EthDeribitMonthlyAtmLongCallRollRegime1D

ETH Deribit Monthly ATM Long-Call Roll with 200-SMA Bull Regime Filter

A long-only single-instrument LONG-CALL OPTIONS strategy on ETH options on the DERIBIT exchange (deepest crypto options venue) using daily bars on the underlying ETH for signal generation and monthly ATM call options for execution. This is the SECOND options strategy in the portfolio (companion to BtcDeribitMonthlyAtmLongCallRollLong) and provides asset-diversified options exposure: ETH and BTC options have correlated but materially different IV regimes, Greeks, and payoff dynamics. ETH IV typically ranges 70-90% annualized vs BTC's 50-70% — ETH calls are MORE EXPENSIVE relative to underlying spot but ALSO DELIVER LARGER convex payoffs when ETH trends. This per-asset edge case (ETH-specific IV regime) is fundamentally different from BTC: ETH's narrative cycles (ETF rebalances, EIP releases, staking yield dynamics, DeFi-cycle catalysts) produce convex moves that linear futures/spot strategies cannot capture efficiently — the gamma exposure of long calls captures the asymmetric upside that ETH's narrative-driven rallies generate. ETH options on Deribit have ~4 years of history (since 2021) with the deepest liquidity in ETH options globally. The mechanism mirrors the BTC Deribit options approach: ETH > 200-day SMA (bull regime) → maintain a 30-day-to-expiry ETH ATM call position. Close existing call when 7 DTE remaining, immediately open a new 30-DTE ATM call. Close all positions if ETH closes below 200-SMA. ETH is currently the most UNDERUTILIZED asset in the portfolio after the HL Pullback failure (only EthDailyGoldenCrossMomentumLong remains — 1 strategy vs 2-3 for every other tracked asset). Adding ETH options coverage (a) doubles portfolio options exposure (from 1 to 2 strategies, both providing convex payoffs in bull regimes), (b) fills the ETH coverage gap, (c) hedges BTC-specific tail risks because ETH and BTC don't always move in lockstep. Convexity is the key portfolio benefit: when ETH delivers a +30% monthly move (which has happened multiple times in ETH-cycle peaks), an ETH ATM call can deliver +500-1000% returns, dramatically improving Sharpe and Sortino at portfolio level. Only 3 explicit parameters (regime SMA period, target DTE on entry, roll DTE on exit) — minimum-parameter options strategy.

Asset-diversified second options strategy, the ETH mirror of the validated in-portfolio BTC Deribit monthly ATM long-call roll. ETH and BTC options are correlated but have materially different IV regimes (ETH ~70-90% vs BTC ~50-70%), Greeks and convex payoff dynamics: ETH calls are pricier relative to spot but deliver larger convex payoffs when ETH trends, and ETH narrative cycles (ETF flows, EIP releases, staking/DeFi catalysts) produce asymmetric upside that linear futures/spot cannot capture efficiently -- long-call gamma captures it. Fills the portfolio ETH coverage gap (most underutilized asset) and doubles options exposure (1->2 convex strategies), hedging BTC-specific tail risk since ETH/BTC do not always move in lockstep; a +30% ETH month can pay +500-1000% on an ATM call, lifting portfolio Sharpe/Sortino. Genuine options exposure (real theta/gamma/premium decay), NOT a synthetic-futures proxy: the rolling expander injects real monthly call legs and the options_no_option_trades gate ensures the analyzed backtest places real option-leg trades; the smoke path fires ONLY in verification where no schedule exists. Deribit is the deepest crypto options venue with ~4y ETH history since 2021. Minimal 3-parameter surface resists overfitting. leverage=1.0 (options venue cap), equity-relative sizing, no unused-leverage gate.

