EthDeribitMonthlyBullPutSpreadVrpSkewIvRankGated

ETH Deribit Monthly Defined-Risk Short Put-Spread VRP + Put-Skew Harvester (Bull Put Spread, IV-Rank Gated, Trend-Filtered, Capped Tail)

A defined-risk, premium-harvesting options strategy on ETH at Deribit (2nd-most-liquid crypto options market, full greeks/IV history in catalog). Each monthly cycle it sells an out-of-the-money PUT and buys a further-OTM protective PUT below it — a BULL PUT SPREAD (vertical) — collecting net credit with a hard-capped maximum loss (spread width minus credit). The edge harvested is TWO structural premia stacked: (1) the volatility risk premium (Deribit ETH implied vol persistently exceeds realized), and (2) the PUT-SKEW premium — crypto puts trade at a persistent implied-vol premium to equidistant calls because of chronic downside-hedging demand, so systematically selling OTM put spreads monetizes overpriced downside insurance. This is mechanistically DISTINCT from the factory's symmetric-vol Deribit structures (short strangle, iron condor, iron fly): those harvest two-sided vol around ATM; this isolates the richly-priced DOWNSIDE wing. It is a legitimate structural extension of the BTC short-put VRP expression (which reached optimize-class) — VRP and put-skew are market-structure properties present in every liquid options market, not an asset-idiosyncratic momentum quirk, so the BTC→ETH transfer is principled (unlike the dead single-alt continuation ports). Defined-risk wings cap the crash-through-strike loss that blows up naked puts. Monthly tenor keeps frequency low (~12 cycles/yr) so the ~0.06% RT × 2-4 legs fee is trivial vs the credit. Sizing is defined-risk-relative (max loss per spread capped at ~2% of equity). ~4 parameters at robust round values to resist overfitting.

Harvests TWO stacked structural premia: (1) the volatility risk premium (Deribit ETH IV persistently exceeds realized) and (2) the PUT-SKEW premium (crypto puts trade at a persistent IV premium to equidistant calls from chronic downside-hedging demand), by systematically SELLING OTM put spreads to monetize overpriced downside insurance. Mechanistically DISTINCT from the factory symmetric-vol Deribit structures (short strangle, iron condor/fly) which harvest two-sided ATM vol: this isolates the richly-priced DOWNSIDE wing. It is a principled structural extension of the BTC short-put VRP expression that reached optimize-class -- VRP and put-skew are market-structure properties of every liquid options market, not an asset-idiosyncratic quirk, so BTC->ETH transfer is sound (unlike dead single-alt continuation ports). Defined-risk wings cap the crash-through-strike loss that blows up naked puts; the IV-rank gate sells only when vol is rich and the trend filter avoids selling into a downtrend. Monthly tenor (~12 cycles/yr) makes the ~0.06% RT x 2-4 legs fee trivial vs the credit. ETH is Deribit second-most-liquid options market with full greeks/IV history. Real options exposure with real option-leg trades in the analyzed backtest (smoke fires only in verification). leverage=1.0 (options venue cap); defined-risk sizing, no unused-leverage gate.

suspected backtest artifact + zero-OOS sparsity, not a validated edge — and over the 2-attempt limit. (1) The optimized config is degenerate: total_trades=0, metrics_reliable=false, extra_instruments=[] — the best-params run never injects the option legs, so the optimization result is unverifiable. (2) Walk-forward is single-window: in_sample_sharpes=[0,0,12.14] / out_of_sample_sharpes=[0,0,9.46] — two of three windows are zero-trade and one carries everything; is_overfitted=false is an averaging artifact over empty windows. PBO=0.623 (>0.5) and deflated_sharpe=null mean the multiple-testing gate cannot be cleared. (3) Plausibility fails decisively: Sharpe 7.5 (optimized WF 9-12), ±100% single-day returns (2025-08-29 +102%, 2025-08-28 -58%, 2022-08-25 +92%), and max_drawdown 146% are IMPOSSIBLE for a defined-risk bull put spread capped at 2% of equity/spread — the harness is not pricing/settling the protective leg as a capped spread (symptom stated; root cause not independently verified — recommend manual/factory-gate review of Deribit option-leg PnL marking before this family is retried). This is the same monthly-single-underlying-options-roll signature as EthDeribitMonthlyAtmLongCallRoll and BTC cash-and-carry: a regime+IV-rank gate makes the chain too sparse for walk-forward, so most windows are empty and a single window manufactures a flashy Sharpe that does not deflate. Not iterate: optimized config produces zero trades, two of three WF windows are zero-trade, PBO>0.5, no robust region to tune toward, and attempt 3 of 2. Not revise_hypothesis: there is no PROMOTED sibling of this mechanism to redirect (the cited BTC short-put VRP only reached optimize-class, not promotion), and the failure is artifact + sparsity in the options harness, not structural decay of a proven edge that a different instrument would rescue. FAILURE PATTERN: monthly single-underlying defined-risk option-spread rolls gated by IV-rank + trend regime produce too few periods for walk-forward (2 of 3 windows zero-trade), yield a degenerate zero-trade optimized config, fail PBO, and report artifact-level metrics (Sharpe 7.5-12, ±100% days, 146% DD that violates the defined-risk premise) — the VRP/put-skew premium is real but is not validatable in this sparse monthly-roll Deribit harness.

