Optimal Trading Filters and the Value of Predictive Signals
Matteo Smerlak
Abstract
The value of a forecast depends on its accuracy, but also on its actionability. In trading, an investor must translate return forecasts into positions while accounting for trading costs and the lingering impact of earlier trades. We solve the stationary quadratic trading problem with transient price impact using tools from prediction theory: Wiener-Hopf factorization gives the optimal policy as the composition of a conditional forecast and an explicit causal linear filter. We recover the Gârleanu-Pedersen aim-portfolio under exponential resilience and derive fractional trading filters under power-law impact decay, requiring memory beyond finite-dimensional LQR control. At fixed positive risk penalty, signals of equal instantaneous predictive strength have equal frictionless Markowitz value. With trading costs, net attainable value is determined by the mean square of the friction-weighted conditional forecast. Under power-law impact with exposure bleeding, we find an intermediate optimal mean-reversion rate for Ornstein-Uhlenbeck signals.
Read the AI summary, key takeaways and discussion on WOBR Quant Research.