{
  "asof": "2026-10-08",
  "purpose": "100 stocks that move on SQ, NOT 100 historically profitable stocks. No live orders or production changes.",
  "universe": "Same collected current-listed domestic ordinary-share universe as sq_gap (1605 expected). Current-listing and collection survivorship bias remains; this is not the point-in-time entire exchange universe.",
  "data": "Use supplied split-adjusted long OHLC through each file end; append later master daily observations. Each event OHLC and previous-session close always use ONE source. Never splice price bases or rescale master. Daily prices can have fractional-yen truncation.",
  "calendar": "Monthly second Friday, previous exchange session if holiday, based on 1306 calendar cross-checked with7011. major=Mar/Jun/Sep/Dec. No incomplete2026-10-09 data.",
  "selection_window": "For EACH monthly SQ, rebuild100 stocks at the previous exchange session close, using the latest12 or6 COMPLETED monthly SQ dates. Ordinary non-SQ Fridays in those same prior months, known by cutoff. Need all12/all6 matched code/month pairs. Test from2020-09, when both windows are available. Test-day outcomes never enter its list.",
  "liquidity": "Median turnover over20 exact exchange sessions ending at cutoff >=500millionJPY. Long data use the actual 売買代金 field (thousandJPY); current master uses Close*Volume proxy. Missing or stale20-session coverage fails. This is not a bid/ask spread or auction capacity test.",
  "features": "SQ range%=100*(High-Low)/previousClose. Absolute opening gap%=abs(100*(Open/previousClose-1)). For each code take SQ median range, SQ median absolute gap, and SQ median range / median of same-month ordinary-Friday median ranges.",
  "score": "Equal-weight mean of cross-sectional percentile ranks of the three features among liquid eligible stocks. Larger is better. Ties: SQ median range descending then code ascending. Predeclared once; no PF, win rate, trade direction, optimized weights or industry quota.",
  "lists": "Rolling12m and6m monthly lists are the primary test. Latest example lists use information through2026-10-08 for2026-10-09 (not yet scored). July/September2026 lists are saved separately. Do not relabel the October9 lists as November lists; refresh after new data arrives.",
  "evaluation": "Movement, ordinary-Friday controls, major/minor and GD0.5/1/1.5/2. Same-day actual Open to Close, equal notional, zero commission and adverse0.05% each side. No TP/SL/CFD gate or11:30 exit. Exact auction entry selected using actualOpen is an idealization, not a preopen fill guarantee.",
  "robustness": "Compare all available stocks and each window liquidity-eligible pool with movement100 on identical SQ dates. Ordinary-Friday controls are Fridays AFTER the test SQ in that same month. Pooling weights dates with more trades more heavily; report date-equal mean, best3 date exclusion and adverse excursion. Year and major/minor splits are exploratory, not pristine prospective OOS.",
  "precision": "Master daily-derived observations get a1JPY uncertainty per OHLC price. Report adverse return bound using (Close-1)/(Open+1) and require (Open+1)/(previousClose-1) at GD threshold for conservative sensitivity. Bounds reflect input quantization only, not execution risk."
}