforeseen events, the python script was not used to calculate the score. An alternate method (hmmmm…Excel…hmmm) was used to apply the same ranking principle.
