Stochastic Controller

Optimal quoting actions for a market-making agent are computed using HJB value functions driven by Bayesian beliefs. The controller applies online Bayesian updates for drift, volatility, and fill intensity to determine the next action.

Stochastic control for market making determines optimal sequential decisions between estimation and execution layers. HJB solvers are needed when computing continuous-time optimal actions that incorporate inventory skew and predictive drift adjustments.

Market regime shifts are detected using a changepoint detector that triggers belief resets. This allows the stochastic controller to adapt to new market conditions and maintain accurate Bayesian updates.

Yes, the stochastic controller requires online Bayesian updates using Normal-Inverse-Gamma, Gamma, and Dirichlet distributions. These distributions update beliefs about drift, volatility, and fill intensity to drive optimal decisions.

Inventory skew and predictive drift adjustments are implemented through HJB-based optimal quoting. The HJB solver computes quote edges by incorporating current inventory state and Bayesian belief updates.

Yes, a continuous-time HJB solver is required to produce optimal actions and manage risk. The solver computes value functions that account for inventory skew, predictive drift, and Bayesian belief updates.