Defined-risk, premium-harvesting BULL PUT SPREAD on ETH at Deribit, rolled monthly. Each cycle the backtest rolling-options expander resolves the new monthly ETH chain from parameters[rolling_options] (underlying ETH, monthly roll, 30 DTE, two 1-DAY put legs at short_offset and long_offset moneyness, same expiry = a supported single-expiry vertical) and the strategy SELLS the higher-strike OTM put and BUYS the lower-strike further-OTM protective put, collecting net credit with a hard-capped max loss = (strike width - credit). Entry fires only when BOTH gates pass: (1) IV-RANK >= iv_rank_min, where IV-rank is the current Deribit DVOL ETH implied-vol ranked within its trailing iv_rank_lookback window (vol is rich); and (2) ETH closes above its sma_period day SMA (non-bear trend regime, since the bull put spread is bullish-to-neutral). Per-leg side is assigned by strike (sell higher, buy lower) since both legs are puts. The base RollingOptionsStrategy owns schedule/roll; this subclass adds the gates, the strike-based two-leg entry, defined-risk sizing and early close on a regime flip. Sizing is defined-risk-relative: contracts = equity*max_loss_frac / strike_width, so max loss per spread is ~max_loss_frac (2%) of equity. Subclasses BOTH RollingOptionsStrategy and FactoryStrategy with the four abstract hooks re-declared (Layer-1 AST). A verification-only long-perp smoke trade on the Deribit primary proves order plumbing in Layer-3 (no schedule there), gated OFF whenever a real schedule exists so the analyzed backtest always trades real option legs. ~4 parameters (short_offset, long_offset, iv_rank_min, sma_period).

DEFINED-RISK by construction: the long protective put caps max loss at (width - credit) x contracts (<=2% equity), so the tail is structurally bounded regardless of daily-bar gap resolution — this fixes the exact un-modeled-tail concern that made the naked short-put sibling (4a121630) suspect.

Greeks-correct premium harvester: 29 spreads, net_entry_vega -137 (short vol), net_entry_theta +222 (theta capture), net_entry_delta +21 (bullish, correct for a bull put spread), avg_iv 0.88 (IV-rank gated); profit_factor 2.19, +451% total.

Positive across every regime (2022 +94%, 2023 +85%, 2024 +145%, 2025 +106%, 2026 +14%) — consistent with harvesting persistent structural premia (VRP + put-skew), and the regime-flip exit cuts losing spreads early (explaining avg_win $307 > avg_loss $122).

Real losses present (largest_loss -$1,461, losing days exist) so it is not a riskless monotonic artifact; mirrors the BTC short-put VRP that reached optimize-class and fills the 0% Options quota.

Thin sample: 29 spreads (monthly cadence) — each walk-forward window holds ~10 trades; DSR is the decisive test.

Sharpe 7.51 (>5) and short_win_rate 93% (>90%) trip the suspected-artifact thresholds — defensible for a defined-risk credit spread but must still clear deflated-Sharpe.

Consistent-positive-every-year could also read as an artifact signature; the holdout must contain a genuine stress window where the spread takes its capped loss.

Options metrics are noisy (data_days 26, entry_diag 0 from the inert primary path) — not bugs, but read the headline ratios with the thin-sample caveat.

robust for walk-forward

capped at (width-credit)x contracts, <=2% equity

bounded loss

29 spreads, vega -137, theta +222

real greeks

Mechanism is economically real: VRP + put-skew are genuine market-structure premia, and the greeks are directionally correct for the structure (net_entry_vega -137 short vol, net_entry_theta +222 positive decay, 58 puts / 0 calls).

