One-day Workshop: Tuesday T-1

ON SYSTEMATIC TRADING STRATEGIES: NEW DIRECTIONS FOR ROBUST CONTROL

B. Ross Barmish, University of Wisconsin- Madison
Farshid M. Asl, New York University

8:30am - 5:00pm
Adams

Abstract
The main goal of the workshop is to formulate new robust control problems associated with systematic trading. While the focal point will primarily be stocks and options, the central concepts will also be applicable in other trading domains such as commodities and foreign exchange. The workshop will include both theoretical and applied components with considerable emphasis on modelling and performance metrics. To this end, the point of view taken will be as follows: Both fundamental and technical analysis will be used to identify an appropriate long buy or short sell point. However, once the trade is ongoing, mainly technical analysis will dictate how one's position should be adjusted over time. Further to the issue of performance, modelling the trader's objectives will also be emphasized. In this regard, we will show by example how one can easily reach erroneous conclusions about the efficacy of a trading method based on a performance index which does not truly reflect a trader's utility. For example, well known strategies such as "Buy and Hold" or "Dogs of the Dow" may result in unacceptable
draw-down which is often neglected in the modelling process. The introduction to trading concepts will be explained in the context of a basic feedback loop with the control corresponding to modulation of the amount invested over time. Taking the point of view of robust control, the formulation of a new set of research directions will not involve any type of stochastic model which includes predictions of future prices. In fact, the avoidance of a stochastic model will be one of the technical noveltiesmotivating new research problems. To this end, the instructors will discuss some limitations of stochastic models when one makes the transition from theory to applications. Consistent with the tenets of robust control, we will pose new research problems which are aimed at providing "certification" of the performance of various trading strategies with respect to a class P of admissible stock price variations.

Finally, the topic of back-testing of candidate trading strategies will also be discussed. To this end, the workshop will include an explanation of the pitfalls of "data snooping" and how one can easily reach erroneous conclusions regarding the efficacy of a feedback control law. This section will also include a discussion of simulation codes and a number of demonstrations.