Four disciplines, continuously validated.
Our framework treats every signal as a hypothesis. Research validates it with out-of-sample evidence; construction combines it with orthogonal siblings; execution sizes it against regime; monitoring tests whether live behaviour still matches its ex-ante distribution. Our framework treats every signal as a hypothesis. Research validates it with evidence; construction combines it with orthogonal siblings; execution sizes it against regime; monitoring tests whether live behaviour still matches expectation.
- 01
Research & Signal Generation Signal Hypothesis
Quantitative researchers develop and validate systematic signals using rigorous statistical methods, out-of-sample testing, and walk-forward analysis. Every idea starts as a falsifiable research question with explicit data, costs, and failure criteria.
- 02
Portfolio Construction Out-of-Sample Evidence
Signals are combined using mathematical optimisation frameworks designed to maximise risk-adjusted returns while controlling for correlation and tail risk. Signals must survive walk-forward testing, robustness checks, and implementation assumptions.
- 03
Execution & Risk Control Portfolio Interaction
Adaptive position sizing and regime-aware exposure management are embedded at the core of every program. Signals must be statistically independent of broader market direction. Approved signals are studied against siblings, capacity, correlation, and drawdown contribution.
- 04
Continuous Validation Live Degradation
All programs are subject to continuous live performance monitoring, statistical hypothesis testing, and systematic review against out-of-sample benchmarks. Research remains under review after launch; live behaviour is compared to the ex-ante distribution.
- Orthogonality, not correlation, is the alpha question Two signals can be uncorrelated on average and still load on the same risk in the worst week.
- Capacity is a risk control We treat capacity as a risk parameter rather than a marketing one. The number cannot be optimised away.
- Where signals decay first Decay is uneven. The first place we see it is execution cost, not raw forecast accuracy.
- Three tells of an overfit backtest Three patterns we look for before a backtest is allowed into the staging environment.