Strategies

ETH Cross-Venue Systematic Covered-Call Overwrite, IV-Rank Gated (Long ETH Spot on BINANCE_SPOT + Rolling 1-Month ~25-Delta Short Call on DERIBIT) — Enhanced-Yield Long, Overwrite ONLY When Call IV Is Rich, 2-Parameter

Outcome: Abandoned

EthSpotCoveredCallDeribitMonthlyIvRankGated

Outcome Summary

This cross-venue strategy aimed to combine long ETH spot on BINANCE_SPOT with a rolling monthly short Deribit call, overwriting only when implied vol was rich by IV-rank, to fill the under-represented options and cross-venue buckets with a mechanism distinct from previously-dead options classes. Although a full 48-contract Deribit chain was resolved and a 2017-2026 backtest ran, the option leg never traded: only a single long underlying trade was recorded, and the metrics were explicitly marked unreliable with a flat 0.0 Sharpe. The pipeline abandoned it at the backtest stage on its first iteration under 'options_no_option_trades', because the strategy had degenerated into buy-and-hold spot rather than the covered call it was designed to be. Optimization, analyst review, and risk review were never reached.

Strategy report

Hypothesis

A CROSS-VENUE, low-frequency, enhanced-yield structure that holds long ETH spot (ETHUSDT.BINANCE_SPOT) and systematically overwrites it with a rolling one-month, out-of-the-money (~25-30 delta) SHORT call on Deribit (ETH options, the most liquid crypto options chain after BTC, monthly expiries with multi-year history from ~2020). This is deliberately NOT the two options mechanisms that already died in this factory: it is NOT a delta-neutral naked short straddle/strangle (which died no_edge because its edge sat inside costs and its tail was unsampled), and it is NOT a long-single-option convexity play (which died on an option-MTM accounting artifact). Here the option leg is SHORT and small-delta, and it is fully COVERED by the spot inventory, so net delta stays long and gamma exposure is bounded by the shares held. The overwrite is CONDITIONAL: we only sell the monthly call when 30-day ATM implied vol is RICH relative to trailing realized vol (IV-rank above a threshold), i.e. when the call premium genuinely over-compensates for the upside we cap. When IV is cheap we simply hold spot un-overwritten and forgo the premium. Fills the two most under-represented portfolio buckets simultaneously (Options 5.8% vs ≥15% target; cross-venue 7% vs ≥15% target) with a mechanism absent from the dead-class list, on ETH rather than BTC to relieve the 62% BTC corpus concentration.

Implements the hypothesis exactly: long ETHUSDT.BINANCE_SPOT overwritten with a rolling 1-month ~25-30-delta SHORT Deribit call, conditional on an IV-rank gate, with two tunable knobs (iv_rank_threshold, iv_rank_window_days) and everything else structural. IV comes from 'deribit_dvol' (ETH DVOL, forward 30d IV) read strictly point-in-time via a precomputed sorted array + searchsorted (O(log n) per bar, no per-bar rescans, no look-ahead); IV-rank is the percentile of the current level within the trailing window. If DVOL is unavailable at a roll the strategy does NOT overwrite — no silent fall-back to an ungated premium sale or a different mechanism. Venue choice follows the structure: the covering inventory is long-only, unlevered spot (BINANCE_SPOT, matching the hypothesis's declared ETHUSDT.BINANCE_SPOT leg) and the option leg is a Deribit European monthly, so the config is cross-venue with leverage 1.0. Two deliberate risk decisions the hypothesis left open: (a) because the legs sit on SEPARATE accounts (no cross-margin), the short call is written against only overwrite_ratio (0.6) of the held ETH and is further capped so a 35% adverse move costs at most half the Deribit account — an uncapped 1:1 overwrite can drive the option account to zero in a parabolic ETH month even though the combined book is economically flat; (b) I gate on IV-rank alone rather than also requiring IV > realized vol, because ANDing two vol filters would cut the ~12 overwrites/year to a handful and starve the sample. Sizing is capital-relative via get_account_equity(); the option leg is sized in whole-ish Deribit contracts (1 contract = 1 ETH) from the actual spot position.

options_no_option_trades: options strategy produced 0 option-leg trades (1 underlying trades) — degenerated to the underlying / smoke path; the option chain never traded

