# Source Map — Full Technical Manuscript

| Source | Used for | Epistemic boundary |
|---|---|---|
| `../source-map.md` | Original cited sources and claim limits | Inherits its boundaries |
| `../html/index.html` | Visual companion terminology only | Teaching artifact, not evidence |
| `docs/architecture/btc-polymarket-5m-prediction-architecture.html` | Component separation and system flow | Design target |
| `docs/planning/btc-polymarket-5m-15m-research-backtest-replay-prompt.md` | Data, replay, tuning, and safety gates | Design target |
| `docs/planning/btc-polymarket-5m-15m-research-backtest-replay-redteam.md` | Known risk controls and audit questions | Design target |
| Polymarket API and market-data documentation, cited in `../source-map.md` | Market metadata and CLOB concepts | External documentation; re-check before use |
| CatBoost, DeepLOB, SimLOB, TLOB, and LiT references, cited in `../source-map.md` | Model concepts | Literature-derived; no production-performance claim |
| Python and scikit-learn official documentation | Illustrative Python interfaces and metric definitions | External documentation; pin versions in an implementation |

## Evidence rules

No notebook, database, model artifact, or historical data corpus is currently the source of truth for measured trading claims. Consequently, worked numerical examples use hand-labelled **Illustrative** values. A future implementation chapter may replace those examples only with a dated dataset manifest, a reproducible command, a configuration hash, and an explicit limitation.