Ronald M. Harstad - Developments in automated short-term stockmarket trading

Financial engineers have developed sophisticated models for estimating both values of particular assets, and portfolio balance adjustments. The developments I will report on are less sophisticated from the standpoint of computational design, but are typically being used over much shorter time horizons. An artificially intelligent agent runs parallel processing on 4 NT workstations, and spots trades that are likely to yield a profit in a very short time. Positions are opened without guidance or approval of a human being. Balancing trades closing the positions usually involve human choices over computer-generated alternatives. The typical position is opened and closed in 10-40 seconds. This agent has recently grown to become responsible for over 3 percent of trading volume on the NASDAQ market.