Binance Spot-Listing Sell-the-News on Perp-First Tokens: Short the USD-M Perp After Its Delayed BINANCE_SPOT Listing, BTC-Hedged, Fixed-Clock Exit (65 Point-in-Time Events, 4H)

Event-driven, market-neutral short that is flat about 85% of the time. Since 2024 many tokens (pre-market and perp-first launches) have had a Binance USD-M perpetual listed days to months before Binance listed the spot pair. The spot-listing announcement draws retail anticipation buying, and perp longs position for 'Binance spot demand'. Once spot trading opens, that demand has been realised. Early holders and airdrop recipients then have a deep cash venue to sell into, and the premium unwinds (classic buy-the-rumour/sell-the-news). The test: at each such spot listing, short the token's USD-M perp, hedge it with an equal-notional BTCUSDT perp long, and hold for a fixed number of days. Universe (point in time, from the platform listing calendar, delisted symbols included): every BINANCE USD-M USDT perp whose BINANCE_SPOT USDT pair was first listed between 3 and 365 days AFTER the perp's own listing, and whose perp was still listed at spot-listing time. That gives 65 events (2020:1, 2021:1, 2023:1, 2024:5, 2025:42, 2026:15) on the 65 perps in `instruments`. The BINANCE_SPOT legs are NOT traded. They are listed only so the strategy can read each spot pair's listing time via listed_since()/is_listed(); subscribe them at 1-DAY to keep the load small. The strategy may instead use the arrival of the first spot bar, but never a future timestamp. Set parameters['point_in_time_universe']=true so later-listed legs do not clamp the window. Traded legs: the 65 perps (4-HOUR bars) plus BTCUSDT.BINANCE (4-HOUR) as the hedge. Execution: market orders at the 4H bar close, the platform's equivalent of a next-bar-open fill on a 24/7 venue. The protective stop is an intrabar reduce-only stop-market on the short perp leg, matched by the platform; no order needs to rest on the exchange. Fee arithmetic: each event round-trips two legs at about 0.10% each, about 0.20% of single-leg notional. Funding comes on top: new-token perps often have negative funding, which the short pays; budget about 0.05%/day x 7 days = 0.35%. Expected capture: after the spot listing, perp-first tokens typically drift several percent to tens of percent over 1-3 weeks. Even a conservative mean hedged move of -4% over 7 days nets about 3.4% per event after costs, more than 10x the fee bar. No prior factory run tests the spot-listing event. Earlier listing-decay tests shorted from the PERP listing date on Binance and Bybit; this one keys on a later, scheduled event and only on perp-first tokens.

Event-driven, BTC-hedged short of Binance perp-first tokens after their delayed BINANCE_SPOT listing. Universe: the 65 BINANCE USD-M USDT perps whose BINANCE_SPOT USDT pair first listed 3-365 days after the perp (perp still listed at that time), chosen point in time from the listing calendar (delisted TONUSDT included). On each perp's own 4H bar the strategy reads the spot pair's listing instant with listed_since() (never a future one; in paper/live, where no calendar exists, the arrival of the first spot 1D bar) and the perp's listing instant, checks the 3-365 day lag, and at the first 4H close at or after spot listing + entry_delay_hours (at most max_entry_slip_hours late, so events before the run starts are not chased) sells the perp with notional_per_event (8%) of equity x leverage and buys an equal hedge_ratio notional of BTCUSDT perp. Optional filter: the perp's run-up from runup_lookback_days before spot listing to spot listing must be >= min_prelisting_runup_pct (baseline -100 = off). A reduce-only intrabar stop-market sits stop_pct (25%) above the short's fill price (8% x 25% = 2% of equity at the stop). Exit after hold_days on a fixed clock from entry, or at the stop / a delisting; the BTC hedge is reduced by that event's fixed hedge quantity. Max 5 concurrent events (gross <= 80% of equity). Execution: market orders at the 4H bar close, the platform's equivalent of a next-bar-open fill on a 24/7 venue. BINANCE_SPOT legs are subscribed at 1-DAY for their listing times only and never traded.

Implements the hypothesis's spot-listing sell-the-news event exactly: point-in-time event detection via listed_since() on the BINANCE_SPOT legs (parameters.point_in_time_universe=true so later-listed legs do not clamp the window), short perp + equal-notional BTC long hedge, fixed-clock hold_days exit, intrabar reduce-only stop (execution_capabilities intrabar_stop: supported; next_bar_open_fill: bar-close market order as the 24/7 equivalent). All research-plan names are implemented with the fixed values; tunables hold_days (baseline 7, the description's horizon), entry_delay_hours (0) and min_prelisting_runup_pct (-100 = filter off, since the description itself specifies no run-up gate). Two helper parameters not in the plan stay fixed: max_entry_slip_hours=8 (two 4H bars; an event is only traded at its own entry bar, not chased later) and hedge_tolerance_frac=0.25 (no BTC churn on rounding). Universe: the listing calendar today yields 67 qualifying perps; I dropped the two non-ASCII symbols (币安人生USDT, 牛来USDT, 2026 events) because their instrument ids are not safely representable, which reproduces the hypothesis's 65 events with the stated year split (2020:1, 2021:1, 2023:1, 2024:5, 2025:42, 2026:15). Risk control route (a): 8% notional x 25% stop = 2% of equity per event; gross at the 5-event cap is 80% of equity at 1x. Alignment: each event decision runs on the perp's own bar and the BTC hedge uses the latest BTC bar already delivered (never later than the leg bar). Caveats for the analyst: (1) adding BINANCE_SPOT legs creates a second (untraded) venue account, so if returns are measured against the sum of venue accounts, total return is diluted ~2x while avg_trade_return_pct and Sharpe are not; (2) the BTC hedge is one netted position scaled in/out across overlapping events, so per-trade stats for the BTC leg are booked per hedge-position lifetime, not per event; (3) Layer-3 load estimate ~1.1M bars over 131 bar types, within the 1800 s cap.