Information Technology Reference
In-Depth Information
i.e., the stock market, to do this, due to the complexity of the information processing
involved. It could be possible, however, for such automatic trading algorithms to be
provided with human-generated estimates of known trading partners. The problem be-
comes even more complex in the case of markets solely populated by artificial trading
agents. However, as mentioned above, if sufficient information is available it may be
possibly to analyse the time series of shouts by a particular agent in order to estimate
the quality of its information. There is obviously a great deal of further work to be done
in this area, particularly in examining the effect of different evaluation rules and the
effect of different market structures on market (and agent) behaviour.
This paper has been more concerned with the performance of the market as a whole
rather than the situation of any one group of traders. There are, however, many interest-
ing question which can be asked about the ways in which information may be used. In
particular it would be interesting to consider how the better connected traders could ex-
ploit their advantageous positions to make a larger profit and also how they could exploit
the knowledge that other agents gain an advantage by considering information quality.
In order to properly understand these issues, however, it may be necessary to make
the market more sophisticated. Currently, traders may only trade once. This effectively
limits the ways in which traders can exploit information because as soon as a trader
makes a trade they are effectively removed from the market. As a consequence some
areas of the market may become stagnant as all available trades are made. This could
be remedied by the introduction of a continuous flow of buy and sell orders entering the
market. Allowing traders to interact multiple times and to develop more sophisticated
strategies, whilst preventing the market from stagnating.
In this paper only a small (but significant) difference between traders using the new
strategy and those not using it is demonstrated. Two points should be noted, firstly this
rule was not chosen as an optimal rule for increasing valuation accuracy, instead it was
chosen for its simplicity in demonstrating a point. Secondly the market employed in
these experiments is by its nature “one-shot” in that all traders only trade once for one
unit of the commodity. This naturally limits the opportunities for making profits, in
particular in eliminates reselling. In real markets, however, this is not the case. Traders
in real markets often trade many times for large volumes of products. A small increase
in accuracy when dealing with large volumes may make a significant difference to the
profit obtained. In the extreme case, foreign exchange markets have a turn over in the
region of one trillion dollars a day. Even very small increases in valuation accuracy in
contexts such as these can results in huge increases in profits.
6Conluion
This paper has demonstrated that simple markets populated by simple trading agents
may function despite trading constraints represented by an explicit, fixed network of
possible agent-agent interactions. We have not attempted to classify or analyse the
effect of topology in general. However, for one particular type of market, some in-
teresting and encouraging results have been shown. We have demonstrated that more
well-connected agents have an informational advantage within a market. Having more
neighbours ensures that a trading agent has a better picture of the market and, as a result,
