Paper:

This repository is a fork of OpenSpiel that was used to run experiments for the paper "Understanding Iterative Combinatorial Auction Designs via Multi-Agent Reinforcement Learning" (d'Eon, Newman, Leyton-Brown, EC'24).

Some useful starting points are:

• Clock auction implementation (open_spiel/python/games/): clock_auction.py, along with supporting files clock_auction_base.py, clock_auction_bidders.py, clock_auction_observer.py, and clock_auction_parser.py
• Value samplers (in /open_spiel/python/examples/): pysats.py (Python implementation of the MRVM from the Spectrum Auction Test Suite) and sats_game_sampler.py
• Database (in web/auctions/auctions/): models.py, savers.py, and webutils.py
• Experiment scripts (in web/auctions/auctions/management/commands/): cfr.py, ppo.py
• Paper experiments: see notebooks/readme.md