Main Conference Day TwoWednesday, 19 November 2025 - GMT (Greenwich Mean Time, GMTZ)

As algo wheels reshape how buy-side firms select brokers, top quality algo performance is no longer a competitive edge—it’s a baseline requirement. Generic factors like bid-ask spread, volatility, top of the book liquidity, order size, etc., are no longer sufficient to navigate today’s intricate market dynamics. This session introduces “Lenses,” a powerful, new framework designed for the dynamic evaluation and enhancement of algorithmic trading performance.

Portfolio allocation is crucial for investment institutions with diverse market functions and strategies. Allocations can be static, based on long-term expectations, or dynamic, incorporating time-varying stochastic factors. A typical dynamic allocation goal consists of an expectation of a deterministic function of the portfolio wealth at a fixed time-horizon. Usually, the portfolio weights depend on prices, wealth, and a time to maturity.

Our paper explores a novel optimization framework without fixed time horizons, focusing on portfolio returns rather than wealth – an approach particularly relevant to sovereign and pension funds. We adopt a comprehensive stationary approach across utility (regularized running forward-looking returns), model (asset returns driven by stationary market factors like sentiments and VIX), and weights (functions of market drivers without explicit time-dependence).

We derived optimal equations for asset weights with arbitrary utility functions, potentially solvable numerically. Additionally, we developed an efficient convex optimization routine for a special case of concave utilities. Morever, for quadratic utilities, we found out that the optimal weights satisfy a matrix Fredholm equation and develop a highly efficient numerical solution. Finally, we obtained explicit analytics for the Kim-Omberg model solving the Fredholm equation. We have confirmed our findings via numerous numerical experiments: we calibrated the Kim-Omberg model using market sentiment as the stochastic drift factor and holistically analyzed analytical and numerical results.

Adjusting exposure to volatility shocks

Join us for an exclusive wellness session in Peninsula East. Rest, relax, recharge, and reset to absorb more in the upcoming sessions.

We investigate portfolio optimization in financial markets from a trading and risk management perspective. We term this task Risk-Aware Trading Portfolio Optimization (RATPO), formulate the corresponding optimization problem, and propose an efficient Risk-Aware Trading Swarm (RATS) algorithm to solve it. The key elements of RATPO are a generic initial portfolio P, a specific set of Unique Eligible Instruments (UEIs), their combination into an Eligible Optimization Strategy (EOS), an objective function, and a set of constraints. RATS searches for an optimal EOS that, added to P, improves the objective function repecting the constraints.
RATS is a specialized Particle Swarm Optimization method that leverages the parameterization of P in terms of UEIs, enables parallel computation with a large number of particles, and is fully general with respect to specific choices of the key elements, which can be customized to encode financial knowledge and needs of traders and risk managers.
We showcase two RATPO applications involving a real trading portfolio made of hundreds of different financial instruments, an objective function combining both market risk (VaR) and profit&loss measures, constrains on market sensitivities and UEIs trading costs. In the case of small-sized EOS, RATS successfully identifies the optimal solution and demonstrates robustness with respect to hyper-parameters tuning. In the case of large-sized EOS, RATS markedly improves the portfolio objective value, optimizing risk and capital charge while respecting risk limits and preserving expected profits.
Our work bridges the gap between the implementation of effective trading strategies and compliance with stringent regulatory and economic capital requirements, allowing a better alignment of business and risk management objectives.

In this talk we shall discuss an approach to forecasting inflation in both DM and EM using machine learning models, contrasting it to more traditional approaches, alongside a discussion of the types of data required. We shall also be discussing how such an approach has performed historically. Later, we discuss ways we can utilise such inflation forecasts within the systematic trading strategies for macro assets such as FX, and how they have performed in a live environment.

We study quality–diversity reinforcement learning for optimal execution, focusing on generating a repertoire of behaviourally distinct execution schedulers. Using MAP-Elites, we construct a structured portfolio of schedulers indexed by liquidity and volatility descriptors, enabling systematic coverage of regime-specialist strategies rather than reliance on a single optimised policy.

To achieve this, we develop GEO (Gymnasium for Executing Optimally), a Gymnasium-based environment for execution under empirically calibrated transient propagator models. Minute-bar equity data are used to estimate impact kernels out-of-sample, producing realistic dynamics consistent with transient, concave impact.

Execution policies are trained with Proximal Policy Optimisation (PPO) and evaluated against established baselines—TWAP, VWAP, POV—under strict train–test splits. Within this setup, schedulers adapt trajectories while respecting completion constraints, and quality–diversity search ensures that the solution space spans different market regimes.

The framework thus combines (i) empirically grounded propagator calibration, (ii) reinforcement learning–based schedulers, (iii) baseline-aligned evaluation, and (iv) quality–diversity methods for execution under regime heterogeneity.

The inflationary environment of 2022 reminded investors that a standard 60/40 portfolio of stocks and bonds that produced double-digit losses was not immune to market turbulence. Experienced investors recognize that crisis periods are challenging to predict and their impact on performance can be devastating since most asset classes and investment strategies struggle during crises. A notable exception is the trend-following strategy used by Commodity Trading Advisors, which is often described as "crisis alpha" due to its propensity to perform well during crisis periods. While investors may allocate to CTAs for various reasons, including strong long-term performance and a low correlation to most asset classes and strategies, crisis alpha is a particularly attractive characteristic of CTA investments. Thus, when crises do occur, CTA investors expect to see strong performance and benefit from crisis alpha. What some investors may not have considered is how sensitive the crisis alpha benefits are to manager selection and to the number of managers in a portfolio. This presentation summarizes our recent research, which investigates the impact of manager selection and the number of managers in CTA portfolios on crisis alpha.

Panesthetic analysis optimises the portfolio not with respect to past returns, but with respect to estimated future returns in a number of potential market scenarios. These scenarios are estimated using nonlinear polymodels to capture possible changes of regime. Bayesian techniques and Kullblack-Liebler divergence are used to replicate a fund with a combination of benchmark strategies and identify its true "alpha" that accounts for its behavior under extreme events.

1. A new kind around the block in insurance finance.

2. Market overview and opportunities

3. Modeling solvency via equity

1. Thousands of pricing models for pricing a derivative
2. Model risk, calibration risk, parameters risk, implementation risk, data risk, …
3. Conic finance as a way out