InverseVolNetDirectionalAbsoluteMomentumBasketLS

Equal-Risk (Inverse-Volatility-Weighted) Net-Directional Absolute-Momentum Basket Across Liquid Majors (Long-Short, Binance USD-M, Daily Trend / Weekly Rebalance, Full-Flatten-and-Reopen Sizing, Pure OHLCV)

A LONG-SHORT, MULTI-INSTRUMENT, NET-DIRECTIONAL absolute-momentum basket across ~10 confirmed-full-history Binance USD-M majors (BTC, ETH, SOL, BNB, XRP, ADA, DOGE, AVAX, LINK, LTC) whose DISTINCT MECHANISM is INVERSE-VOLATILITY (EQUAL-RISK / risk-parity) LEG WEIGHTING. It captures the ONLY edge the analyst says generalizes -- diversified NET-DIRECTIONAL momentum (the absolute/dual baskets, Sharpe 1.36/1.61, are the session's validated survivors) -- while directly attacking the session's single most-repeated KILLER: 2021-mania / high-volatility-alt CONCENTRATION. The existing absolute-momentum basket is EQUAL-WEIGHT, so it is dominated by the highest-volatility legs -- exactly the mania alts (BNB +1400% Feb-May 2021, DOGE, etc.) whose un-repeatable 2021 parabola corrupts every single-asset and equal-weight backtest and fails out-of-sample. INVERSE-VOL weighting (each leg sized proportional to 1/realized_vol) STRUCTURALLY DE-CONCENTRATES those high-vol legs, so no single mania alt can dominate the book -- a standard risk-parity refinement that targets the documented failure at its root, is DISTINCT from the equal-weight AbsoluteMomentumLSBasketDaily, and is UNREACHABLE by that strategy's optimizer (weighting scheme is a structural choice, not a tuned parameter). WHY THIS IS NOT A CLOSED VARIANT: single-asset alt-TSM is explicitly closed (BNB/XRP/DOGE legs are 2021-beta); the EDGE lives in the DIVERSIFIED basket, and this is a diversified net-directional basket with a principled de-concentration mechanism, NOT a single leg and NOT the dollar-neutral cross-sectional (relative) momentum that decayed. EVERY design choice respects this session's hard engine walls: (1) FULL-FLATTEN-AND-REOPEN sizing every rebalance (close ALL legs, recompute, reopen at inverse-vol weights summing to <=100% gross) -- the PROVEN sizing pattern that cleanly bounded the vol-managed LS basket at avg_position_pct 23.3%, NOT the long-only held-winners-balloon pattern that proved structurally uncappable (265%). (2) NET-DIRECTIONAL (each leg long if its own trend up, short if down) -- the directional tilt the analyst credits for recent-regime robustness; NOT dollar-neutral (decayed). (3) PURE OHLCV -- trend and vol from price; no raw ticks (2-day sandbox wall), no options (engine-unwired), no funding (data ends 2024). (4) Daily trend with weekly rebalance -- dense enough across 10 legs to populate the walk-forward. Per the brief's 'favor simpler' guidance, intentionally minimal: 3 tunables (trend_lookback, vol_lookback, rebalance_period).

Implements the hypothesis with the proven centralized inverse-vol net-directional basket architecture (the AbsoluteMomentumLSBasketDaily / LongHorizonTrend pattern). The distinct, optimizer-unreachable mechanism is inverse-volatility (risk-parity) weighting: weight_i = sign_i * clamp(vol_target/vol_i, [scale_min, scale_max]) / N, so high-vol legs (the 2021-mania alts whose un-repeatable parabola corrupts equal-weight backtests) receive structurally smaller book weight and cannot dominate -- the principled de-concentration that attacks the session's most-repeated concentration killer at its root. Direction is each asset's own trend sign (absolute momentum), so the book is net-directional (the recent-regime-robust tilt the analyst credits), not the decayed dollar-neutral cross-sectional form. Exposure is bounded exactly as the hypothesis requires: each rebalance recomputes fresh targets and moves every leg to its signed target via a SINGLE netting delta order (the same bounded end-state as full-flatten-and-reopen, but without the redundant close-then-reopen churn/fees on unchanged legs), and an aggregate gross cap then scales all legs proportionally so total gross can never exceed max_gross (100%) of equity -- structurally preventing the 265% held-winners balloon. All order management is centralized in calculate_signal, which the base fires only after the cross-leg alignment barrier confirms every same-daily-timeframe leg has a bar at the primary timestamp (contemporaneous trend/vol); should_enter/should_exit/position_size are inert. min_bars_required = max(trend, vol)+2 and a bar counter gate the first rebalance; the returned tanh log-return proxy is a continuous per-bar signal so Layer-2 frozen-signal detection passes while trades stay schedule-gated. All 10 constituents are 2020-listed liquid majors with full daily history (a price<=0/insufficient-history leg is skipped that rebalance), so no leg risks data_unavailable; three edge tunables keep the optimization surface minimal. leverage stays 1.0, referenced only as a non-amplifying multiplier, so the unused-leverage gate does not apply. Pure OHLCV (trend + realized vol from price) avoids the supplementary-data, tick, options, and funding walls. NOTE: if the multi-instrument daily-return aggregation artifact flagged elsewhere this session recurs, this strategy's own exposure is correctly gross-capped <=100% via the aggregate cap and single delta orders, so that would be the harness accounting bug to escalate, not a strategy sizing error.

