The ICPC AutoAnalyst Project
Fredrik Heintz, Fredrik Niemelä, Stein Norheim
This talk presents the AutoAnalyst project. The goal of the project is to automatically analyze ICPC style contests in real time and make the result of the analysis available in a machine readable form. The underlying purpose is to enhance the contest experience of the audience through fascinating insights and exciting news. The analysis includes extracting important events such as first team to solve a problem and significant changes in the scoreboard, estimating what problems the different teams are working on, and computing when a team must solve its next problem in order to advance in the ranking. This type of analysis is currently provided by human analysts at the ICPC World Finals. This project makes that type of analysis available to other ICPC style contests with a minimum effort.
The AutoAnalyst tool uses different sources of inputs such as the CCS event feed, background information from the ICPC registration system, and backups of the contestants' directories if they are available. These must be provided by the contest organizers. The more information sources made available, the more interesting analyses the tool can make. The tool makes its output available both in the form of APIs and machine accessible feeds. The tool is documented and easy to set up and configure for someone with reasonable experience of Unix, databases and web servers. The tool does not provide graphical user interfaces to the information as that is part of other ACM ICPC projects such as SecondScreen. The intended users are the ACM ICPC World Finals and Regionals.
In this talk we demonstrate the tool, describe its design and present the future plans of the project.