Implementation

Cross-venue systematic covered-call overwrite on ETH. Leg 1 buys and holds ~95% of the BINANCE_SPOT account in ETHUSDT (long-only CASH, no leverage). Leg 2 sells a rolling MONTHLY ~10%-OTM (~25-30 delta) Deribit ETH call against a fraction of the ETH actually held, but ONLY when 30-day implied vol (Deribit DVOL) sits at or above its iv_rank_threshold percentile of the trailing iv_rank_window_days of DVOL observations (point-in-time). When IV is cheap the spot is held un-overwritten and the premium is forgone. Each short call is closed 48h before expiry and rolled into the next monthly. Net delta stays long; the short call is always covered by spot inventory and is additionally capped so a stress rally cannot cost more than half the Deribit account. Rolling chain resolution, entry timing, the pre-expiry close and the roll are owned by RollingOptionsStrategy; there is no underlying smoke path (an unresolved chain trades nothing at all, not even spot).

Verification Results

Keep as-is if the base's _leg_size/_enter_leg contract is stable; otherwise add a lightweight assertion/log confirming the override is being invoked by the base each roll so a signature drift fails loud.

_leg_size(self, spot) and _enter_leg(self, iid, kind, spot, bar) shadow RollingOptionsStrategy framework methods (flagged by static analysis). The signatures match the base's, and the developer documents these as deliberate overrides (covered/gated sizing and gated entry), so behavior is intended — but this couples the strategy tightly to the base's private call contract; a future base-template change to these signatures would silently bypass the gate.

No change required given the 7-day entry window; just be aware the first monthly overwrite of the backtest may be skipped if spot fills late.

On the very first roll bar the spot market order is submitted in _manage_spot but has not filled when _enter_leg reads _spot_eth() (returns 0 -> 'no_spot_cover' block). This is mitigated by the 168h entry window (later bars within the window find the spot filled), so overwrites are not permanently lost, but a roll whose entry window is unusually tight relative to spot fill latency could skip an overwrite.

Two things for the analyst to weigh at backtest review, both about measurability rather than code correctness: (1) The overwrite leg is monthly (~12 opportunities/yr) AND IV-rank gated, so over the Deribit ETH option-chain history actually collected in this factory (recent runs found ETH option chains beginning only ~2024-08, far short of the hypothesis's claimed ~2020 depth), the number of ACTUAL call overwrites may fall well below the ~30-trade floor needed to distinguish edge from noise (cf. options-cadence trade-count pattern). (2) The Layer-3 sandbox recorded only 1 trade (the spot buy-and-hold, 962% 'return', Sharpe 0.0, 2399d hold) with no observed call fill over 2400 bars — consistent with either the IV-rank gate not firing on that draw or an empty/short option chain in the sandbox window. Confirm the Deribit monthly-chain + deribit_dvol coverage is deep enough to produce a testable overwrite sample, and that per-overwrite premium clears Deribit option fees, before spending optimization compute.

Outcome Summary

An options overwrite strategy must be verified to actually enter option legs during the backtest; if the chain resolution, IV-rank gate, or coverage sizing never fires a call, the run collapses to plain spot and cannot test the premium-harvesting thesis.

It was abandoned at the backtest gate with reason 'options_no_option_trades': the strategy produced 0 option-leg trades against 1 underlying trade, so the Deribit option chain never traded and the position degenerated to the underlying/smoke path — the intended edge was never exercised.

Hold long ETH spot on BINANCE_SPOT and overwrite it with a rolling monthly ~25-30-delta short Deribit call, selling the call only when ETH implied vol (DVOL) was rich versus its own trailing history (IV-rank gated), to harvest option premium as enhanced yield while staying net-long and fully covered.

The backtest ran over 3,308 daily ETH bars (2017-2026) but was flagged unreliable: only 1 trade occurred (a single long underlying trade), Sharpe/Sortino/Calmar were 0.0 and only one day moved the equity curve. The reported 347% total return and 732% avg trade return reflect a degenerate buy-and-hold of spot, not the covered-call mechanism.

Backtest Review

Total return: 347.32%
Max drawdown: 0.00%
Trades: 1
Win rate: 100.0%
Profit factor: 0.00

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