Information Technology Reference
In-Depth Information
1
Day
Finished Goods
Ratio Offer/Demand
Normalized Price
0.8
0.6
0.4
0.2
0
−0.2
−0.4
−0.6
−0.8
−1
Over−supply
Balanced
Scarcity
Regime
Fig. 8.
Game 1189@tac4 (Final TAC SCM 04) - Correlation coefficients between regimes and
day in the game, the ratio of offer to demand, normalized price (np), and quantity of finished
goods inventory in the low market segment
Figure 7 shows the quantity of the finished goods inventory (FG), the ratio of offer
to demand, which represents the proportion of the market demand that is satisfied, and
the normalized price (np) over time
1
.
The regimes identified by our approach are superimposed, where
S
(or
R
3
)repre-
sents scarcity,
B
(or
R
2
) balanced, and
O
(or
R
1
) over-supply. These factors clearly
correlate with market regimes, but they are not directly visible to the agent during the
game. For example, the figure shows that when the offer to demand ratio is high (i.e.
over-supply) prices are low and vice versa. We can observe that the ratio of offer to de-
mand changes significantly during the game. For instance, on day 141 the ratio of offer
to demand is approximatively 1.4 and prices are high. On day 179 the ratio of offer to
demand is much higher, approximatively 4.5, and prices are lower. We can also observe
that prices tend to lag changes in ratio of offer to demand.
A correlation analysis of the market parameters is shown in Figure 8. The p-values
for the correlation analysis are all less than 0.01. Regime
R
1
(over-supply) correlates
strongly and positively with time, ratio of offer to demand, and quantity of finished
goods inventory, and negatively with normalized price. On the other hand, in Regime
R
3
(scarcity) we observe a strong negative correlation with the selected market parameters.
Figure 9 shows the relative probabilities of each regime over the course of a game.
The graph shows that different regimes are dominant at different points in the game,
and that there are brief intervals during which two regimes are almost equally likely.
An agent could use this information to decide which strategy, or mixture of strategies,
to follow.
1
The quantity of the finished goods inventory is affected by other factors, such as storage cost,
which have changed in the TAC SCM 2005 games. In 2005 games agents tend to build to order
and keep most of their inventory in the form of parts, not finished products.
