DeribitBtcMonthlyDefinedRiskIronCondorVrpIvRankGated

Deribit BTC Monthly Defined-Risk Short Strangle Harvesting the Variance Risk Premium (IV-Rank-Gated Entry, Protective Wings, Theta/VRP Income, Delta-Bounded, Market-Neutral)

A SHORT-VOLATILITY, DEFINED-RISK options income strategy on Deribit BTC monthly European options that systematically harvests the crypto Variance Risk Premium (VRP) — the persistent, well-documented tendency of option-implied volatility to exceed subsequently-realized volatility. Each cycle, when implied vol is ELEVATED relative to recent realized vol, the strategy sells a delta-balanced out-of-the-money strangle (short OTM call + short OTM put, struck near the ~16-delta / ~1-sigma points) and BUYS further-OTM wings (long call above, long put below) to convert it into a fully DEFINED-RISK iron condor. This caps the catastrophic gamma/tail exposure that destroys naked short-vol books, while still collecting net premium (theta + vol-mean-reversion) over the monthly hold. This is deliberately ORTHOGONAL to everything in the current portfolio: it is NOT a trend/momentum signal (so it cannot suffer the 2024-2026 single-asset trend recent-regime decay that abandoned the BTC funding-confirmed-trend, ER-gated TSM, VWAP-continuation, and dual-TF-confluence strategies), NOT a basket refinement (so it does not repeat the four monotonically-degrading Absolute-Momentum-basket variants), and NOT a sub-1H impulse-continuation (so it is immune to the 15m fee-domination that killed the BTC/ETH 15m impulse strategies). It also fills the single largest portfolio quota gap: options coverage is currently 0.1% (1 of 1196 experiments) against a >=15% target, and the strategy is market-neutral (helping the 87% long-only over-concentration) at a fee-forgiving monthly horizon.

Iteration-2 fix targeting the reported catastrophic loss (-167% single day; -131%/-142% years) which is mathematically impossible for a 2%-capped condor unless position size is inflated. Root cause traced via the sizing identity: netted max loss = width * contracts = equity*max_loss_frac (self-correcting), so a >2% loss requires (a) the long wings not offsetting the shorts AND (b) a collapsed `width` inflating `contracts`. Deribit's coarse away-from-ATM strike grid can snap the 0.10 short and 0.20 wing onto near-adjacent strikes, collapsing `width` and exploding `contracts`; the sibling 2-leg bull-put spread (bc8310ca) caps correctly on the SAME engine, confirming the defect is this strategy's sizing, not the netting engine. Two minimal, additive fixes (everything else, incl. IV-rank gate, DVOL load, leg resolution and smoke path, is unchanged so earlier layers stay green): (1) require BOTH call_width and put_width >= min_width_frac*spot before sizing -- rejects degenerate strike-grid collapses that inflate qty and guarantees each side's breach is capped at its own real wing width; (2) hard-cap contracts by max_gross_frac*equity/spot so that even if a wing fails to net, gross exposure -- and therefore worst-case realized loss -- is bounded, making the -167%/-131%/-142% outcomes impossible. Deribit options are modeled linear-USD here, so width_usd*contracts is the exact USD loss, validating the bound. Venue stays DERIBIT (options, leverage=1, no leverage in sizing so no leverage_set_but_unused).

Not worth optimizing, and cannot be validated with the available data. (1) The metrics are a classic short-vol small-sample artifact: Sharpe 14.29 (CI [5.94, 30.0]) with a near-flat 0.64% max drawdown is implausible and reflects an 11-month sample (Jul-2025..Jun-2026) that apparently contained no wing-breaching vol spike — the tail event that defines a short-premium book's entire risk is absent, so the Sharpe massively understates downside. (2) The sample is structurally too small to optimize: only 11 monthly iron-condor cycles / 11 monthly return observations exist (Deribit option data effectively begins mid-2025 despite the 2018 start_time), so walk-forward's 3 windows would each hold ~3-4 returns and the post-optimization DSR/holdout gates cannot produce a statistically meaningful result — sending it to optimization would fit ~4 tunables to noise and waste 2 hours. (3) Near-zero commission (0.0001 across 44 option legs) suggests option fees are not being materially charged, further flattering the edge (a symptom to verify, not an asserted root cause). This is not an iterate: the strategy already functions correctly with the iteration-2 loss-bounding fixes, and the binding constraints — one year of monthly option history in a single no-crash regime — are not fixable by the developer. The VRP premise is legitimate and worth revisiting ONLY if multi-year monthly Deribit option history spanning at least one vol-crash regime becomes available; as tested it is a benign-window artifact that cannot be validated.

Defined-risk short-volatility iron condor on Deribit BTC monthly options. When DVOL-based IV-rank is elevated (>= iv_rank_min), sells a delta-balanced ~1-sigma OTM strangle and buys further-OTM protective wings, forming a bounded-risk iron condor to harvest the crypto variance risk premium (theta + vol mean-reversion) over a monthly hold. Market-neutral, options-category, monthly (fee-forgiving) horizon.

Strategy functions and matches the hypothesis: 11 monthly iron condors (44 legs, 22 calls / 22 puts), net short vega (-81.96) and net positive theta (+119.22) at entry, avg IV 0.60 — a genuine market-neutral VRP-harvest structure that fills the real 0.1% options coverage gap.

