All entries were scored before the

All entries were scored before the conference through run-ning them on ten levels of increasing difficulty, and using the total distance traveled on these levels as the score. The scoring procedure was deterministic, as the same random seed was used for all controllers, except in the few cases where the controller was nondeterministic. The scoring method uses a supplied random number seed to generate the levels. Competitors were asked to score their own submissions with seed 0 so that this score could be verified by the organizers, but the seed used for the competition scoring was not generated until after the submission deadline, so that competitors could not optimize their controllers for beating a particular sequence of levels.For the second phase of the competition (the CIG phase),we discovered some time before the submission deadline that two of the submitted controllers were able to clear all levels for some random seeds. We therefore modified the scoring method so as to make it possible to differentiate better between high-