BtcTrendConfluenceDeribitWeeklyDebitSpread

BTC Trend-Confluence Expressed as Weekly Deribit Defined-Risk DEBIT SPREADS (Cross-Venue: BINANCE 4H+1D momentum signal → DERIBIT BTC weekly call/put vertical spreads, convex directional, bounded MTM, roll weekly)

A CROSS-VENUE, DEFINED-RISK, directional-convexity strategy that transports the factory's ONE proven, currently-promoted edge (BTC Dual-Timeframe Momentum Confluence, 4H primary + 1D confirm, long-short, Sharpe ~1.99 in paper) off the perp and onto the DERIBIT BTC weekly option chain, expressed exclusively as VERTICAL DEBIT SPREADS. Fills the biggest quota gaps: options (5.8% vs >=15%) and cross-venue (7% vs >=15%). Why debit spreads not naked long: the prior naked-long Deribit run was abandoned as a suspected MTM/scaling artifact and never tested its edge; a vertical spread's payoff is BOUNDED by strike width so a mis-scaled mark cannot run away, it halves theta cost, and max loss = net debit (the stop). Why NOT short-vol: the two dead Deribit short-vol runs failed because their payoff is defined by rare tails that ~2yr of chain cannot sample — a DIRECTIONAL spread payoff is NOT tail-dominated, so ~25 months / ~100+ weekly rolls IS diagnostic. Mechanism: each weekly expiry, compute BTC 4H+1D confluence on BINANCE; both bullish -> buy ATM/OTM call debit spread; both bearish -> put debit spread; disagree -> flat. Hold to expiry or take-profit, roll weekly. Fee math: 1wk ATM premium ~3.0-3.5% of spot, OTM short leg finances ~40%, net debit ~1.8-2.2%, max payoff ~2.6-3.2% (R:R ~1.4:1); Deribit fees ~0.06% RT (<1% of capture) — opposite of the fee graveyard. Risk 2.0-2.5% equity per spread (debit=max loss). Sizing in explicit USD/contracts to avoid the coin-vs-USD confusion behind the prior artifact.

Implements the hypothesis literally. (1) Signal transport: the promoted BTC dual-timeframe confluence is reproduced (fast/slow momentum must AGREE in sign, flat on disagreement) on BINANCE 4H bars plus a genuine 1-DAY confirm series, and only the EXPRESSION moves to Deribit — the perp is never traded. Vol-normalised momenta (not absolute % thresholds) keep the firing rate scale-invariant across regimes. (2) Defined risk: the payoff is bounded by the strike width, so a mis-scaled mark cannot run away (the failure mode of the abandoned naked-long Deribit run), and max loss = net debit is the stop — sized directly as risk_frac of equity, in explicit USD contracts, never coin units. (3) Directional, not tail-dependent: ~77 weekly rolls in the 540-day chain (~130 option-leg positions, ~85% of weeks expected to find a confluent entry inside the 48h window) is a diagnostic sample for a spread whose P&L is driven by ordinary weekly drift rather than rare tails. (4) Fee posture: net debit ~2% of spot per contract vs Deribit's 0.03%-per-side fee on premium — costs are a small fraction of the captured move, the opposite of the fee graveyard. Structure: I subclass RollingOptionsStrategy so the factory owns chain resolution and the roll schedule (four legs listed per expiry: ATM+5%-OTM calls and ATM+5%-OTM puts) and this class only decides direction, per-leg BUY/SELL by strike, sizing and exits. No underlying smoke path exists: if the chain is empty the strategy trades nothing. One caveat for the record: this engine charges Deribit fees on the USD premium, whereas the real venue charges 0.03% of the UNDERLYING per contract (capped at 12.5% of premium), so realised fees on a ~2%-of-spot debit will be a few percent of premium higher than the backtest shows — still small against a 5%-wide spread.

Negative expectancy on a diagnostic 114-structure / ~18-month sample (the full real Deribit weekly-chain window): profit_factor 0.898 (<1 and gross-negative, since total_commission is ~$0.002 so fees are not the cause), expectancy -$169.65/structure, total_return -9.67%, Sharpe -0.643, Sortino -0.464, recovery_factor -0.389, losing in both 2025 (-3.9%) and 2026 (-6.0%). Optimization cannot lift a gross-negative directional-options signal above zero with 3 parameters, and the losing side is the mechanism's own CORE expression: the confluence's bullish (call debit spread) leg wins only 28.1% vs 87.7% for puts, so the proven perp momentum edge does not transport onto weekly ATM debit spreads (theta + the ~1.4:1 R:R needs a fast ~5% move the weekly confluence didn't deliver) — a premise flaw, not a tunable parameter. QA's edge concern is borne out on full history, and the honestly-disclosed fee under-modeling (~5-6% of debit unmodeled) means live results would be worse than this already-negative result. The premium-notional avg_trade_return_pct of 7.99% is not fee-comparable and is contradicted by the negative USD expectancy. Not worth 2 hours of optimization. abandon_class=negative_expectancy.

Cross-venue, defined-risk directional options strategy. The BTC dual-timeframe momentum-confluence signal (fast ~2-day momentum on BINANCE BTCUSDT 4H bars plus a 1-DAY confirm series resampled from them on UTC day boundaries, each divided by its own realised-vol scale) is computed on BINANCE and EXPRESSED on the DERIBIT BTC weekly option chain as a vertical DEBIT spread. Each weekly expiry, at the roll entry (expiry - 7d = the prior Friday 08:00 UTC): both horizons bullish -> BUY the ATM call / SELL the ~5% OTM call (bull call debit spread); both bearish -> BUY the ATM put / SELL the ~5% OTM put (bear put debit spread); disagreement -> flat, with the decision retried each 4H bar for 48h before the week is skipped. Both legs are submitted on the same bar (never a naked short) and only when both contracts have a live, non-stale market. Exits: take-profit at 60% of the way from the entry debit to the spread's maximum value (the strike width), an early exit if the confluence flips against the structure, otherwise the scheduled roll close 12h before expiry. Sizing is explicit USD: a Deribit leg is modelled linear with bars carrying the USD premium of one contract (1 BTC), so contracts = equity * risk_frac / net_debit_usd, making the per-trade MAX LOSS exactly 2.2% of equity by construction.