Degenerate out-of-sample validation on a thin, regime-beta long-call roll — not fixable by parameter changes. The mechanism is implemented correctly (real long call: vega +113, theta -108, positive skew +2.0, convex), but the optimization gates fail decisively and uniformly: walk-forward out_of_sample_sharpes = [0.0, 0.0, 0.0] across ALL three windows (avg OOS 0.0 vs avg IS 1.63), is_overfitted=TRUE, pbo 0.53; the holdout is degenerate (holdout_sharpe 0.0, ratio 0, passed=false); and the optimized re-run produced ZERO trades (metrics_reliable=false). The optimizer searched 150 trials and found no config that both trades and wins out-of-sample. This is the over-selection / zero-OOS / degenerate-holdout signature, compounded by a structurally thin sample: only 24 trades over ~4 years (monthly roll gated by the 200-SMA regime), with sharpe_ci_low -8.44 (the headline Sharpe 2.96 is statistically indistinguishable from zero), max_drawdown 61%, and a sma_period sensitivity cliff (0.81@180 vs 2.96@200). The positive full-sample return is regime-dependent long-ETH convexity carried by a few bull months (2024-02 +95%, 2024-05 +103%, 2025-07 +158%; 2023 annual -5%) — bull beta, not validated alpha. Not iterate: there is no robust parameter region to tune toward when every OOS window is zero, the holdout is degenerate, and the sample is 24 trades; narrowing the (already minimal) parameter set cannot manufacture out-of-sample trades or significance. Not revise_hypothesis: the failure is overfitting + zero-OOS + thin sample, not a structurally-dead target for a proven mechanism — and the long-call-roll family is essentially regime-gated long-asset convexity whose 'edge' is bull beta that does not survive walk-forward on either the BTC or ETH instance. FAILURE PATTERN: regime-gated monthly ATM long-call roll on a single crypto underlying produces a flashy bull-carried full-sample Sharpe but collapses to zero out-of-sample across all walk-forward windows with a degenerate holdout and a zero-trade optimized config — too few trades (monthly roll x SMA gate) for the optimizer to validate, and the convex payoff is long-beta, not generalizable alpha.

Long-only ETH Deribit MONTHLY ATM long-call roll gated by a 200-day SMA bull regime. The primary instrument is ETH-PERPETUAL.DERIBIT on 1-DAY bars (the roll clock and the SMA(200) underlying source); the tradeable structure is a SEQUENCE of monthly near-ATM ETH call contracts resolved by the backtest rolling-options expander from parameters[rolling_options] (underlying ETH, roll monthly, dte_target 30, ATM offset 0.0, 1-DAY legs, 4y lookback) into concrete legs + a per-period roll schedule. Each monthly expiry the strategy opens ONE ATM call ~entry_target_dte_days (30) before expiry and rolls it out ~exit_roll_dte_days (7) before expiry, BUT only while ETH closes above its 200-SMA (confirmed bull regime). It closes the open call immediately if the regime flips bearish (ETH <= SMA) or on a premium-loss catastrophe stop (stop_loss_pos_pct). Sizing is equity-relative (risk_frac) with a 1-contract floor. The base RollingOptionsStrategy owns schedule parsing/entry/roll; this subclass adds the regime gate and early exit. Subclasses BOTH RollingOptionsStrategy and FactoryStrategy and re-declares the four abstract hooks so Layer-1 AST recognizes it while the MRO routes roll logic through the rolling base. A verification-only long-perp smoke trade proves order plumbing in the Layer-3 sandbox (no chain expansion there) and is gated OFF whenever a real schedule exists, so the analyzed backtest always trades real option legs. 3 economic parameters (sma_period, entry_target_dte_days, exit_roll_dte_days); pure OHLCV signal, zero supplementary data.

Genuinely convex, greeks-aware options exposure (24 real ATM Deribit call trades: avg_moneyness 1.018, avg_iv 0.48, net theta -108, net vega +113) — not a synthetic-futures proxy; the monthly chain resolves and trades concrete legs.

Textbook positive convex payoff: win_rate 37.5% with avg_win 3.5x avg_loss (losses capped at premium), profit_factor 2.13, total_return +242%, Sortino 17.9, driven by a few large rally captures (the expected long-call shape).

Mirrors an already-promoted BTC Deribit long-call sibling (validated mechanism class), fills the under-covered ETH bucket, doubles portfolio options/convexity exposure, and is low-parameter (3 knobs: SMA period, entry/exit DTE).

Orthogonal to the linear spot/futures book by payoff shape; theta/decay behavior is normal and evaluated over 24 expiry cycles.

Thin sample: only 24 trades (one regime-gated monthly roll over ~4 years) — inherent to the strategy, but each walk-forward window will have ~8 trades, and the sharpe CI is enormous (-8.44 to +7.62).

Result is concentrated in a few big convex wins (2024-05, 2025-07) — robustness hinges on catching ETH bull rallies; 2023 was slightly negative.

Net long delta (+29.5): this is a regime-gated convex long-ETH bet, correlated with ETH direction — it diversifies via convexity, not market-neutrality.

