Frozen-Rule Breadth Transfer of the Absorbed-Flush Rule to Hyperliquid: buy a Hyperliquid alt perp at the close of a down day whose range blew out (>=2x its 60-day return std) but that closed in the upper 45% of its range, fixed 3-day clock, 10% per leg, max 6 concurrent, on 15 HL perps disjoint from the Bybit panel (1-DAY, long-only)

Original Research Lead hypothesis, written as a breadth replication. It is NOT a new parameter search. MECHANISM (same as the promoted Bybit 'Intraday Flush-Absorption Panel': 200 trades, PF 1.72, 1.92% per trade): on a daily bar a leveraged perp makes an abnormally deep intraday excursion. The high-low range, measured against the prior close, is at least range_mult times the trailing 60-day std of daily log returns. A move that size is mostly stops and long liquidations being swept. The bar still closes BELOW the prior close, but in the UPPER part of its own range: close location (C-L)/(H-L) > close_loc. Resting demand absorbed the forced selling inside the same day, and the marginal forced seller is gone. Expect a multi-day rebound. Lesson L258 asked for this rule (a), absorption confirmed on the bar and a multi-day clock, to be tested against rule (b), buy a bottom-of-range close and hold 1 day. Rule (b) was the frozen Hyperliquid capitulation-close transfer, which ended underpowered: 425 trades, 0.42% per trade. Rule (a) has never been run on Hyperliquid. This hypothesis is that test, on a venue where long-side shock panels have survived (L242/L250: the HL daily volume-confirmed absorption panel, the HL 4H capitulation panel and the HL BTC-beta-hedged idiosyncratic down-shock panel were all promoted). The Bybit OKX-style transfer failures (L242) motivated choosing HL rather than OKX. UNIVERSE (fixed before measurement, disjoint from the Bybit 13 names, from the factory's BTC/ETH/SOL/BNB/XRP/ADA/LINK core and from the L250 list of over-mined HL mid-caps): ARB, OP, SUI, LDO, CRV, SNX, RUNE, DYDX, GMX, STX, WLD, BLUR, kPEPE, SEI, TIA, each as <SYM>USD.HYPERLIQUID 1-DAY. Point in time: a leg is eligible only after >= 70 daily bars of its own history (universe_at()/is_listed(), parameters.point_in_time_universe = true). A leg that stops trading is closed at its last price by the base template. Residual survivorship: these are perps that still trade on Hyperliquid today; HL serves no delisted perps. DATA NOTE: the catalog's HL 1-DAY history for several of these names (CRV, SNX, RUNE, DYDX, OP, LDO) starts in 2020-2022, before they traded on Hyperliquid's current venue. The backtest uses the catalog as it is. 166 of the 200 pre-study events fall in 2023 or later. EXECUTION SUBSTITUTION (capability next_bar_open_fill = use_equivalent): the market order is sent at the signal bar's close (00:00 UTC daily close). On a 24/7 venue that price is the next bar's open. No stops: risk control route (b), a fixed time exit plus capped unlevered sizing. PRE-REGISTERED MEASUREMENT (Research Lead, catalog HL 1-DAY bars Nov-2020 to Oct-2026). Fixed rule, non-overlapping per coin, entry at the signal close, exit at the close 3 days later, 0.09% HL round trip deducted, funding NOT deducted, no concurrency cap. Baseline range_mult 2.0 / close_loc 0.55 / hold 3: 200 events on 96 distinct entry dates. Mean net forward return +3.64% per event, win rate 0.58, date-clustered t 2.74. Bybit's frozen values range_mult 2.0 / close_loc 0.60 / hold 3: 143 events, 75 dates, +2.06%, t 2.16. In the 27-cell grid range_mult {2.0,2.5,3.0} x close_loc {0.5,0.6,0.7} x hold {2,3,5}, every cell has a positive mean (+1.08% to +7.49%), with date-clustered t from 0.74 to 2.63. Cells with close_loc 0.7 are thin (12-59 events). By year at the baseline: 2021 +2.3% (20 events), 2022 +2.2% (12), 2023 +2.7% (37), 2024 +3.2% (72), 2025 +11.2% (28), 2026 -0.02% (29). 2026 is flat, and Bybit's 2026 was also negative (-2.2%): the recent edge is weaker, and the 2025 mean is driven by a few large rebounds. Per name, 10 of 15 means are positive; BLUR, RUNE, LDO, STX and kPEPE are negative on 4-14 events each. These figures are context only; the full backtest and holdout judge the idea. Evidence counts per L237/L259: 96 distinct event dates in the overlap where every leg has bars (no other feed needed), well above the ~30 floor. Counts exclude the concurrency cap, which can only drop events.