An Analysis of the 2004 Supply Chain Management Trading Agent Competition - Agent-Mediated Electronic Commerce - page 111

Ta b l e 6 . Average values of four factors correlated with market ASPs given for the PCs each agent
won and bid on. Day is the simulation day, Demand is market demand, Bids is the bid density
(defined in text), and PC inv is average manufacturer PC inventory.

Agent Sold, Day Bid, Day Sold, Demand Bid, Demand

FreeAgent 122 130 160 147

Mr.UMBC 124 129 150 148

UMTac-04 135 126 143 140

Botticelli 163 164 153 146

Deep Maize 122 123 168 158

SouthamptonSCM 105 112 143 136

Agent Sold, Bids Bid, Bids Sold, PC Inv Bid, PC Inv

FreeAgent 2.89 3.39 588 832

Mr.UMBC 2.58 3.26 761 813

UMTac-04 2.90 3.48 933 925

Botticelli 3.75 3.99 730 822

Deep Maize 2.40 3.35 551 675

SouthamptonSCM 3.38 3.71 836 939

ished in the order they did. The following is a brief synopsis of the important points
about each agent, starting from the lowest-scoring agent and working up.
SouthamptonSCM bought far more components than it sold, leaving it with large
amounts of unsold inventory and high storage costs. The agent had reasonably high
ASPs and winning percentages in the customer market, but bid on a very low fraction
of customer orders. The underlying problem was that the agent was not successful at
getting all of the complementary components needed for production of finished prod-
ucts.
Deep Maize had very high ASPs in the customer market. However, it also had in-
ventory management problems with complementary components. The agent was often
left with insufficient non-CPU inventory to manufacture and sell aggressively, and had
large numbers of unsold CPU components. The agent paid low average prices for non-
CPU components, so it may have been too selective about prices for these components.
Botticelli did not make any supply purchases at the start of the game, and did not
enter the market substantially until very late in the game. By this time there was too
little time remaining in the game and the customer market was too competitive to allow
for large profits.
UMTac-04 compared favorably to FreeAgent on virtually every metric we consid-
ered except average daily inventory and customer market ASP. This agent had the lowest
ASP of any agent. It sold later in the game in markets with relatively low demand, high
PC inventory; all of these factors correlate with low market prices.
Mr UMBC had a better ASP than UMTac-04 , but lower than FreeAgent .Itsold
in lower demand and higher inventory markets than FreeAgent . This agent also had a
sizable amount of non-CPU inventory unsold at the end of the game and higher daily
inventories than FreeAgent .
FreeAgent made some interesting strategic choices for the final round. It opted to
block other agents only on non-CPU inventory, while acquiring CPUs at low prices on