Hong Kong Metal-Miner Idiosyncratic Up-Shock Continuation: buy an HKEX gold/base-metal miner at the next HK open after a session whose peer-beta-adjusted return rose >=2.5 sigma ABOVE its miner peer basket, fixed 10-session open-to-open clock (HKEQ 1-DAY, 9-name pre-registered panel, long/flat cash, 8% per leg, max 6 concurrent)

Original Research Lead hypothesis (not seeded). Venue HKEQ (Hong Kong main board, CASH account, long-only), which has almost no history in the corpus (a handful of large-cap basket tests, none on miners, none event-driven on a commodity-homogeneous panel). MECHANISM: Hong Kong-listed Chinese metal miners (Zijin, Zhaojin, Shandong Gold, China Gold Intl, CMOC, MMG, Jiangxi Copper, Chalco, Hongqiao) share one dominant driver: metal prices and mainland industrial policy. A session in which ONE of them rises far more than the rest of the complex carries stock-specific news (resource/reserve upgrade, acquisition, output guidance, policy approval, southbound Stock Connect buying, index inclusion). In the HK H-share market this information diffuses slowly: mainland investors arrive through Stock Connect over several days, local coverage is thin, and the foreign holder base reprices late. So the idiosyncratic move CONTINUES rather than reverting. This is the opposite sign from the promoted TSX/ASX miner down-shock REBOUND panels; it is a continuation (underreaction) bet, the same family as the Bybit idiosyncratic attention-shock continuation that passed optimization, and it trades only the long side because HKEQ is a cash account. SIGNAL (at each HK daily close, all inputs known at that close): for each panel name i, r_i = close-to-close log return; peer_i = equal-weight mean log return of the OTHER panel names that have a bar that session; beta_i = 120-session rolling OLS slope of r_i on peer_i, lagged one session; residual e_i = r_i - beta_i * peer_i; z_i = e_i / (60-session rolling std of e_i, lagged one session). An UP-SHOCK is z_i >= z_entry (baseline 2.5) AND r_i > 0. EXECUTION: market order decided at the shock session's close, filling at the NEXT HK session open (next_open_fill, supported on session venues). Exit by market order decided at the close of the 9th session after the entry session, filling at the open of the 10th session after entry: exactly 10 sessions open-to-open. No stop, no target: risk-control route (b), time exit plus capped sizing (8% of equity per leg, at most 6 concurrent legs = 48% gross, unlevered cash). A new up-shock on a name already held is ignored (no pyramiding, no clock reset); a shock that arrives when 6 legs are open is skipped (if several qualify the same day with fewer free slots, take the largest z first). PANEL (fixed, pre-registered, all have HKEQ daily bars; 1787 from its 2018-09-28 HK listing): gold 2899.HKEQ (Zijin, gold+copper), 1818.HKEQ (Zhaojin), 1787.HKEQ (Shandong Gold), 2099.HKEQ (China Gold Intl); base metals 3993.HKEQ (CMOC), 1208.HKEQ (MMG), 0358.HKEQ (Jiangxi Copper), 2600.HKEQ (Chalco), 1378.HKEQ (Hongqiao). 2259 (Zijin Gold Intl) excluded: listed 2025-09, too short. HOMOGENEITY (L269): median full-sample correlation of each name's daily return with its leave-one-out peer basket = 0.72 (the promoted TSX miner panel is the comparable narrow-commodity design; US oil-complex measured 0.81). PRE-STUDY (Research Lead, frozen rule, these exact names, 2018-01 to 2026-08, gross of costs, open-to-open; reported only to pre-register sample size and sign — performance is judged by the backtest). Looks taken, all disclosed: down-shock rebound 3/5/10 sessions x z 2.0/2.5 (all NEGATIVE: at z<=-2.5, 5 sessions, -1.41% raw, t -2.55 — idiosyncratic shocks continue in this panel, which is what motivated the long-side continuation test); then up-shock continuation 3/5/10 x 2.0/2.5: H3 z2.0 +0.35% (t1.26), H3 z2.5 +0.58% (t1.46), H5 z2.0 +0.57% (t1.53), H5 z2.5 +0.84% (t1.57), H10 z2.0 +1.34% (t2.58), H10 z2.5 +2.00% (t2.70). The baseline is the H10/z2.5 cell: 300 events on 261 distinct entry dates (~30 dates/year, well above L271's ~60-event floor), mean raw +2.00%, mean peer-excess (trade minus equal-weight peer basket's same 10-session open-to-open return) +1.07%, raw mean positive in 8 of 9 calendar years and for 7 of 9 names; max same-date events 4. Every cell of the up-shock grid is positive and the effect grows with the horizon and the threshold, so the baseline is not an isolated spike, but it IS the best of 6 continuation cells after 6 rebound cells (12 looks): the analyst should apply that multiplicity. SURVIVORSHIP DISCLOSURE: names come from today's HK universe file; all nine were large, listed HK miners in 2018 (1787 from its 2018-09 listing), and no miner that collapsed is in the catalog, so the long arm may carry selection bias. Control: the backtest report must give each trade's peer-excess return; if raw returns are positive but peer-excess <= 0, the continuation premise is NOT shown (it would be miner beta in the 2024-2025 gold bull market). COST ARITHMETIC: HKEQ charges ~0.11% per side statutory (stamp duty 0.10% each side + levies) plus a 4 bp half spread: ~0.30% round trip. Expected capture: about half to two-thirds of the pre-study gross +2.0% after skipped slots and decay = ~1.0-1.3% per trade, 3-4x the round trip.