Information Technology Reference
In-Depth Information
The number of PCs that can be assembled by an agent during the game is:
TotalPCs
= 440000
/
5
.
5 = 80000
This calculation assumes that the agent is making full use of its assembly capacity
from day 0 to 219. But this is not realistic as the first components cannot arrive before
day 3 and therefore the earlier that production can start is on day 4. Moreover, produc-
tion on the last day is useless as the agent is unable to deliver the PCs manufactured
that day. The number of effective cycles and total number of PCs then would be:
EffectiveCycles
= 215
2000 = 430000
TotalEffectivePCs
= 430000
/
5
.
5 = 78181
∗
We also have to take into account the assembly capacity of the suppliers in order
to determine how many components a day they can produce. The expected assembly
capacity of a supplier is 500 components a day, and the real production capacity for
every day for each supplier is calculated as follows:
C
p
(d)=max(0,C
p
(d-1)+rnd(-0.05,0.05)*C
nominal
+0.01*(C
nominal
-C
p
(d-1)))
Where C
nominal
denotes the nominal or expected assembly capacity. We have deter-
mined experimentally that the average number of components that a supplier manufac-
tures every day is 400 of each of the 2 types of components it can produce. Thus, the
total number of components that can be produced during the whole game and the to-
tal number of PCs that can be manufactured can be calculated. CPUs are produced by
two suppliers, Pintel and IMD, each of which provide two varieties of CPUs. If every
supplier produces 400 CPUs of each variety daily, then 1600 CPUs are produced every
day, which represents 1600 PCs. Considering that a supplier cannot produce any com-
ponents on the first two days and that the production on the last two days is useless, we
can approximate the total number of PCs that can be produced in a game:
AllP Cs
= 1600
∗
216 = 345600
Assuming that all agents manufacture approximately the same number of PCs:
345600
PCs/
6
agents
= 57600
PCs
This number is only an approximation and can vary from game to game.
4.2
Supplier Strategies
We explored a number of strategies to deal with the suppliers in the TAC SCM 2004
competition as well as in a number of controlled experiments. The underlying strategy is
based on what we call the Massive Simple Strategy (MSS) which simply sends 5 RFQs
with big quantities to every supplier for every component they supply on day 0. The 5
RFQs are split taking into consideration the number of days that the components will
last in the inventory, which can be computed by taking into consideration the assembly
capacity of the agent and the average number of cycles needed for one PC to be manu-
factured. The 5 RFQs request a number of components enough to manufacture between
55000 and 65000 PCs during the entire game, which ensures a production between 150
and 180 days. For instance, the following RFQ bundle shows how Socrates splits the