Deliberate override is fine; leaving as-is. No change required unless the base's own min_bars logic is needed.

min_bars_required() overrides a framework method (flagged by static analysis base_shadow). This appears deliberate — it correctly returns the 4H warmup requirement (fast_lookback + vol_bars + 2) — but confirm the base does not rely on its own implementation for gating extra-leg subscription warmup.

Optional: skip the trade instead of flooring when the computed size falls below min_contracts, to keep the defined-risk guarantee strict.

The min_contracts floor (0.1) can override the risk-based size when equity is small or the debit is large, breaking the 'max loss = risk_frac of equity' invariant at the floor. Impact is minor at typical equity levels but means a handful of trades could risk slightly more than 2.2% of equity.

Sandbox smoke test is negative (total_return -9.67%, Sharpe -0.64, PF 0.90 over 114 structures) on one unoptimized draw, so not blocking. Two things for the analyst to weigh on full history: (1) the loss is asymmetric by leg-direction — long/call spreads win only 28% vs 88% for short/put spreads over this window, i.e. the confluence's bullish expression carried the negative P&L here; watch whether that survives optimization or is a sample artifact. (2) Fee under-modeling the developer honestly disclosed: the engine charges Deribit fees on USD premium (commission_pct_of_gross=0.0, total_commission=$0.0023 — essentially nil), whereas real Deribit charges ~0.03% of the UNDERLYING per contract per leg (capped at 12.5% of premium). With 4 fills per round trip on a ~2%-of-spot debit, realized fees are on the order of ~5-6% of the debit/edge — modest but not zero, so live returns will be somewhat worse than the backtest reports. Neither concern is a correctness defect.

Correctness looks sound: 114 defined-risk vertical debit structures, both legs on the same bar, USD-denominated sizing, no naked-short-leg path; the artifact/scaling failure of the prior naked-long Deribit run is genuinely avoided (max_drawdown a bounded 24.8%).

Cross-venue plumbing works: Binance 4H+1D confluence drives Deribit weekly spreads and the sample (114 structures over ~18 months of real chain) is diagnostic for a directional payoff, as the hypothesis intended.

Negative expectancy on a decisive sample: total_return -9.67%, profit_factor 0.898 (<1, gross-negative BEFORE fees since total_commission is ~$0.002), expectancy -$169.65/structure, Sharpe -0.643, Sortino -0.464, recovery_factor -0.389, CAGR -6.83%. Loses in both calendar years (2025 -3.9%, 2026 -6.0%).

Mechanism mismatch on the core expression: the momentum-confluence BULLISH signal (call debit spreads) wins only 28.1% of the time vs 87.7% for the bearish/put side — the strategy's intended trend-following long expression is the source of the loss, confirming QA's edge concern on full history. This is a premise flaw, not a tuning knob.

avg_trade_return_pct 7.99% is measured against option premium notional (not fee-comparable; same signature as prior defined-risk options runs) and is contradicted by the negative USD expectancy.

Developer-disclosed fee under-modeling: real Deribit RT fees (~5-6% of the debit across 4 fills) are essentially absent here, so live returns would be WORSE than this already-negative backtest.

BtcTrendConfluenceDeribitWeeklyDebit-b24d55acb7

This cross-venue strategy aimed to fill the factory's options and cross-venue quota gaps by expressing the promoted BTC momentum-confluence signal as defined-risk Deribit weekly vertical debit spreads, deliberately avoiding the MTM/scaling artifact that killed the prior naked-long Deribit run. The plumbing worked and the defined-risk design held (bounded 24.8% drawdown, no naked short legs), but over 114 structures the edge simply did not transport: the strategy was gross-negative before fees, with the bullish call-spread expression — its own core thesis — winning only 28.1% of the time. The analyst abandoned it at the backtest-review gate as a premise flaw that three optimization parameters could not lift above zero, noting live results would be worse still given disclosed fee under-modeling. It never advanced to optimization, risk review, or paper trading.

A directional momentum edge proven on the perpetual does not automatically transport onto weekly ATM debit spreads, where theta bleed and a ~1.4:1 R:R demand a fast ~5% move the weekly confluence did not deliver — and premium-notional avg_trade_return_pct (7.99%) is not fee-comparable and can mask negative USD expectancy.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate: negative expectancy on a diagnostic sample that was gross-negative before fees (commissions ~$0.002), a premise flaw rather than a tunable one, so it never reached optimization or later stages.

Transport the factory's proven BTC dual-timeframe (4H+1D) momentum-confluence signal computed on Binance onto the Deribit BTC weekly option chain, expressed as defined-risk vertical debit spreads (call spreads when bullish, put spreads when bearish, flat on disagreement) rolled weekly.

Over 114 spread structures across ~18 months of real weekly-chain data it lost money: total return -9.67%, Sharpe -0.643, Sortino -0.464, profit factor 0.898, expectancy -$169.65 per structure, and a 24.85% max drawdown, losing in both 2025 (-3.9%) and 2026 (-6.0%). The bullish call-spread leg won only 28.1% of the time versus 87.7% for the bearish put leg.