Strategy architecture
Breaking a trading idea into explicit signal, filter, sizing, exit, state, and risk modules so each assumption can be examined.
Research
Quantitative research built to challenge the system.
Our work starts with falsifiable questions and explicit constraints. A strategy is useful only to the extent that its behaviour, dependencies, and failure modes can be understood.
01 / RESEARCH POSITION
Explainable before impressive.
We do not treat a backtest as a product demonstration. It is an experiment produced by a dataset, a simulator, a specification, and a set of assumptions.
That distinction shapes the work: define the mechanism, isolate the moving parts, test the implementation, search for fragility, and document what remains uncertain.
02 / AREAS
Research across the full system.
Current work is centred on algorithmic trading and MetaTrader 5, with an architecture intended to support broader instruments and environments over time.
Market behaviour
Studying how spread, volatility, liquidity conditions, session structure, and price-path dependence affect system behaviour.
Robustness
Looking beyond a preferred configuration through sensitivity checks, alternative periods, and deliberately adverse assumptions.
Execution mechanics
Treating order handling, fills, stops, costs, and platform behaviour as part of the strategy—not an implementation footnote.
AI-assisted research
Exploring where specialised models can organise evidence, support diagnostics, or surface hypotheses while keeping decisions reviewable.
Research tooling
Building utilities that reduce manual friction in data preparation, experiment setup, diagnostics, and result review.
03 / WORKING LOOP
A process designed to find problems.
- 01
Frame the question
State the proposed mechanism, expected conditions, decision rules, and what evidence would weaken the idea.
- 02
Audit the inputs
Review data provenance, coverage, gaps, timestamps, symbol settings, spread representation, and any transformations.
- 03
Build observable modules
Separate strategy logic, execution, position state, and risk controls so behaviour can be traced and tested.
- 04
Run staged tests
Use quick diagnostic runs first, then more demanding data and assumptions when the strategy requires them.
- 05
Interrogate the result
Examine sensitivity, concentration, path dependence, cost assumptions, edge cases, and failure periods—not only aggregate output.
- 06
Document the boundary
Record what was tested, what was not, what changed, and which conclusions the evidence does not justify.
04 / NEXT
Research is only as credible as its inputs.
See how we approach tick data, testing modes, execution assumptions, sensitivity, and reproducibility.
Data & methodology