Adding Supply/Demand Imbalance-Sensitivity to Simple Automated Trader-Agents

In most major financial markets nowadays very many of the participants are "robot traders", autonomous adaptive software agents empowered to buy and sell quantities of tradeable assets, such as stocks and shares, currencies, and commodities; in the past two decades such robots have largely replaced human traders at the point of execution. This paper addresses the question of how to make minimally simple robot traders sensitive to any imbalance between supply and demand that may occur in the market in the course of a trading session. Such imbalances typically are transient, and their occurrence is unpredictable, but when they do arise any human traders would automatically shift the prices that they quote, to reflect their expectations (grounded in basic microeconomics) that a momentary excess supply is an indication that prices are about to fall, while an excess demand is an indication that prices are about to rise. This can result in prices moving against a trader that is attempting to buy or sell a large quantity of some asset. In this paper we describe our work on exploring the use of multi-level order-flow imbalance} (MLOFI) as a usefully robust instantaneous measure of supply/demand imbalance, and we show how MLOFI can be used to give simple robot traders an opinion about where prices are heading in the immediate future, which mean that our imbalance-sensitive trading robots can serve as a platform for experimental study of issues arising in Nobel laureate Robert Shiller's recent work on Narrative Economics