Information Technology Reference
In-Depth Information
Profits for 10 agents and 5 strategic bidders
Profits auctioneer, capacity agents 50
2
34
17,17,17
15,15,15
13,13,13
9,9,9
32
1.5
30
28
1
26
24
0.5
22
20
0
18
1
2
3
4
5
6
7
8
9 10
0 1 2 3 4 5 6 7 8 9 10
agents
number strategic bidding agents
Fig. 2. Profits bidders (a) and auctioneers (b)
The strategic bidders use the possibility to overbid in auctions with good effect.
Results for one strategic bidder competing with 9 myopic bidders are even more
skewed as the strategic bidder is able to achieve an average profit of 1 . 8and
is filled to capacity. For 10 strategic bidders, agents are however at the same
level of capacity use as for all myopic bidders and loads are distributed evenly.
However, the average profit of the agents has dropped to 0 . 7. This is worse than
in the all myopic case (
1 . 2), as agents strongly compete for the items.
In Figure 2b, we show the total average profit of the auctioneers as a function
of the number of agents using a speculative bidding strategy. We present various
settings of the number of loads available in fruitful regions <F 1 ,F 2 ,F 3 > .
Clearly, the auctioneers profit from the agents trying to outthink and outbid
each other in competitive settings. The slope of the curve is determined by the
bidfactor employed by the agents and bounded by the total complementary value
of all loads for the agents. The added profits of the auctioneers are reduced as
the number of loads offered in the auctions approaches the total capacity of the
agents. For such scenarios it becomes more useful for agents to wait for cheap
resources in auctions and not bid strategically. Plots of the profits of the agents
like in Figure 2a for the < 17 , 17 , 17 > scenario show that agents with strategic
bidding perform near identical to myopic players.
In Figure 3a, we compare the profits of the auctioneers as function of the
number of strategic bidders in a more stochastic setting. We have plotted the
curve as usual for the traditional case of 10 agents as above for 9 loads per 3
fruitful regions ( < 9 , 9 , 9 > ). We also present results for the same number of
fruitful regions, but with a random number of loads of 8, 9, or 10 equiprobably
and independently available for each of the three fruitful regions each epoch ( <
9 , 9 , 9 >S (tochastic)). The agents therefore are faced not only with competition,
but also with a varying supply of loads for sale per fruitful region.
The agents react by optimizing for the worst case scenario, that of least avail-
able supply. We also observed that competition between the agents can be more
varied for a larger number of fruitful regions. The agents experience a varied
level of competition as the agents oscilate in their choice of targetting the fruit-
ful regions, and hence the competition between the agents induces stochasticity
in the supply.
 
Search WWH ::




Custom Search