Sensitivity grid is clean on economic params (0 economic cliffs; the single cliff is roll_buffer_hours, an infra knob).

Fee drag is trivial (commission_pct_of_gross ~0, monthly cadence) — costs are not the problem.

Optimized backtest is DEGENERATE: total_trades=0, metrics_reliable=false, extra_instruments=[] — the best-params config does not trade, so the optimization outcome is unverifiable.

Walk-forward is effectively single-window: in_sample_sharpes=[0,0,12.14], out_of_sample_sharpes=[0,0,9.46]. Two of three windows have zero trades; the entire 'edge' rests on one window. is_overfitted=false is an artifact of averaging over empty windows.

PBO=0.623 (>0.5): parameter selection is more likely than not overfit. deflated_sharpe is null (not computable) — the multiple-testing gate cannot be cleared.

Suspected backtest artifact: headline Sharpe 7.5 (optimized WF 9-12) plus ±100% single-day returns and a 146% max drawdown directly CONTRADICT the defined-risk (max loss ~2%/spread) premise. The option-leg payoff is not being priced/settled as a capped spread (mechanism not independently verified — recommend harness review).

Economic strike-offset params (short_offset, long_offset, iv_estimate) are completely inert in sensitivity (identical Sharpe across all variations) because strikes are pre-resolved in a fixed schedule — the optimizer only moved non-edge knobs.

Over the iteration limit (attempt 3 of 2).

implausible — suspected artifact, not a real edge

146% DD impossible for a defined-risk 2%/spread structure

not computable; PBO 0.623 > 0.5

2 of 3 windows zero-trade

all-positive

EthDeribitMonthlyBullPutSpreadVrpSkewIvRankGated sold IV-rank-gated, trend-filtered ETH bull put spreads to harvest the volatility-risk and put-skew premia with a structurally capped tail, a principled extension of the BTC short-put VRP work. The greeks were correct and the baseline dazzling — +451%, Sharpe 7.51, profit factor 2.19, positive every year — earning a full optimization, but that exposed it as artifact + sparsity: the optimized config traded zero times, two of three OOS windows were empty, PBO 0.623, and a 146% drawdown with ±100% days that a 2%-capped spread cannot produce, indicating the protective leg isn't being settled as a capped spread. The analyst ruled it un-validatable in the sparse monthly-roll Deribit harness (not a structural edge a different instrument could rescue, with no promoted sibling to redirect to) and over the attempt limit. It ended as abandoned, reaching optimization and analysis but never risk review.

Monthly single-underlying defined-risk option-spread rolls gated by IV-rank + trend regime are too sparse to validate (two of three walk-forward windows zero-trade) and the Deribit option-leg PnL is being marked implausibly (Sharpe 7.5-12, ±100% days, a 146% drawdown that violates a 2%-capped spread) — the VRP/put-skew premium is economically real but un-validatable in this sparse monthly-roll harness, which needs a manual harness review before the family is retried.

It passed the pre-optimization backtest-review gate (verdict: optimize) but was abandoned at the post-optimization analyst gate (verdict: abandon) as a suspected backtest artifact plus zero-OOS sparsity, not a validated edge — the optimized config produced zero trades, two of three walk-forward windows were empty, PBO exceeded 0.5 with DSR uncomputable, and the 146% drawdown / ±100% days contradict the defined-risk premise (the harness isn't settling the protective leg as a capped spread) — and it was over the 2-attempt iteration limit, so it never advanced to risk review.

A defined-risk monthly bull put spread on ETH Deribit — selling an OTM put and buying a further-OTM protective put (capped max loss = width − credit, ~2% equity/spread), gated to enter only when IV-rank was rich and a non-bear SMA trend regime held, closing on roll buffer or regime flip — designed to harvest two structural premia (the volatility risk premium and the persistent put-skew) and isolate the richly-priced downside wing, as a principled extension of the BTC short-put VRP expression.

Across 29 spreads (58 put legs, correctly short-vol: net vega -137, theta +222, IV-rank gated, avg IV 0.88) the baseline was strong — total return +451%, Sharpe 7.51, profit factor 2.19, short win rate 93%, positive every year (2022 +94% through 2026 +14%) — but optimization was degenerate and implausible: the best-params config traded ZERO times (metrics_reliable=false, no legs injected), walk-forward had two of three windows zero-trade ([0, 0, 9.46] OOS), PBO 0.623, DSR null, and the headline showed ±100% single days and a 146% max drawdown that is impossible for a 2%-capped defined-risk spread.