Weekly report — work week 2 combined (2026-W33 + 2026-W34)

Weekly report — work week 2 combined (2026-W33 + 2026-W34)

• Period: 2026-08-10 ~ 2026-08-23
• Stage: stage 1 research and backtesting kickoff
• Status: in progress

On the site, the experiments, verdicts and research logs created in 2026-W34 are also shown on this work-week-2 page. Each record's actual creation date and permanent URL are preserved as they are.

Combined follow-up records

• Added the full-period EXP-003 C4 baseline backtest and its rejected verdict.
• Identified consistency problems among cost, target price and holding period, and interpreted the C4 hypothesis itself separately from the current execution contract.
• The trade and signal ledgers and the 2026-08-20 research log are linked from the same page.

This week's goals

1. Fix the common contract and report structure of the strategy research repository.
2. Finalize the measurement items and pass criteria for CREON sample collection.
3. Separate the validation scopes of bnf_daily, bnf_intraday and orb.

Work done

• Organized the common validation philosophy and folder boundaries based on the README.
• Wrote the decision log and the data, execution, experiment and report policies.
• Registered three strategies in the strategy catalog as proposal.
• Implemented the CREON continuation-query adapter, a sample-collection CLI, and CSV, manifest and EXP-001 report output.
• Measured the longest CREON daily, 1-minute and 5-minute histories for Samsung Electronics (A005930) and generated the EXP-001 coverage report containing the source hashes.
• Implemented a reusable daily backtest kernel and bnf_daily_c4_baseline, and ran EXP-002.
• Implemented the symbol master, symbol history, corporate-action ledger and date-based universe contract for all-symbol expansion, and the REF-001 audit script.

Strategy, data and experiment status

Performance metrics

Per EXP-002, the C4-only baseline is a negative result. It is a single symbol and a recent period, without C1~C3 or regimes, so it is not interpreted as a profitability or failure verdict on full BNF.

Risks and decision requests

• A long-range data correction/exclusion policy is needed for the 2 OHLC-relationship anomalies in 1984 daily bars.
• BNF's C1~C3 formulas, sector classification and exact definitions of stop, take-profit and time exit are needed.
• ORB's catalyst filter and intraday entry/exit rules must be received as a separate specification.

Completion criteria for next week

1. Collect the KRX symbol master, delisting history and DART corporate-code mapping to generate REF-001.
2. Build the all-symbol daily collection queue, then accumulate per-symbol manifests.
3. Complete the comparison of the corporate-action ledger against original filings and effective dates, and the verdict on the 2 long-range OHLC anomalies.