Approach
Systematic macro research & technology.
Our research focuses on directional and relative-value opportunities in liquid futures. Signals are evaluated across market conditions and time horizons, with attention to portfolio risk, trading costs and practical implementation.
THE PROCESS
Information
A broad, structured feature library brings together economic and market information to support research across assets and strategies.
Models
Statistical and machine-learning models provide complementary views of the data. Evaluation spans training windows and forecast horizons.
Forecasts
Forecast evaluation examines predictive strength, persistence and consistency through time, including where models agree and where they diverge.
Risk
A multi-asset risk model separates common factor risk from asset-specific risk, informing portfolio construction and exposure analysis.
Portfolio
Portfolio research examines how forecasts translate into long and short positions, accounting for volatility, leverage, factor exposures, concentration and drawdown risk.
The thinking behind the process
Investment decisions, methods and the work behind them.
Define the investment problem
Mandate and opportunity set
- Systematic macro and futures
- Directional and relative-value opportunities
- Investment instruments and research predictors
Our research focuses on liquid futures across equity indices, rates, currencies and commodities. We investigate directional opportunities alongside relationships between markets and along futures curves. The investment universe and the information set serve different purposes: an instrument may help explain a market without becoming a portfolio holding. Defining the tradeable exposure, its economic rationale and its implementation requirements establishes the starting point for each study.
Make the hypothesis testable
Research discipline
- Economic hypotheses and target choice
- Information timing and reproducibility
- Validation and selection bias
Research begins with a question about market behaviour, a defined target and an explicit information set. Our platform supports repeatable experiments across periods, markets and model choices, with attention to when inputs become available and when outcomes can be observed. Results are assessed for sensitivity to assumptions and selection decisions. A useful study explains where an idea works, where it breaks down and which questions remain unresolved.
Evaluate the information in a signal
From models to forecasts
- Feature groups and model diversity
- Forecast horizons, calibration and efficacy
- Combination, concentration and uncertainty
Different feature groups, model families and training windows offer different views of the same market. We examine their forecasts across holding horizons, looking at direction, magnitude, stability and realised outcomes. The research then considers how to combine forecasts, how much concentration to accept and how uncertainty should affect conviction. Predictive statistics provide one view; the connection to a clearly defined position and holding period provides another.
Connect expected returns to tradeable exposures
From forecasts to portfolios
- Return and exposure consistency
- Risk budgets and portfolio construction
- Costs, turnover and position granularity
A forecast becomes economically useful through the positions it can support. Portfolio research brings expected returns together with factor exposures, covariance estimates and risk budgets. Particular attention goes to consistency between the return being forecast and the instruments held, including spread hedge weights and contract sizing. Costs, turnover, liquidity and capital requirements then determine whether an attractive statistical result can support a practical portfolio decision.
Set the conditions for committing capital
Deployment and oversight
- Research acceptance and operational readiness
- Monitoring, attribution and review
- Capital limits and change control
Deployment requires a reviewed research case and an operational process that can be observed and controlled. Acceptance criteria should cover data quality, reproducibility, implementation assumptions and the behaviour of the portfolio under stress. Capital limits, monitoring and attribution provide a basis for subsequent review. Changes to models or operating rules should be traceable, with a clear distinction between an exploratory result, an approved release and a live investment decision.