Original Research Lead hypothesis (not seeded). A local-market calendar flow trade on the Tokyo Stock Exchange, a venue the factory has used 4 times out of 5,001 experiments. MECHANISM: Japanese turn-of-the-month demand starts EARLIER than the US/Canadian window because Japanese salaries and bonuses are paid on/around the 25th, and the largest retail accumulation channels (NISA tsumitate monthly installment plans, iDeCo and corporate DC pension contributions, monthly-plan investment trusts) are debited from payroll accounts at month end and invested in index/large-cap trusts over the last sessions of the month and the first sessions of the next. Ziemba (1991, 'Japanese security market regularities', Japan and the World Economy) documented that the Japanese turn-of-the-month return cluster spans trading days -5 to +2, not the US -1 to +3. Our promoted calendar survivors (SPY ToM 858fbc5b, EM country-ETF ToM 98089727, TSX 16-name ToM de78ff1c: 1,664 trades, PF 1.35, Sharpe 0.57) are all long/flat fixed-session windows (lesson L263); the developed-ex-US ETF version traded from New York failed (76e5d391), which is why this test trades the LOCAL shares on the LOCAL exchange, open-to-open in Tokyo, with the Japan-specific window shape. Universe (pre-registered, fixed, 16 TOPIX Core/Large70 names across banks, trading houses, autos, telecom, pharma, insurance, industrials, all with board-lot costs well under the per-name allocation): 8306 MUFG, 8316 SMFG, 8411 Mizuho, 8766 Tokio Marine, 8058 Mitsubishi Corp, 8031 Mitsui & Co, 8053 Sumitomo Corp, 7203 Toyota, 7267 Honda, 6752 Panasonic, 6503 Mitsubishi Electric, 7011 Mitsubishi Heavy, 9432 NTT, 9434 SoftBank Corp, 2914 Japan Tobacco, 4502 Takeda. A name enters the basket only once is_listed() and it has >= 60 sessions of history (point-in-time; 9434 listed Dec 2018). SURVIVORSHIP: these are today's large caps; because the strategy is exposed only ~30% of sessions, any survivorship drift is proportional to exposure. The analyst should therefore compare the window return against the same basket's mean return over all non-window 6-session blocks (the pre-registered comparator), not only against zero. EXECUTION: JPEQ is a session venue; decisions are made on the daily close and fill at the NEXT session's open (next_open_fill, supported). Month boundaries and session counts come from the platform exchange calendar (src.strategies.helpers.sessions / session_calendar capability, supported); nothing is hard-coded. Board lot 100 (size with self.lot_size(), round down). Account and metrics in JPY. Distinct calendar periods over 2018-01..2026-10: ~105 month windows (this, not the ~1,600 per-name trades, is the evidence count per L238/L263).