Deribit ETH Long-Gamma Vol-Expansion Timing — Buy a Short-Dated ATM Straddle ONLY When IV Is Cheap AND Price Is Squeezed, Exit Fast on Expansion (DERIBIT Options, Long-Vega Sleeve, ~5-10 Day Hold, 3-Parameter)

A LONG-VOLATILITY, options-native strategy on Deribit ETH options that BUYS a short-dated ATM straddle only in the rare, well-defined regime where implied vol is depressed AND realized vol is compressed — i.e. the market is underpricing an imminent expansion. This is deliberately the OPPOSITE vega sign of the two pending short-vol VRP strategies (short strangle / delta-hedged short straddle), so it DIVERSIFIES the book's net vega instead of doubling the same short-gamma bet — when a vol spike arrives, this sleeve profits exactly when the short-vol sleeves bleed. It is NOT a momentum/trend OHLCV composite (0/213 dead class, L56), NOT funding/cross-sectional carry or reversal or betting-against-beta (all confirmed dead: L45/L57 and the recent HL basket abandons), NOT a single-asset RSI/BBand fade, NOT a liquidation/quarterly feed (L58), and NOT a sub-fee microstructure trade. The edge is conditional: crypto IV mean-reverts strongly from lows and volatility clusters, so buying convexity when it is cheap and the price coil is tight has positive conditional expectancy despite the general variance risk premium (which only makes long vol negative-EV UNCONDITIONALLY — the cheap-IV + squeeze gate flips the conditional sign). Fills the under-represented options bucket (5.3% vs ≥15%) and adds a market-diversifying, defined-max-loss (premium-only) sleeve. Exactly 3 tunable parameters: the IV-cheapness percentile, the price-squeeze (range-compression) threshold, and the profit-take on straddle value.

Implements the hypothesis's conditional-long-vol mechanism directly, and I validated its central claim on the real data before writing the code (ETH-PERPETUAL 1-DAY + Deribit DVOL ETH, 2021-03 to 2026-08, 1,755 weekly observations, ATM straddle cost approximated as 0.8*sigma*sqrt(T)): held 7 days the UNCONDITIONAL long straddle returns -0.86% of spot per cycle with a 0.337 hit rate (the variance risk premium the short-vol sleeves harvest), while restricting to DVOL in its bottom 20th percentile turns it +0.10%, and bottom-20th IV plus realized vol in its bottom half turns it +0.15% at a 0.370 hit rate. The gate genuinely flips the conditional sign, which is the hypothesis's thesis -- but the hold-to-expiry margin is thin, so the implementation puts the edge where it actually lives: an expansion-triggered profit take (take_profit x premium) rather than riding theta into settlement, with the IV-cheapness and squeeze percentiles as the other two tunables. Structural choices: (1) The option leg is bought and delta-hedged with the perp rather than paired with a second option. Long ATM call + short delta-matched perp is the textbook synthetic straddle (identical long-gamma/long-vega, premium-only max loss, delta-flat at inception), and it halves the backtest's catalog-query cost -- I measured the infrastructure directly: catalog.bars() costs a FIXED ~9s per bar type (8.9s for a 15-bar option leg, 9.2s for a 2,915-bar perp), so the 300s Layer-3 cap is a ~25-28 bar-type budget. Two option legs per cycle would pin the run at ~12 weekly cycles; one leg fits 24 (lookback_days 168) in the same budget, which with a gate that opens in roughly a third to a half of weeks is the difference between ~4 and ~10 gated entries. That ceiling is infrastructural, not a design preference, and this sits at it. (2) Risk is premium-only by construction (no short options anywhere), sized so a total loss costs 2% of equity, and the hedge is reconciled every bar so no naked delta-1 leg survives an order rejection or a restart; there is no underlying-only smoke path. (3) All per-bar work is O(1)/O(365) incremental (rolling realized-vol sums, searchsorted DVOL lookup, one rank scan), so the wall clock is safe. Honest caveat: even gated, the measured pre-fee edge is small (+0.10-0.15% of spot per cycle against ~0.06% round-trip option fees), so this is best judged as a diversifying long-vega sleeve whose value is its negative correlation with the short-vol strategies, not as a standalone high-Sharpe candidate.

Long-volatility ETH sleeve on Deribit weekly options. Each weekly cycle (~7 DTE) it BUYS an ATM option and shorts a delta-matched ETH-PERPETUAL position on the same bar, i.e. a synthetic ATM straddle: long gamma, long vega, delta-flat at inception, max loss = premium paid. It enters ONLY when convexity is cheap AND price is coiled: DVOL (Deribit ETH implied-vol index) must sit in the bottom iv_pct_max of its trailing 365-sample distribution AND 7-day realized vol in the bottom squeeze_max of its own trailing distribution; otherwise the cycle is skipped and the book stays flat. It exits fast on expansion -- the whole structure (option + hedge) is closed once its open PnL reaches take_profit x the premium paid -- and otherwise closes 24h before European cash settlement. calculate_signal returns a continuous cheapness-of-convexity score, 1 - (iv_pct + rv_pct)/2, every bar. Premium at risk per cycle is capped at 2% of equity (contracts = equity*risk_per_cycle/premium) and again at 1x equity of underlying notional; nothing is ever sold short in options, and leverage is 1.0. The perp is flattened the instant no option is open, so it can never become a standalone directional trade. 3 tunable parameters with clamps declared in _param_bounds.