| 2026-10-03 17:57:34 | attribution_sign_repair | Rule attribution signs repaired through 2026-10-02: 767 short-position rows negated, 2 flat-position rows set to NULL. |
| 2026-05-25 00:00:00 | methodology | Universe expansion + drift cleanup. Traded universe grew from 265 to 269 instruments by promoting four candidates whose roll calendars were restored after a stale-instrument audit: ALUMINIUM_LME (LME aluminium, the third LME base metal after COPPER_LME and ZINC_LME), CAD2 (Canadian 2y bond, extending the CAD curve), GILT5 (UK 5y bond, extending the GILT curve), and MSCIEMASIA (Asia-only EM equity subset, distinct from broader MSCIEM). Same day, five permanently dead or duplicate instruments were silenced at the sampler layer to stop noisy CRITICAL roll alerts: SP500-GROWTH and SP500-VALUE (CME-delisted ~2002), BRENT-LAST (duplicate of BRENT), BB3M (Bloomberg discontinued BSBY 2023), and WATER-CALI (NQH₂O market essentially untraded — 0–5 nonzero-volume days per contract). Two more duplicates (EUROFIRST100, NIKKEI_large) were added to the sampler skip list because rob_system already trades the equivalent broad indices. Strategy logic, vol target, forecast cap, and capital — all unchanged. Position sizing scales proportionally under vol targeting, so adding instruments spreads risk further; no individual position grows. Triggered by a ContractNotFound crash in the daily contract-sampling cron that surfaced 8 stale instruments simultaneously; investigation distinguished structurally blocked from cleanly recoverable. URANIUM remains in the queue — its 2021–2023 contract history has a 2+ year gap that blocks roll-calendar formation; deferred pending a data backfill. |
| 2026-05-10 06:51:56 | attribution_cutover | Rule-level attribution unrecoverable for the full backfill window; all rows in the window carry NULL rule_contributions. |
| 2025-01-15 00:00:00 | methodology | Universe expansion + capital increase. Traded universe grew from 80 to 200 instruments; account size stepped from the $50k–$150k range to the $150k–$300k range (personal capital). Position sizing scales proportionally under volatility targeting, so strategy behaviour was unchanged; only absolute position sizes increased. Strategy logic, risk targets, and all other parameters remained the same. The exact date is approximate — lost in an accidental git commit squash during a repo cleanup. |
| 2025-01-15 00:00:00 | methodology | Universe expansion + capital increase. Traded universe grew from 80 to 200 instruments; account size stepped from the USD 50k–150k range to the USD 150k–300k range (personal capital). Position sizing scales proportionally under volatility targeting, so strategy behaviour was unchanged; only absolute position sizes increased. Strategy logic, risk targets, and all other parameters remained the same. The exact date is approximate — lost in an accidental git commit squash during a repo cleanup. |
| 2024-11-01 00:00:00 | methodology | Live trading begins. Systematic trend-following on 80 futures instruments using pysystemtrade (rob_system), 25% annualised volatility target, personal capital only. |