Converting stock-market patterns to investment strategies
Research innovation, technology and investment expertise work together
From an idea to a live strategy
Evidence accumulates through eight stages. Weak ideas exit early; only validated systems reach a portfolio
From an idea to a live strategy
Evidence accumulates at every stage. Weak ideas exit early; only validated systems reach a portfolio.
Economic hypothesis
A reason the relationship may persist
Stage 1 of 8Investment thesis to strategy
Define a thesis, test historical outcomes and investigate the fundamental reasons behind those outcomes
Define the economic relationship worth investigating.
Each step needs a different expertise
Research innovation begins the process; technology scales it; investment expertise validates and implements it
We use two frameworks to discover investment ideas
MAGIC searches transparent rules, while NEURO uses explainable machine learning for more complex relationships
Best suited to investment relationships that can be expressed clearly as transparent rules.
Rules-based MAGIC Framework
Human-understandable investment theses and rules are generated, tested and combined at scale
MAGIC finds human-understandable transparent rules, at scale
Advanced technology discovers, tests and combines investment rules across vast historical data
300,000+ ideas
A broad universe of transparent rules
Four pillars of MAGIC
Rule generation, filtering, fundamental validation and ensemble construction form one research architecture
1. Rule engine
Advanced automation works with human expertise and machine intelligence to search for promising strategies
right strategy
Generate → train → analyse → improve → validate
1. Rule engine: 300,000 rules created and tested over 25 years
Scale helps us explore many more investment relationships than a traditional manual process
1. Rule engine: cutting-edge technology and intelligence
Reinforcement learning, rule-space dispersion and human expertise guide the search process
Millions of portfolios, rolling-return data points, filters and computations.
Reinforcement learning and genetic-algorithm dispersion.
Investment logic guides what the system is allowed to search.
2. Analyse & filter: select the top 0.1% rules
Evaluate each rule from every angle—return, risk, distribution, probabilities, drawdowns and consistency
300,000+ ideas
A broad universe of transparent rules
Only robust ideas survive
The strongest rules combine statistical persistence with economic coherence. Return alone is never enough
Only robust ideas survive.
Rules are judged on statistical persistence and economic coherence—not return alone.
Promote to portfolio
Persistence score 71 · economic evidence 75 · successful windows 89%
3. WHY — do business fundamentals support the returns?
We examine profitability, growth, cash generation, financial strength and sensible valuations
Evidence must agree from three directions
Economic rationale, statistical reliability and fundamental validation must reinforce one another
Evidence must agree from three directions.
A strong historical result is not enough. The economic thesis, the statistical behaviour and the underlying businesses must reinforce one another.
4. Ensemble: combine several strong rules together
Complementary rules reduce dependence on one investment relationship
The strategy becomes less dependent on one factor, one market regime or one investment relationship.
That is what we call MAGIC
Multiple AI-ML Generated Investment Rules Combined
Explainable ML: NEURO — our next-generation machine-learning model
A more flexible framework for complex fundamental relationships while preserving explainability and research discipline
NEURO finds relationships that simple rules may miss
Explainable machine learning uncovers complex connections between fundamentals and future market outcomes
Designed to capture nonlinear interactions while retaining explainability and validation discipline.
This is how NEURO learns from financial data
Features are created, models are trained, tested out of sample and combined into reliable predictions
176 curated features
Fundamental, valuation, market and consistency features are engineered with economic intent.
Two frameworks. One research standard.
MAGIC and NEURO are judged through the same return, risk, fundamental and live-evidence disciplines
Evidence must agree before a strategy is trusted.
Transparent rules are selected only when statistical evidence, economic logic and fundamentals reinforce one another.
Technology is at the core of everything we do
Research, portfolio monitoring, compliance and client operations are strengthened through purpose-built automation and AI
300k rules · ML models · portfolio simulations
Rolling returns · drawdowns · fundamental validation
Rebalancing · monitoring · live diagnostics
Controls · records · review support
Research archive · learning paths · CRM