Strategic trading agents via market modelling

Abstract

This paper presents some key strategies applied in jackaroo agent. Most of the strategies are rooted in theoretical modelling and statistic analysis of TAC-03 SCM game. We model the product market with a variation of Cournot game and specify the component market by constant-supply model. We outline the basic theory and algorithms dealing with component procuring, product pricing, production scheduling and price forecasting.

Learning market prices in real-time supply chain management

“…(10) ensures production does not exceed factory capacity. The expected component need is calculated in (11), and supply constraints are enforced by ensuring that the closing component inventory for any day is non-negative (12). Finally, (13) ensures that the closing product inventory level for each day is non-negative.…”

“…There have been many different approaches to pricing strategies in the competitions to date. Game theory and a variation of the Cournot game is used in [11] to model the customer market. They investigate the relationship between market demand and product price and use this as a way to predict market-clearing price.…”

Markets come to bits: Evolution, computation and markomata in economic science

“…One suspects if they had extended their consideration to the constitution of preference functions, they would find that the problem of non-computability (and not just computational complexity) extends all the way down to the root of their initial assumptions concerning agent intentionality; although in practice, AI researchers have been happy to work with higher-level primitives, such as given piecewise linear demand curves. Other researchers have explored the use of genetic algorithms to search over a space of possible variant auction mechanisms in order to arrive at some notion of an improved mechanism design, although their findings have been hampered by an inadequate parameterization of the full range of species of auctions (unable to 18 The latest round of TAC is described in Wellman et al (2004) and Zhang et al (2004) encompass even the taxonomy in Wurman et al, 2001), and an undue dependence upon zerointelligence agents to reduce their domain to a closed form. 20 Whatever the present obstacles which hamper their attempts to feel their way through the economic landscape, it is interesting to see that the AI researchers have essentially traversed the same trajectory from initial stress on agent models to treating different species of markets as algorithms, in much the way the other groups have done.…”

Strategic Interactions in a Supply Chain Game

“…Early in the qualifying rounds we noticed Jackaroo's distinct saw-toothed profit timeline, indicating a steady increase in wealth with large periodic drops corresponding to bulk deliveries of supplies. This pattern was the result of large supply orders placed early in the game (over the first 7 days, not just day 0) for delivery at regular intervals (Zhang et al 2004).…”