Alexander Barzykin, a director in the foreign exchange, rates and commodities team at HSBC (NYSE: HSBC) in London, published a quantitative modelling framework addressing informational risks in over-the-counter foreign exchange market-making. Developed alongside Philippe Bergault, Olivier Guéant, and Malo Lemmel, the model formalizes internal production logic to optimize liquidity management against adverse selection and price reading.
The Bottom Line
- Model Architecture: Built as a stochastic optimal control problem solved via dynamic programming, the framework models how dealers skew bid/ask quotes to manage inventory without leaking actionable market signals.
- Informational Dynamics: Identifies adverse selection from asymmetric client latency and subtle price reading where algorithmic competitors reverse-engineer internal risk-management footprints.
- Strategic Shift: Suggests that rational acceptance of specific adverse selection flows from long-term signal providers can be leveraged to hedge the broader trading franchise.
Decoding the Over-the-Counter Liquidity Puzzle
Foreign exchange trading operates predominantly over-the-counter, meaning dealers quote prices directly to individual clients or aggregators rather than executing through a centralized central limit order book. This bilateral structure leaves dealers exposed to idiosyncratic, client-specific risks that are considerably more pronounced than those found in centralised markets. Managing asset inventory while continuously calculating competitive bid and ask quotes transforms into a complex multi-period optimization problem for quantitative modellers.
According to research detailed by Barzykin and co-authors, the fundamental hazard stems from information leakage. When a dealer adjusts or skews prices to manage accumulated inventory, that adjustment transmits a supply-demand imbalance signal to the broader market. Algorithmic participants read these footprint signals to anticipate future price trajectories. Left unmitigated, this dynamic causes prices to drift unfavorably against the dealer’s book, creating severe inventory drawdown.
| Risk Factor | Primary Mechanism | Quantitative Treatment |
|---|---|---|
| Adverse Selection | Client possesses superior information or latency advantage in OTC bilateral setup. | Stochastic differential equations within a stochastic optimal control framework. |
| Price Reading | Algos detect inventory-driven price adjustments made by the dealer’s risk desk. | Dynamic programming optimization balancing risk management and information leakage. |
Stochastic Control and the Economics of Information Leakage
To combat information risk, Barzykin’s framework relies on stochastic optimal control solved via dynamic programming. This technique allows risk modellers to treat inventory management as a continuous control problem where prices are perturbed by stochastic differential equations. The resulting strategy dictates precisely how a dealer should skew quotes to minimize inventory risk while limiting the informational value transferred to competing algorithms.
Interestingly, the research challenges the conventional Wall Street dogma that adverse selection is universally destructive. Barzykin points out that clients exhibiting adverse selection often do so because they rely on stable, long-term informational signals. Market-makers can rationally choose to absorb some level of adverse selection from these specific counterparties if the resulting order flow provides actionable intelligence that helps risk-manage the remainder of the institutional franchise.
This insight bridges micro-structural FX mechanics with broader macroeconomic liquidity provision.
Future Horizons in Quantitative Market Structure
While the current mathematical framework was conceptualized specifically for foreign exchange, its underlying architecture applies to any bilateral, quote-driven asset class where information asymmetry dictates execution quality. Future academic and industry research streams are expected to focus heavily on joint modeling. By treating the dealer’s optimization problem and the client’s execution problem simultaneously through game theory, quants hope to better capture equilibrium dynamics.
Another critical frontier involves reputation feedback loops. A market-maker’s operational behavior—such as the frequency with which it rejects quotes or throttles toxic flow—directly impacts the future composition and volume of incoming client orders. Incorporating these behavioral feedback loops into production algorithms will dictate which financial institutions maintain pricing dominance across global funding markets.
