The emergence of computational markets presents new opportunities for automated trading by autonomous software agents. To spur research and focus attention on common problems, we instituted an annual Trading Agent Competition (TAC). To date, TAC has featured four market games:
- TAC Travel. The original (classic) TAC market game presents a travel-shopping scenario, where agents compete to assemble travel packages for their clients through simultaneous interacting markets (Wellman et al., 2007).
- debut: 2000; designed at Michigan (operated by SICS starting 2002)
- Michigan agent: Walverine
- TAC/SCM. In the TAC Supply Chain Management game, agents representing PC manufacturers compete to procure components and sell finished PCs, over a simulated 220-day horizon.
- debut: 2003; designed by CMU and SICS
- Michigan agent: Deep Maize
- CAT. In the TAC Market Design game, players are market specialists competing to lure trading agents to participate in their respective markets. Because the trading agents are built into the environment, the Market Design game is a “reverse TAC”.
- debut: 2007; designed by CUNY, U Liverpool, and U Southampton
- Michigan agent: Manx
- TAC/AA. The TAC Ad Auction game presents a sponsored search scenario, where advertisers compete for search advertising slots, in order to lure simulated users to click on ads and purchase their products.
- debut: 2009; designed at Michigan
Related Projects and Publications:
- A trading agent competition for the research community
- Trading Agents
- Strategy and Mechanism Lessons from the First Ad Auctions Trading Agent Competition
- Distributed feedback control for decision making on supply chains
- Autonomous Bidding Agents: Strategies and Lessons from the Trading Agent Competition
- Learning Improved Entertainment Trading Strategies for the TAC Travel Game
- Forecasting Market Prices in a Supply Chain Game
- Designing the Ad Auctions Game for the Trading Agent Competition
