Quantitative research in the foreign exchange markets
Overview
This section is a living record of my work on a practical time-series problem: how to design and validate systematic decision rules for trading under uncertainty. The emphasis is research discipline and engineering, not “holy grails” or hindsight narratives.
Note: This is research documentation, not financial advice, not a recommendation to trade, and not an invitation to invest.
Current focus
My research programme is structured as a two-stage funnel:
Stage 1 — Stage gating (robustness first)
The goal is to identify simple, interpretable veto rules that avoid low-quality conditions and improve conditional outcome probabilities. This is intentionally conservative: avoiding bad conditions is often more robust than trying to predict perfect entries.
Stage 2 — Microstructure entry refinement (only after Stage 1 is stable)
Once gating criteria are stable out-of-sample, I test whether tick-level microstructure variables can improve entry timing and trade quality without increasing adverse excursion.
How I validate (summary)
To reduce self-deception in time-series research, I use:
- Purged + embargoed walk-forward splits (train on past, test on future with leakage controls)
- Block bootstrap uncertainty (confidence bounds that respect autocorrelation)
- Multiple-testing discipline (so “best of 500 rules” doesn’t masquerade as a discovery)
- Strict separation of decision-time features from anything that uses future information
Where helpful, I use lightweight ML (e.g., shallow decision trees and regularised logistic regression) as an interpretability-first tool to generate candidate gates and quantify relationships—then promote only what survives purged walk-forward and bootstrap validation.
Where beneficial, I include concise mathematical notation, for example this event label:
y_i = \mathbf{1}{\text{hit } +X \text{ before } -Y \text{ within horizon } H}Research stack
I’m building an end-to-end, telemetry-driven workflow so experiments are reproducible and auditable:
- Strategy execution + telemetry capture
- Market data (bars and ticks)
- Dataset construction (labels, features, join logic)
- Validation harness (walk-forward + bootstrap + promotion rules)
- Reporting artifacts (notes, charts, CSV summaries)
This is not presented as a product. It’s the working system I use to iterate quickly and publish findings responsibly.
White papers
White Paper — Part 1
Stage Gating for Robust FX Strategy Research: A Purged Walk-Forward + Bootstrap Framework
Status: Published (living document)
- Read online: Analytics white-paper
- Download PDF: https://lucitech.co.uk/wp-content/uploads/2026/02/lucitech_quant_research_part1.pdf
- Version: 1.0
- Last updated: 3rd February 2026
Research notes
Short updates tied to completed experiments, written as: “what I tested / what held up / what failed / what’s next”.
- (Coming soon)
Feedback
I welcome feedback that is specific and technical, especially around:
- leakage controls and decision-time feature validity
- validation design and multiple-testing discipline
- cost/execution assumptions and failure modes
To keep this sustainable, I can’t respond to open-ended trading questions. If you’d like to leave feedback, please include one concrete critique or suggestion (a paragraph is perfect).
Contact: research@lucitech.co.uk
Disclaimer
This material is provided for information and discussion only. It is not financial advice, not a recommendation, and not an invitation to invest. Past or simulated results do not guarantee future outcomes. Markets can change without warning, and execution costs materially affect results.