Several metrics are options-specific and noisy (data_days 22, entry_diag 0 from the inert primary path) — not bugs, but the headline ratios should be read with the thin-sample caveat.

robust for walk-forward

+242%, Sortino 17.9

positive

avg_win 3.5x avg_loss, PF 2.13

positive convexity

24 ATM calls, greeks present

real convex legs

Mechanism correctly implemented: real long-call roll with the right greeks (net_entry_vega +113 long vol, net_entry_theta -108 pays decay, net_entry_delta +29.5), positive skew +2.0 and tail_ratio 4.0 confirming genuine convexity; fills the under-represented Deribit options / ETH buckets.

Sensitivity is mostly stable (only 1 cliff) and the headline backtest is positive (Sharpe 2.96, PF 2.13, total_return 242%).

Walk-forward out-of-sample Sharpe is 0.0 across ALL THREE windows (avg OOS 0.0 vs avg IS 1.63), is_overfitted=TRUE, pbo 0.53 — the edge does not generalize out-of-sample.

Holdout is degenerate: holdout_sharpe 0.0, ratio 0, passed=FALSE; and the optimized re-run produced ZERO trades (metrics_reliable=false) — the optimizer found no validatable config across 150 trials.

Only 24 trades full-sample (monthly roll x 200-SMA gate is structurally thin) and sharpe_ci_low -8.44 — the headline Sharpe is not statistically distinguishable from zero; max_drawdown 61%.

Returns are regime-dependent long-ETH convexity carried by a handful of bull months (2024-02 +95%, 2024-05 +103%, 2025-07 +158%; 2023 annual -5%), not a generalizable alpha; sma_period is a sensitivity cliff (0.81@180 vs 2.96@200).

ratio>=0.70

no cliff

EthDeribitMonthlyAtmLongCallRollRegime1D bought monthly ATM ETH calls while ETH held above its 200-SMA, a clean greeks-correct convexity play companion to the promoted BTC sibling, aiming to diversify the book with asymmetric bull-rally upside. The baseline dazzled — +242%, Sharpe 2.96, 3.5:1 payoff — so it earned a full optimization, but that exposed the edge as bull beta carried by a few months: walk-forward OOS Sharpe was zero in all three windows, the holdout degenerate, and the optimized re-run traded zero times across 150 trials on a 24-trade sample. The analyst ruled it overfit + zero-OOS + structurally thin, not a tunable miss, with the long-call-roll family failing walk-forward on ETH as on BTC. It ended after one iteration as abandoned, reaching optimization and analysis but never risk review.

A regime-gated monthly ATM long-call roll on a single crypto underlying is essentially bull beta, not generalizable alpha: a flashy full-sample Sharpe carried by a handful of bull months collapses to zero out-of-sample with a degenerate holdout, and the monthly-roll × 200-SMA gate produces too few trades (24 over 4 years) for the optimizer to validate — the family fails walk-forward on ETH just as on BTC, regardless of the minimal parameter set.

It passed the pre-optimization backtest-review gate (verdict: optimize) but was abandoned at the post-optimization analyst gate (verdict: abandon): the edge does not generalize — every walk-forward OOS window is zero, the holdout is degenerate, and the optimizer found no config that both trades and wins — compounded by a structurally thin 24-trade sample (monthly roll × SMA gate) whose headline Sharpe is statistically indistinguishable from zero, so it never advanced to risk review.

A long-only single-instrument convex options strategy on ETH Deribit — holding one near-ATM monthly call (opened ~30 DTE, rolled ~7 DTE) but only while ETH closed above its 200-day SMA (bull regime), closing immediately on a regime flip — a 3-parameter, pure-OHLCV-signal companion to the promoted BTC Deribit long-call roll, meant to add asset-diversified convexity and fill the under-covered ETH bucket.

Across 24 real ATM call trades (correctly convex: net vega +113, theta -108, delta +29.5, skew +2.0, tail ratio 4.0) the baseline was flashy — total return +242%, Sharpe 2.96 (CI [-8.44, +7.62]), profit factor 2.13, Sortino 17.9, win rate 37.5% with avg_win 3.5x avg_loss — but concentrated in a few bull months (2024-02 +95%, 2024-05 +103%, 2025-07 +158%; 2023 -5%) with a 61% max drawdown; optimization then failed degenerately: walk-forward OOS Sharpe [0.0, 0.0, 0.0] (is_overfitted=true, PBO 0.53), holdout Sharpe 0.0 / ratio 0 / passed=false, and the optimized re-run produced zero trades across 150 trials.