Strategies

Post-Listing Lottery-Premium Decay on Binance USD-M Perps — Short Each NEWLY LISTED Perp From Its Own Day+3 for 45 Days, BTC-Hedged Basket of ~37 Never-Traded Listings (2024-2026), Hard +30% Squeeze Stop, 3 Parameters

Outcome: Abandoned

PostListingLotteryPremiumDecayBasketLS

Outcome Summary

PostListingLotteryPremiumDecayBasketLS proposed an event-driven long/short book: short every newly listed Binance USD-M perp starting at its own day+3 for 45 days, capped by a +30% squeeze stop and hedged with a long BTCUSDT position sized to aggregate short notional, with only three tunable parameters. The code was written and reached the data stage, but data collection failed after five retries — one basket leg returned zero daily bars against a 50-bar / 90-day minimum — so no backtest, optimization, analyst verdict or risk review was ever produced. The run ended as abandoned for data_unavailable, which the abandon note explicitly attributes to the data layer rather than the strategy. The hypothesis remains re-evaluable once the missing listings are backfilled.

Strategy report

Hypothesis

An EVENT-DRIVEN, LONG-SHORT, primary-market strategy whose signal is a LISTING EVENT, not a price pattern: a brand-new Binance USD-M perp opens into attention-driven retail long demand with no borrow market, no spot inventory and no basis trade to supply the other side, and that premium decays over the following weeks. The book holds rolling SHORTS in recently-listed perps (entry at the close of each name's own 3rd bar after its first-ever bar, 45-day hold) hedged with a LONG BTCUSDT.BINANCE position sized to aggregate short notional. MEASURED ON OUR OWN CATALOG BEFORE PROPOSING: across 659 perps first listed 2023-01→2026-06 with >$20m/day early volume, day+3 / 45-day / +30%-stop BTC-hedged returned mean +6.53%, median +0.76% per trade, 51% win rate, POSITIVE IN ALL FOUR YEARS (2023 +7.3%, 2024 +13.7%, 2025 +4.8%, 2026 +3.4%), and every cell of the entry-day × hold × stop grid was positive (+2.4% to +7.6%).

Mechanism as written: a brand-new perp opens into attention-driven retail long demand with no borrow market and no spot inventory to supply the other side, so the listing premium decays over the following weeks; shorting it while long BTC isolates that decay from market beta. Implementation choices: (1) BTCUSDT.BINANCE is the PRIMARY (it is the only leg with data over the whole span and it carries the hedge), the 37 listings are extra_instruments; calculate_signal returns a continuous variable — the mark-to-market short notional as a fraction of equity — which moves every bar with both the composition and the marks of the short book, and should_enter/should_exit/position_size manage the BTC hedge against it. (2) The base template's same-timeframe alignment barrier is deliberately disabled: the legs list on different calendar dates and never all coexist, so waiting for a contemporaneous bar from all 37 would block the hedge for the entire run. (3) A leg only fires when its first observed bar is within 5 days of its known listing date, so a backtest window that starts mid-history cannot fabricate a listing event. (4) The squeeze stop triggers on the daily CLOSE and fills at that close rather than triggering on the intrabar high and filling at the close, which would book a better price than a real stop. (5) Basket selection is mechanical and observable at entry — the 19 instruments the hypothesis declares, plus the highest early-3-day dollar-volume listings in each remaining half-year to reach ~37 and to cover 2025-H2 and 2026-H1, excluding tokenized equities/commodities and stablecoins. Two caveats the analyst should price in, since I implement the hypothesis as asked rather than change it: the basket can only contain names that are STILL LISTED today (perps delisted since 2024 have no current data and cannot be backtested), so there is residual survivorship in the sample; and BTCUSDT daily history starts in 2020 while the first listing event is 2024-01, so the full-history backtest carries roughly 1,500 idle zero-return days that dilute the annualized Sharpe by about a third relative to the traded span.

data_unavailable: backtest data for BTCUSDT.BINANCE could not be collected after 5 retries: - bars (bar_type=REUSDT.BINANCE-1-DAY-LAST-EXTERNAL): got 0, needed 50, or 90+ days span The data layer (not the strategy) is the blocker — re-evaluate this hypothesis once collection is fixed / the symbol is backfilled.

Implementation

Event-driven long/short basket on Binance USD-M perps, daily bars. The signal is a LISTING EVENT: for each of 37 perps first listed between 2024-01 and 2026-06 (all still trading, all with >$200m/day early dollar volume), the strategy SHORTS the name on the bar 3 calendar days after its first-ever bar and holds 45 calendar days, exiting early if the daily close ever sits 30% above the entry price (the squeeze stop). Each short is sized at equity x leverage x short_gross_frac / max_concurrent_shorts (= 10% of equity at the defaults), and at most 6 legs are held at once, so gross short exposure is capped at 60% of equity. The aggregate short book is hedged with a LONG BTCUSDT.BINANCE position equal to the basket's mark-to-market short notional, rebalanced whenever the gap exceeds 2% of equity, so total gross stays at or below 1.2x equity against a 2x leverage request. Holding periods, entry offsets and the roll of the book are all derived from each name's listing DATE and the bar's own timestamp, never from a bar counter, so restarts change no decision.

Outcome Summary

A basket keyed to dozens of recently listed symbols is only as testable as the thinnest leg — confirm catalog coverage for every declared instrument (especially short-history new listings) before proposing a multi-leg event-driven strategy.

The run was abandoned at the data stage after 5 retries: required bars could not be collected (bar_type REUSDT.BINANCE-1-DAY-LAST-EXTERNAL returned 0 bars against a need of 50 bars / 90+ days span), so backtest-review, analyst and risk stages were never reached and no verdict was ever issued.

Short each newly listed Binance USD-M perp from its own day+3 for a 45-day hold with a hard +30% squeeze stop, as a BTC-hedged basket, betting that the listing-day 'lottery premium' from attention-driven retail long demand decays over the following weeks.

No backtest metrics exist — the backtest report is empty and optimization never ran; the only numbers on record are the hypothesis author's own pre-proposal catalog study (mean +6.53% / median +0.76% per trade, 51% win rate, positive in all four years 2023-2026), which the pipeline never independently reproduced.

Backtest and paper results are hypothetical. Trading involves risk of loss.