Humanness results

Humanness % is calculated by dividing the number of times a player (bot or human) was judged to be human by the total number of times it was judged. For this calculation, only judgements made by humans are counted. Even though we call this "humanness", it really only measures how humanlike a player seems to the judges. This explains why it is possible for bots to get a rating higher than the ratings of the real humans - the bots are good at tricking the judges.

Each player was judged around 25 times. This makes it unlikely that the high humanness ratings are a fluke. For example, if a bot was really only 35% human, then there is only about a 5% chance of getting more than a 50% rating. If a bot was really only 30% human, then this drops to about 2%.

As well as humans and competition bots, a few epic bots, the bots built into the game, briefly took part. These bots don't play by the same rules as the competition bots, but their ratings are reported here for interest.

Most human bots

bot name humanness %
MirrorBot 52.2 %
UT^2 51.9 %
ICE-CIG2012 36.0 %
NeuroBot 26.1 %
GladiatorBot 21.7 %
AmisBot 16.0 %
average 34.2 %

Most human humans

player name humanness %
Samaneh Rastegari 53.3 %
Craig Speelman 52.2 %
John Weise 30.8 %
Chris Holme 26.3 %
average 41.4 %

Most human epic bots

humanness %
average 37.8 %

Judging results

Judging accuracy is calculated as the number of correct judgements made by a player, divided by the total number of judgements made by that player. Note that the bots also judged (but their judgements were not used in the calculation of humanness ratings).

Best bot judges

bot name accuracy %
MirrorBot 60.0 %
NeuroBot 56.4 %
UT^2 54.6 %
GladiatorBot 50.5 %
ICE-CIG2012 45.8 %
AmisBot 44.4 %

Best human judges

human name accuracy %
Chris Holme 60.9 %
John Weise 60.8 %
Craig Speelman 50.0 %
Samaneh Rastegari 47.8 %