EvidenceHunter

An AI-assisted market observation and evidence evaluation framework built around research-grade practices

EvidenceHunter is also an experiment in AI-native software development: can someone without a traditional software engineering background use AI coding agents to independently build, verify, document, and maintain a complex research software system?

EvidenceHunter focuses on systematic market observation, evidence collection, provenance, validation, reproducibility, and explicit research governance.

It is not designed as an automated trading system or financial advice tool.

---

Why EvidenceHunter Exists

This project began with two connected questions.

Research question

Can market observations be collected and evaluated through a workflow that keeps evidence, assumptions, provenance, and conclusions clearly separated?

Engineering question

Can a person without formal software engineering training use modern AI coding agents as engineering collaborators while still maintaining:

• automated testing
• reproducible workflows
• explicit governance
• dataset lifecycle controls
• documented research decisions
• version-controlled development
• continuous integration

EvidenceHunter is an attempt to explore both questions through a real, continuously maintained software project.

---

What EvidenceHunter Does

The current framework includes components for:

• market data collection
• order book and order flow observation
• dataset management
• gap and completeness tracking
• archive materialization and verification
• feature construction
• shadow outcome analysis
• discovery experiments
• provenance tracking
• research readiness checks
• governance and implementation-state tracking

The emphasis is not on producing confident predictions.

The emphasis is on making the path from observation → evidence → evaluation → conclusion inspectable.

---

What EvidenceHunter Is Not

EvidenceHunter is not:

• a guaranteed-profit system
• a price prediction service
• financial advice
• an autonomous trading bot
• a substitute for independent research or risk assessment

---

AI-Assisted Development

AI coding agents have been used extensively throughout the development of EvidenceHunter.

AI is treated as an engineering collaborator rather than an unquestioned authority.

The development process emphasizes:

• evidence before claims
• verification before declaring work complete
• tests after meaningful code changes
• explicit scope and implementation records
• documented failures and corrections
• human-controlled decisions
• iterative review of AI-generated work

The repository intentionally preserves governance artifacts and research records because the development process itself is part of the experiment.

See:

docs/AI_DEVELOPMENT_STORY.md

docs/ARCHITECTURE.md

docs/RESEARCH_METHOD.md

AGENTS.md

---

Evidence of Engineering Discipline

The public repository includes:

Automated tests

The tests/ directory contains regression, integration, lifecycle, collector, archive, dataset, and research-workflow tests.

Continuous integration

GitHub Actions automatically runs the Python test suite on changes to the repository.

The current public baseline passes CI on Windows.

Research governance

research/governance/ contains artifacts for:

• scope declarations
• implementation handoffs
• change roles
• readiness reports
• test results
• reconciliation
• execution contracts

Dataset lifecycle controls

dataset_lifecycle.json and related modules track dataset state and help prevent accidental reuse or mutation of frozen research datasets.

Provenance and reproducibility

The repository includes provenance, summary, readiness, continuity, and cross-platform verification artifacts for research executions.

---

Architecture

EvidenceHunter is organized around several layers.

Collection Layer

Collects and records structured market observations.

Evidence Layer

Preserves observations, provenance, completeness information, and dataset state.

Evaluation Layer

Supports feature construction, shadow outcomes, discovery experiments, and structured comparisons.

Governance Layer

Tracks scope, implementation decisions, readiness, testing, and research-state transitions.

More detail is available in:

docs/ARCHITECTURE.md

---

Repository Structure

EvidenceHunter/
│
├── EvidenceHunter_*.py
├── audit_order.py
├── governance_guard.py
├── dataset_lifecycle.json
│
├── tests/
│
├── research/
│   └── governance/
│
├── research_design_next/
│
├── docs/
│   ├── AI_DEVELOPMENT_STORY.md
│   ├── ARCHITECTURE.md
│   └── RESEARCH_METHOD.md
│
├── AGENTS.md
├── CONTRIBUTING.md
├── requirements-collector.txt
└── README.md