Defined-risk construction with iteration-2 loss-bounding rails (min_width_frac, max_gross_frac) in place; low realized drawdown so far.

Implausible metrics = short-vol small-sample artifact: Sharpe 14.29 (CI up to 30), max_drawdown 0.64%, omega 12.8. A short-premium book looks spectacular only until a wing-breaching vol spike; the 11-month sample apparently contained none, so the numbers grossly understate tail risk.

Sample far too small to optimize or validate: only 11 monthly option cycles / 11 monthly return observations (option data effectively starts mid-2025, not 2018). Walk-forward windows would hold ~3-4 returns each — the post-optimization DSR/holdout gates cannot produce a meaningful verdict.

Single benign regime: all cycles fall in Jul-2025..Jun-2026 with no crash — the exact scenario that defines short-vol risk is absent from the test.

Near-zero commission (0.0001 across 44 option legs) suggests option fees are not being materially charged, flattering the premium-collection edge (symptom to verify).

single benign 2025-2026 window, no vol crash

multi-regime incl. crash

nonzero

Do NOT optimize — the defined-risk implementation is broken, so the metrics are unreliable. The iron condor mechanism is sound and genuinely market-neutral (net delta -0.5, short vega -306, theta +384), but the backtest shows -167% SINGLE DAYS (2022-11-24) and catastrophic annual losses of -131% (2022) / -142% (2023), which are mathematically impossible for a strategy capped at max_loss_frac=2% per condor (~12 condors/yr -> worst case ~-24%/yr). Required fixes: 1. ROOT CAUSE - 4-leg P&L netting. The long protective wings (long call above the short call, long put below the short put) must offset the short legs' losses. The backtest appears to book the short call/put losses without crediting the long-wing gains, so the condor behaves like an unbounded naked short strangle. Verify that at marking AND at expiry the FULL 4-leg structure is netted, so realized max loss = max(call_width, put_width)*contracts - net_credit, bounded near 2% of equity. Cross-check against the 2-leg bull-put-spread sibling (bc8310ca), whose losses ARE correctly capped — the 4-leg accounting should behave the same way. 2. VERIFY SIZING. contracts = equity*0.02/width: confirm `width` is the actual wing width (not e.g. 0), and that qty isn't inflated. A -167% day implies the realized condor loss is ~80x the intended 2% cap — confirm whether that is a sizing error or the wing-netting error in (1). 3. RE-RUN AND RE-JUDGE. Once the defined risk is genuinely bounded, the -167% days and -131%/-142% years should collapse to <= ~-24%/yr worst case. Then assess the true two-sided VRP edge net of the 4-leg fees: if the bounded iron condor is positive with a believable Sharpe (not the current unreliable 2.69) and survives the 2022/2023 stress within its cap, proceed to optimization; if, properly bounded, it is a coin-flip (the two-sided structure gets run over on one side too often to clear costs), abandon. Note the bull-put-spread sibling already works as a one-sided VRP harvester, so this two-sided condor must justify its extra complexity.

This was the factory's serious options attempt — a defined-risk BTC iron condor harvesting the variance risk premium, built market-neutral with iteration-2 loss-bounding rails (min-width and gross-notional caps) and correctly executing 11 monthly cycles of real option legs. On paper it looked extraordinary (Sharpe 14.29, 0.64% max drawdown, profit factor 2.50), but the analyst recognized this as the classic short-vol trap: the 11-month Jul-2025–Jun-2026 sample contained no wing-breaching vol spike, so the metrics understated tail risk, and near-zero charged commissions further flattered the edge. With only 11 monthly return observations in a single benign regime, it could not be meaningfully optimized or validated, so it was abandoned at the backtest-review gate — the VRP idea deemed legitimate but revisitable only when multi-year, multi-regime option history exists. It never reached optimization, risk review, or paper trading.

A short-premium options strategy cannot be validated on a benign, crash-free window — its entire risk lives in the tail, so an implausibly high Sharpe with near-zero drawdown over only 11 monthly cycles is a small-sample artifact; the VRP premise needs multi-year monthly Deribit history spanning at least one vol-crash regime, plus confirmation that option fees are actually charged.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate: with only 11 monthly option cycles in a single benign no-crash regime, the results could neither be optimized nor validated (the downstream DSR/holdout gates would fit ~4 tunables to noise), and the Sharpe grossly understated the tail risk that defines a short-vol book.

A defined-risk, short-volatility iron condor on Deribit BTC monthly options that harvested the variance risk premium — selling a delta-balanced OTM strangle and buying further-OTM protective wings when implied vol was elevated (IV-rank ≥ threshold) — as a market-neutral, fee-forgiving structure to fill the factory's near-empty (0.1%) options coverage quota.

It functioned as intended, trading 11 monthly iron condors (44 option legs, 22 calls / 22 puts) that were net short vega (-81.96) and net positive theta (+119.22) at entry with avg IV 0.60, and posted spectacular-but-implausible metrics: Sharpe 14.29 (CI 5.94 to 30.0), max drawdown 0.64%, profit factor 2.50, omega 12.8, and a 40.9% win rate. Those numbers were flagged as a small-sample artifact — the 11-month window (Jul-2025 to Jun-2026) contained no wing-breaching vol spike, and commissions were near-zero ($0.0001 across 44 legs), flattering the premium-collection edge.