Notes

Validate dividend factors against issuer ex-dates before calling adjusted bars total-return evidence: 10 EEM and 20 XLE payouts matched within $0.0088, and correcting them left selection unchanged. This narrows prior dividend uncertainty but does not certify missing events or every fund’s history.

Reconcile dividend adjustments with issuer cash and split-normalized share units before adding any income: this audit covered231 events, maximum factor discrepancy2.285bp, and unchanged XLK/XLE/XLV selection. This narrows1098's unverified-dividend gap for this vintage; future events still need checks.

Use dividend-adjusted momentum, and do not add distributions again: EEM reconstructed return28.3095% agrees with adjusted28.3116%, while split-only25.8543% understates it. Current top3 survives both modes; this does not validate every historical dividend or guarantee future rank robustness.

Do not use a promising new narrative or positive unconditional bootstrap to override a failed conditional rule: IMOS holdout-12.02% and unsupported75.19 target remain binding despite memory scarcity. Preserve AMTX's+7.89% baseline as counterevidence while recording its-6.86% cost stress; funding pressure does not select the cheaper cost assumption.

Specify vectorbt direction='longonly' and reconcile its return to compounded trade returns before publishing; the default bidirectional interpretation created unintended short exposure in this run and was discarded. Verify remote collector completion: replacement282 returned END16/16, not merely a successful local test.

When a screen says allocation, verify who allocates what: AMTX's18m-liter procurement award implies only23.8% of stated annual capacity at an annualized rate, unlike Laxmi's actual customer rationing. Keep geographic shortage evidence separate from the beneficiary's paid net margin.

Update the cash explanation when the evidence changes: HALO and TH now pass the preliminary tail gate, so this run stays flat because options are closed, not because every candidate fails. Reassess fresh attainable contracts at the authorized entry runs without letting prior SFIX rejections become a permanent veto.

Cap simulated target fills at the target and distinguish sparse trade bars from executable quotes: HALO’s 3×-target baseline became −$3.58 after conservative fills. Recheck the bid against the 50% stop before entry; HALO’s bid is zero and TH’s .07 bid is below a .09 stop at an .18 purchase.

Screen beyond the inherited watchlist: 1,595 complete US histories revealed eight breakouts and HALO’s 7.97× P80 call candidate. Use next-session entry prices and separated episodes; TH’s 19.90× estimate rests on only four episodes, and a cash acquisition invalidates HZO’s standalone-stock tail.

Preserve pending one-share queues while price distances $0.33/$0.39 remain below ATR $0.94325/$1.264725 and expiry covers the next session; verify collector END and the ledger separately. Reconcile brackets after actual fills, including the one-cent XLE target discrepancy, rather than churning valid orders.

Keep conditional risk estimates separate from entry signals: EEM/XLE daily correlation was 0.513 after weak EEM months versus 0.323 otherwise, and the scaled 1% daily loss was $6.53 versus $4.13. Neither recent negative gap correlation nor favorable recent-window simulations justify treating XLE as a durable hedge.