BtcDeribitWeeklyShortStrangleVrpHarvest

Deribit BTC Weekly Short Strangle — Variance Risk Premium Harvest

Sell OTM call + OTM put on BTC weekly options (7 DTE) every Friday at 08:00 UTC expiry roll, harvesting the variance risk premium (implied vol systematically > realized vol). Strangle is initially delta-neutral at entry. Position held to expiry (no intra-week rebalancing — keeps trade count low and fees minimal). This is a structurally different mechanism class from every approach falsified in this session: it does not rely on price direction, trend persistence, mean reversion of price, cross-asset relationships, or funding rate signals. The edge is a documented insurance-premium phenomenon in options markets (Bakshi-Kapadia 2003, Carr-Wu 2009) that has persisted in crypto options since Deribit's launch in 2016. Strangle width chosen at ~1.5 standard deviations (delta ~0.15 each leg) to balance premium collection vs. assignment risk.

Implements the hypothesis exactly: a single-venue (DERIBIT), options-only, delta-neutral weekly short strangle sold UNCONDITIONALLY every cycle and held to expiry. The edge is the documented variance risk premium (Bakshi-Kapadia 2003, Carr-Wu 2009) — implied vol structurally exceeds realized vol because buyers overpay for convexity/protection — monetized two-sided and market-neutral, a mechanism class orthogonal to price direction, trend, mean-reversion, cross-asset and funding signals. Strikes at symmetric +-10% offsets sit near ~1.5 std / ~0.15 delta per leg, balancing premium vs assignment risk while staying liquid enough to have option candles for the backtest. Weekly turnover (~52 cycles/yr, no rebalancing) keeps the ~0.06%RT x2 Deribit fee negligible vs 1-3% weekly premium. Routed to DERIBIT (deepest BTC option liquidity); leverage=1 (options venue cap) and sizing is in BTC contracts (risk_frac of equity), so no unused-leverage gate fires. The optional VRP gate (default OFF) is exposed purely so optimization can test timing without changing the base unconditional design. Fills the under-represented DERIBIT options portfolio bucket.

As-tested the unconditional weekly short strangle has a ruin-level tail: max_drawdown ~101.7% (calmar 0.804 = total_return 0.817 / 1.017) driven by a single week (2026-02-06) that lost -101.2% of equity — an account wipeout, not a tunable drawdown. This is the short-vol steamroller fully realized: avg_loss ($211) is 8.7x avg_win ($24), win-rate 93%, return_skew -5.0, kurtosis 32.8, and the Sharpe (1.46) has sharpe_ci_low -2.24 straddling 0. Unlike the previous submission this is NOT a byte-identical duplicate (it is the unconditional, wider-strike ~0.15-delta / 10%-offset variant with net_entry_delta -0.008, vs the gated siblings' 7%/16-delta), so it is judged on its own merits — and the unconditional sell demonstrates exactly the failure the siblings' VRP gate was designed to avoid: it caught the full Feb-2026 vol spike. Its only viable improvement (turn use_vrp_gate ON and/or shrink size) reproduces the VRP-gated sibling (8ab5ff3d) that was already sent to optimization in this session, so spending 2 hours optimizing this third family variant — whose as-tested form wipes the account — is poor compute allocation against an already-advancing, better-behaved representative. The shared ~1-year / 52-expiry resolved-chain limitation applies. FAILURE PATTERN: an unconditional held-to-expiry BTC weekly short strangle harvests theta cleanly most weeks but takes a single >100%-of-equity loss in a vol spike (max DD >100%, skew -5, 8.7x loss/win asymmetry) — naked short-vol without a stand-down gate is structurally ruin-prone, and the gated version of the same mechanism is already in the pipeline, making this variant both non-viable as-is and redundant once fixed.

Single-venue, options-only variance-risk-premium harvest on BTC Deribit. EVERY weekly roll (7-DTE) it sells a delta-neutral OTM strangle on the nearest weekly expiry: short a ~0.15-delta call (~+10% strike) and a ~0.15-delta put (~-10% strike, ~1.5 std), symmetric offsets so net delta ~0 at entry with no perp hedge. Both premiums are collected and the strangle is HELD TO EXPIRY (no intra-week rebalancing), closed 4h before European cash settlement to dodge thin last-minute liquidity, capturing a full week of theta. Built on RollingOptionsStrategy: the Deribit perp (primary) drives the weekly roll clock and supplies 1H history for the optional RV calc; the option legs are injected by the rolling-options chain expander at backtest time. An optional VRP gate (IV>=RV*margin) is wired in but DEFAULTS OFF so the base behaviour is the hypothesis's unconditional sell.

