Historical research evidence, 22 September 2026.

600 mechanical runs across 27 strategies; three separate monthly family analyses. No verified profitability claim. Native source prices remain in the local data-v2 archive: the public package supplies their SHA256 manifest, native SAR buffers, research code and simulated trade logs. Re-running requires numpy, tzdata and the matching source CSVs as described by the manifest. Monthly research additionally requires xlrd and the MIT workbook linked in portfolio-data-audit.json.

The Python SAR approximation is not used for ID9; timestamp-matched native MT5 iSAR buffers override it. Current swap assumptions are not historical swap data. Consult protocol.json and all per-row coverage flags.
