Collusion features from a simple two-dimensional game

Data and supplementary files for TUCS Technical Report "Detecting a colluding subset in a simple two-dimensional game" [1]

The files aoi.arff, block.arff and tabu.arff contain features for area-of-intererst, blocking and tabu colluders respectively. Each file contains 23800 rows from 500 simulated games. Each row represents one subset of players in a single game. Simulations used 2 to 6 colluders and 100 games were performed with each number of colluders. For each game only data of subsets of players with a size equal to the number of colluders were collected. The data contains features proposed in [2] and in addition the following features:

- colluderRatio: fraction of colluders in the subset
- colluders: number of colluders and subset's size
- collusion: 1 if all members of the subset are colluders, 0 otherwise.

swiftness.csv contains mean time used to generate data for each subset size.

wins.csv contains collusion pay-off as wins per 1000 games per colluder in dispenser setting from [2].

experiment_final.Rmd is the R Markdown file for generating the figures in the report.

[1] Jussi Laasonen, Jouni Smed, Detecting a Colluding Subset in a Simple Two-Dimensional Game. TUCS Technical Reports 1074, 2013.

[2] J. Laasonen, T. Knuutila, and J. Smed, “Eliciting collusion features,” in Proceedings of the 4th International ICST Conference on Simulation Tools and Techniques, 2011, pp. 296–303.