7 SBFoundation as a Research Platform
Author: Todd B. Adams Companion to: PhD Research Proposal — Dynamic Heterogeneous Multiplex Networks for Factor-Based Systemic Risk & Market Phase Transitions Status: draft evidence package (supervisor-facing)
7.1 Thesis
The proposal argues that systemic risk is an emergent property of a multi-layered financial network, and that a spatial-temporal graph model can detect its phase transitions. Testing that claim credibly needs more than a notebook and a static academic panel — it needs a live, point-in-time-correct, survivorship-adjusted, cost-aware, executable empirical instrument. SBFoundation is that instrument. It runs the full chain from raw vendor data to real paper-broker fills every night, it already implements the proposal’s stock–factor layer in production, it already carries correlation-network and crowding machinery, and it has already stress-tested the supervisor’s own published connectedness methodology on its universe. This folder is the evidence for those claims, mapped gate-by-gate to the proposal.
7.2 How to read this folder
The docs are numbered as a reading order. Each is skimmable (~1–2 pages) and cross-links into the deeper platform documentation rather than restating it.
| # | Doc | Answers | Reinforces proposal § |
|---|---|---|---|
| 00 | README (this file) | What is this and why does it exist? | §1 |
| 01 | Platform as research substrate | Why is a live E2E trading platform the right instrument for this research? | §1, §3 |
| 02 | The end-to-end pipeline | Does it really run data → live trading, reproducibly? | §3 |
| 03 | The multiplex data substrate | Can it source the proposal’s three network layers? | §5 |
| 04 | The factor layer as multiplex Layer 1 | Is the bipartite stock–factor layer hypothetical, or built? | §3, §6 |
| 05 | Systemic-risk & crowding groundwork | Is there prior, supervisor-aligned network work here? | §2, §3 |
| 06 | The integrated target architecture | Where does the Supra-Laplacian / ST-GNN engine mount? | §3 |
| 07 | Gap analysis & research roadmap | What is built vs. proposed, honestly? | §3, §5 |
| 08 | Validation & anti-overfitting | How are network claims protected from self-deception? | §6 |
| 09 | Reproducibility & experiment infrastructure | Is each experiment auditable and repeatable? | §3 |
| 10 | Literature bridge | Where do the proposal’s papers live in the codebase? | §2 |
7.3 What this package is engineered to demonstrate
- Reinforce the proposal — the research question already has platform-grounded precedent (docs 01, 04, 05, 10).
- The platform supports the research — substrate, mount points, honest gaps, and validation rigor (docs 03, 06, 07, 08, 09).
- End-to-end, data to live trading — a production chain terminating in real broker fills, not a simulator (docs 02, 09).
7.4 Status legend
Every forward-looking claim in this folder carries one of three tags, so nothing is over-claimed:
BUILT— running in production today; traceable to committed packages and nightly runs.PARTIAL— substrate exists and is adjacent, but the specific capability is incomplete.PROPOSED— the research extension; not yet implemented. This is the PhD.
7.5 Relationship to docs/phd/
The proposal and its bundled source papers stay in docs/phd/. This folder is the supporting evidence: read the proposal for the research argument, read here for the platform that makes it testable.