U-Mart2004

To estimating parameters for experiments with human agents, 2 kinds of rule were tried. These were different in sessions interval, and number of itayose per day. This was the first time to held multi-location experiment by point to point connection.

･Date
   2, October, 2004
･Place
   Tokyo Institute of Technology
& Osaka City University
(Demo Session in Autumn Conference of Japan Association of Evolutionary Economics)
･Participants
   Machine Agent:　5 Agents
Human Agent: 　30Agents
･Agent type
  Machine & Human

For 1st Trial
1,　# Sessions per day; 2 time
2.　Session interval;　 30Secondss
3.　Trading period: 20days
4.　Total # sessions; 40 times

For 2nd Trial
1.　# Sessions per day; 90 times
 2.　Session interval;　 1Seconds
3.　Trading period; 20 days
4.　Total # sessions; 1800 times


【Machine】

Team Agent name Member ID
Kinki University

MarketMaker

 h31
Kinki University TaniguchiSeminar h32
Kinki University TaniguchiSeminar2 h33
Kinki University TaniguchiSeminar3 h34
Kinki University TaniguchiSeminar4 h35

【Human】

Team Member ID Member name
Chuo University h1〜h15 Takada・Kousin・Hutakushi・Kobayashi・Otuka・Chiba・Shinjyo・Abe・Yagyu
（Itai・Kanami・Akazaki・Maeyama・Segawa・Kamio）（Shinohara・Koyama・Kanada・Tahira・Osako・Nakamura）
Kinki University h16〜h20 Ikeda・Matuura・Yamamoto・Asahina・Noguchi
Osaka City University h26〜h28 Nakajima・Ueki・Morimoto

●Result of the 1st trial：Total（Machine＆Human）

Rank Agent name Member ID Team CASH（\）No.
No.1 Matsuura h17

Kinki University
1,683,575,000
No.2 Kamio h15 Chuo University
1,146,053,000
No.3 Nakajima h26 Osaka City University
1,117,100,000
No.4 Chiba h7 Chuo University
1,047,100,000
No.5 Kanami h11 Chuo University
1,035,990,000

●Result of the 2nd trial：Total（Machine＆Human）

Rank Agent name Member ID Team CASH（\）
No.1 Yamamoto h18 Kinki University
3,968,359,000
No.2 TaniguchiSeminar2 h33 Kinki University
3,426,796,000
No.3 TaniguchiSeminar h32 Kinki university
3,088,683,000
No.4 TaniguchiSeminar3 h34 Kinki University
2,918,254,000
No.5 Noguchi h20 Kinki University
2,559,112,000

【U-Mart2004】

In the first trial, human agents could show good performance because they can carefully consider to trade in 30 seconds for 1 session. However, in the next trial, itayose was held in each 1 second.
This settings were proposed to realize continuous session. Human agents were confused and machine agents could show good performance. Indeed, only 2 human agents were bankrupted in the first trial, but it was increased to 11 agents in the second trial. In this experiment adding to the agents in pursuit of good performance, a market maker participated. The agent was participated to check the availability and feasibility. The agent could gain a surefire profit constantly in both trials.