The Gavel expo judging system, as used at HackMIT, HackHarvard, and many other hackathons, is based on asking judges to perform pairwise comparisons of different teams. The algorithm it uses is based on the CrowdBT algorithm defined and developed in this paper, and is effectively a maximum likelihood estimate for the ranking of entries using a slightly modified version of the Bradley-Terry model for pairwise comparisons. This results in a global ranking of the goodness of all teams, despite the fact that no judge has seen all teams and no team has been seen by all judges.
Built With
- jupyter
- pymc
- python
Log in or sign up for Devpost to join the conversation.