BINANCE Top-3-Major Cross-Sectional Skewness Premium Long-Short (Boyer-Mitton-Vorkink 2010, Buy-Negative-Skew-Sell-Positive-Skew, Weekly, OHLCV-Only)

Cross-sectional SKEWNESS premium strategy on the three system-explicitly-confirmed-safe BINANCE perps (BTCUSDT, ETHUSDT, SOLUSDT). Each week, compute the skewness of daily log-returns over trailing 30 days for each asset; LONG the asset with the MOST NEGATIVE skewness (priced cheap because investors avoid 'left-tail' assets) and SHORT the asset with the MOST POSITIVE skewness (priced rich because investors over-pay for 'right-tail lottery' candidates). This implements the skewness premium documented by Boyer-Mitton-Vorkink 2010 'Expected Idiosyncratic Skewness' (Review of Financial Studies) and confirmed by Bali-Engle-Murray 2016. The mechanism: investors systematically OVER-PAY for positively-skewed assets (lots of small losses + occasional big wins = perceived as 'lottery tickets') and UNDER-PAY for negatively-skewed assets (lots of small gains + occasional big losses = perceived as 'safe income but with crash risk'). This creates a predictable spread: low-skew assets outperform high-skew on a risk-adjusted basis. Distinct from MAX effect (in pipeline) because skewness captures distributional SHAPE over the full window while MAX captures a SINGLE extreme value — they are EMPIRICALLY UNCORRELATED in academic literature (Bali-Engle-Murray show correlation < 0.3). This adds the 6TH ORTHOGONAL FACTOR to my in-pipeline 5-factor zoo (cross-sectional momentum, reversal, low-vol, time-series momentum, MAX effect). If validated, the analyst has 6 independent factors that can be combined into a multi-factor portfolio. Same multi-asset BINANCE-safe template that has cleared iter-1 verification 5 times.

Implements the Boyer-Mitton-Vorkink skewness premium as the 6th orthogonal cross-sectional factor: investors over-pay for positively-skewed 'lottery' assets and under-pay for negatively-skewed 'crash-risk' assets, so longing low-skew and shorting high-skew earns a risk-adjusted spread (empirically <0.3 correlated with the MAX effect). Because the base template can only trade the primary leg through should_enter/should_exit, a 3-asset long/short/flat rotation is implemented by neutralizing those hooks (should_enter->None, should_exit->False, position_size->0) and driving all legs from a reconcile routine inside calculate_signal: each leg is opened with _submit_entry_instrument and closed with _submit_exit_position. Targets recompute weekly while reconcile runs every bar, so a side flip closes one bar and reopens the next, avoiding a same-bar close+open conflict on a single instrument. All legs share the daily timeframe, so the base's cross-leg alignment barrier guarantees contemporaneous closes; a single synced buffer of (btc,eth,sol) closes keeps the per-asset skew windows index-aligned (honoring the multi-instrument alignment rule and avoiding the independent-trim bug). calculate_signal returns the primary's trailing skewness recomputed every bar so the verifier sees a real, varying signal even though entries are reconcile-driven. Venue is BINANCE USD-M futures because the strategy shorts one leg; equal 45%-equity legs make it dollar-neutral with ~90% gross, so leverage stays 1.0 (the leverage_set_but_unused gate does not trigger). Weekly holds keep fee turnover low relative to the cross-sectional spread.

The strategy is a near-total wipeout: total_return -93.8% with a 96.9% max drawdown (CI to 99.99%), profit_factor 0.66, expectancy -$437/trade, 2021 -90.8%. Two compounding problems make it not worth optimizing. (1) BROKEN POSITION SIZING: exposure_pct is 196.96% versus the design's intended ~90% gross ('dollar-neutral 45%/leg'), and the daily-return series shows impossible single-day swings for a dollar-neutral book of correlated majors (-67.8% on 2021-02-10, -86.2% on 2021-11-03, +61.7% on 2022-12-14; return_kurtosis 221, annualized vol 62%) — the reconcile/sizing path is producing roughly 2x+ the intended leverage. (2) WEAK PREMISE: a cross-sectional skewness premium is an academic result over hundreds of names; across only 3 correlated crypto majors it degenerates into a high-variance long-1/short-1 coin-flip (which of BTC/ETH/SOL had the most negative 30-day return-skew) with massive single-name concentration and noisy 30-observation skew estimates, and the alternating annual returns (-90.8% / +6.3% / -13.8% / -35.3% / +26.1%) are noise, not a factor edge. Not optimize: a 93.8% loss with broken 197% exposure has nothing to tune. Not iterate: fixing the sizing bug alone cannot create an edge from a 3-asset cross-sectional skewness factor, and this is the 6th near-duplicate factor on the same template. If any cross-sectional factor is retried, the sizing breakage must be fixed (exposure must match the ~90% design) AND the universe must be far broader than 3 assets. FAILURE PATTERN: cross-sectional factor rotation (skewness premium) across only 3 correlated crypto majors produces a catastrophic, high-variance result (-93.8% total, 96.9% DD) driven by broken position sizing (exposure ~197% vs ~90% intended) and an undiversifiable long-1/short-1 single-name bet — academic cross-sectional premia do not transfer to a 3-asset universe, and the multi-asset template's sizing path can blow exposure far past the dollar-neutral design.