2017 Competition Results

Last week, the 2017 Text-Based Adventure AI Competition results were announced at the IEEE Computational Intelligence and Games Conference in New York. The results were:

1st. BYU 2016 Agent - Nancy Fulda, Daniel Ricks, Ben Murdoch, David Wingate. Brigham Young University, USA

2nd. CARL – Nancy Fulda, Daniel Ricks, Ben Murdoch, David Wingate. Brigham Young University, USA

3rd. The Golovin Agent – Bartosz Kostka, Jarosław Kwiecień, Jakub Kowalski and Pawel Rychlikowski. University of Wrocław, Poland

Our winners again this year, was last year’s winning agent that the team from BYU have made available as open source and published at IJCAI.

2nd and 3rd place was a close decision as both agents achieved the same score on our test game. Therefore, we went to our second scoring criteria (a subjective decision by the competition judges on the agent’s freedom from a priori bias) and awarded 2nd place to BYU’s new agent CARL.

Our 3rd place competitor, The Golovin Agent, is also available as open source and was both published and nominated for the best paper award at CIG.

We would like to take this opportunity to publicly thank all competitors for the world-class research they have produced in participating in this competition. In particular we are taking inspiration from the evaluation methods they have used and considering options to adopt similar approaches in future iterations of the competition.

Finally, as shown by the evaluation in both papers and the competition results, there is still significant room for improvement in general text-based adventure AI performance. We hope this competition, the codebases available and the papers published all help to engage more of the game AI community in pursuing this challenging open problem!