Styled Algorithmic Trading and the MV-MVP Style
18 Pages Posted: 9 Oct 2014
Date Written: October 7, 2014
Abstract
Except for pre-trade assessment, algorithmic trading for optimal execution has to be dynamically adapted to real-time market environments and inventory positions. This makes dynamic programming (DP) the most natural approach. Due to the curse of dimensionality, however, it is highly challenging to design a rigorous DP strategy that invites a tractable solution procedure. Approximate dynamic programming (ADP) has hence emerged as a practical solution, and the proposed framework of styled trading follows the ADP paradigm.
A styled model makes a particular choice for its state variables as well as the parametric forms of its trading policies. Its calibration then becomes a nonlinear, constrained, but lower-dimensional optimization problem. The current framework also goes beyond the additive objectives normally required by a traditional DP strategy and its associated Bellman solution procedure. It works with the mean-variance utilities of Almgren and Chriss (J. Risk,
Keywords: Dynamic programming, style, moneyness, aggressive, passive, participation, parametric, stochastic, gradient descent, binomial trees
JEL Classification: G24, C61
Suggested Citation: Suggested Citation