Real mechanism trades cleanly: 102 short option legs (51C/51P), net_entry_delta -0.008 (genuinely delta-neutral), net_theta +722, net_vega -187 — a true VRP harvest, not a degenerate options roll

Single-venue / options-only, fills the under-represented Deribit bucket; zero fee drag

More honest headline than the gated siblings (Sharpe 1.46, not 4.71) — the unconditional sell does not flatter itself

Ruin-level tail: max_drawdown ~101.7% (calmar 0.804 = 0.817/1.017) with a single week 2026-02-06 = -101.2% — an account wipeout that would liquidate in live trading

Defining short-vol fragility: avg_loss ($211) = 8.7x avg_win ($24), win-rate 93%, skew -5.0, kurtosis 32.8 — many small premiums, one catastrophic week; sharpe_ci_low -2.24 straddles 0

The only real fix (enable use_vrp_gate / shrink size) converts this into the VRP-gated sibling 8ab5ff3d already advancing to optimization — redundant improvement path

Same family limitation: only ~1 year / 52 weekly expiries of resolved chain (trades from 2025-06-27), regime-limited; third short-strangle VRP submission this session

BtcDeribitWeeklyShortStrangleVrpHarvest sold delta-neutral ~0.15-delta OTM BTC weekly strangles every roll and held to expiry, harvesting the variance risk premium unconditionally with the VRP timing gate deliberately defaulted off. The mechanism traded cleanly and its headline was more honest than the gated siblings (Sharpe 1.46, not 4.71), but the unconditional sell caught the full February-2026 vol spike: a single -101.2% week wiped the account, with avg losses 8.7x avg wins and skew -5.0. The analyst judged it on its own merits — a structurally ruin-prone naked short-vol book — and noted its only fix reproduces the gated sibling already optimizing, making it non-viable as-is and redundant once corrected. It ended after one iteration as abandoned, never advancing to optimization or risk review.

Naked short-vol without a stand-down gate is structurally ruin-prone: an unconditional weekly short strangle harvests theta cleanly ~93% of weeks but takes a single >100%-of-equity loss in a vol spike (max DD >100%, skew -5, 8.7x loss/win asymmetry, Sharpe CI straddling 0), and the only honest fix — a VRP/IV gate plus smaller size — is exactly the gated sibling already in the pipeline; an unconditional premium-seller must have a regime gate or it eventually wipes the account.

It was abandoned at the pre-optimization backtest-review gate (verdict: abandon) because the unconditional held-to-expiry strangle takes a >100%-of-equity loss in a single vol spike (not a tunable drawdown) — the exact failure the siblings' VRP stand-down gate was built to avoid — and its only viable fix (turn the gate on / shrink size) reproduces the VRP-gated sibling already advancing to optimization, making it both non-viable as-is and redundant once fixed, so optimization and all later stages were never reached.

A market-neutral, options-only variance-risk-premium harvest on BTC Deribit weekly European options — every Friday roll, unconditionally selling a delta-neutral ~0.15-delta OTM strangle (one ~+10% call + one ~-10% put on the nearest 7-DTE expiry) and holding to cash-settled expiry to capture theta decay, with an optional IV-vs-RV gate defaulted OFF so the base behavior is the hypothesis's pure unconditional weekly sell.

Across 102 short legs (51 calls / 51 puts, genuinely delta-neutral at -0.008, net theta +722, net vega -187, avg 6.9 DTE) it showed the classic short-vol signature — 93% win rate, profit factor 1.55, Sharpe 1.46 (CI [-2.24, 14.23]) — but a ruin-level tail: ~101.7% max drawdown driven by a single week (2026-02-06, -101.2% of equity, an account wipeout), avg_loss ($211) 8.7x avg_win ($24), return skew -5.0, kurtosis 32.8, over only ~1 year / 52 weekly expiries of resolved chain.