About the TAC SCM competition
The Trading Agent Competition is an annual competition with the goal of encouraging research into the design of autonomous trading agents.
In the supply chain management scenario (a travel agent scenario exists as a separate competition), six agents compete as computer manufacturers in a simulated economy.
Agents must procure components such as CPUs and memory from suppliers,
bid for
sales contracts with customers,
decide what types of computers to manufacture
from these components as constrained by factory resources, and decide which computers to deliver
to whom and by when. Component suppliers and customers of computers are managed
by a game server, and a single game lasts 220 simulated days, each taking 15 seconds of real time.
Uncertainty concerning future supply capacities and prices, customer demand, and opponent strategies results in a challenging domain.
The competition is organized as follows. After a qualifying round and a seeding round, the top 24 teams advance to the final round of competition, which takes place over three days. Each day, agents are divided into groups of six, with each group playing a number of games against itself.
The top three agents from each group then advance to the next day.
Because each agent faces the same opponents during a single day's games, there is an opportunity for adapting to the behavior of opponents from game to game.
Any adaptation must be automatic, however, as no human-made changes are allowed during a day's games.
Additional background on the TAC SCM scenario is available
here.
The full specification is available as a PDF
here.