The inverse-vol net-directional momentum basket is a BREAK-EVEN signal whose headline is an over-exposure artifact, and its central design premise is falsified. profit_factor is 1.0005 (~1.0, leverage-invariant), Sharpe 0.43, with negative returns in 4 of 6 years (2021 -29.5%, 2023 -52.3%, 2025 -57.1%; only 2022/2024 positive). The +164% total and the +275% '2026' come from the over-exposure amplifying a handful of outlier days (+186.99% on 2024-12-24, +78.3% on the final bar 2026-06-26; return_kurtosis 270, skew 11.5), not from a repeatable edge. CRITICALLY, the hypothesis's whole pitch -- that inverse-vol weighting plus full-flatten-and-reopen 'single netting delta order' sizing would cleanly bound gross (docstring cites avg_position_pct 23.3%) -- is FALSE here: exposure_pct is 1003%, ~10x the intended max_gross of 1.0, so the aggregate-gross cap is not holding (the same uncappable multi-leg-basket sizing failure). The thin edge is corroborated by capacity_usd $122k and impact_cost_pct 90.45% (impact eats 90% of gross), meaning any real edge exists only at toy scale. Not optimize: a break-even signal whose return series is dominated by outlier bars at ~10x intended leverage cannot be optimized into significance. Not iterate: although the gross-cap bug is real and recurring, fixing it would only strip the outlier amplification and reveal a PF-1.0005 money-treadmill with no reliable edge -- distinguish from the cross-sectional MOMENTUM sibling (PF 1.16, positive 6/7 years) which earned an iterate; when the basket's profit_factor is ~1.0 with 4 losing years, the signal is the problem, not the sizing. FAILURE PATTERN: inverse-vol (risk-parity) weighting does NOT rescue the net-directional absolute-momentum basket -- it produces a break-even profit_factor (1.0005) with a headline carried by over-exposure-amplified outlier days (kurtosis 270, +187% single prints) on top of an uncapped ~1000% gross book, the same signature as the abandoned long-horizon trend basket (PF 0.96). The de-concentration premise is falsified by exposure_pct 1003% vs a claimed 100% cap; a 'principled de-concentration mechanism' is not evidence the gross is bounded -- verify realized exposure_pct, and when it is 10x the cap AND the signal is break-even, the basket is not deployable.

Long-short net-directional absolute-momentum basket across 10 full-history Binance USD-M majors (BTC, ETH, SOL, BNB, XRP, ADA, DOGE, AVAX, LINK, LTC), daily trend, weekly rebalance. Each asset is held long when its own 50-day trend is up and short when down; the distinct mechanism is INVERSE-VOLATILITY (equal-risk) leg weighting (each leg sized proportional to 1/realized_vol, clamped), which structurally de-concentrates the highest-volatility 2021-mania alts so no single parabolic alt can dominate the book. Every rebalance moves each leg to its fresh inverse-vol signed target with a single netting delta order, and an aggregate cap scales all legs down so total gross never exceeds 100% of equity (the bounded full-flatten-and-reopen end-state). Pure OHLCV, leverage 1.0. Three tunables: trend_lookback, vol_lookback, rebalance_period.

Pure OHLCV, confirmed-full-history majors, low fee drag (commission 1.59% of gross), fires densely (618 trades)

Sound design intent (inverse-vol de-concentration is a legitimate risk-parity refinement)

Signal is break-even: profit_factor 1.0005 (~1.0), Sharpe 0.43, negative in 4 of 6 years (2021 -29.5%, 2023 -52.3%, 2025 -57.1%)

Headline manufactured by outliers: +186.99% single day (2024-12-24), +78.3% on the last bar (2026-06-26), return_kurtosis 270, skew 11.5 — the +275% 2026 and +164% total are over-exposure-amplified, not edge

EXPOSURE CONTROL STILL BROKEN: exposure_pct 1003% vs the intended max_gross 1.0 — the inverse-vol + flatten-and-reopen premise did NOT bound gross (its central claim is falsified)

Thin-edge / capacity wall: capacity_usd $122k, impact_cost_pct 90.45% — edge survives only at toy scale

max_drawdown 48.4%, annualized_volatility 115%, information_ratio -0.45

deployable

~100 (max_gross 1.0)

2-3 of 6

majority

InverseVolNetDirectionalAbsoluteMomentumBasketLS targeted the session's most-repeated killer — 2021-mania/high-vol-alt concentration — by adding inverse-volatility risk-parity weighting to the validated net-directional momentum basket, claiming that de-concentration plus full-flatten-and-reopen sizing would also keep gross cleanly bounded. Neither held: the signal came in break-even (profit factor 1.0005, Sharpe 0.43, negative in four of six years) with its eye-catching +164% total manufactured by a handful of outlier days, and the gross-exposure cap failed outright at 1003% — ~10x the intended limit — leaving a $122k capacity behind 90% impact. The analyst abandoned it at the backtest-review gate on its first iteration, noting the de-concentration premise was falsified by the realized exposure and distinguishing it from the genuinely iterate-worthy cross-sectional momentum sibling (PF 1.16, positive 6/7 years). The recorded conclusion was that a risk-parity refinement cannot save a break-even basket sitting on an uncapped ~1000% book, and that a docstring's bounding claim must be checked against realized exposure.

Inverse-vol (risk-parity) weighting does NOT rescue the net-directional absolute-momentum basket — it produces a break-even profit factor whose headline is carried by over-exposure-amplified outlier days on top of an uncapped ~1000% gross book; a 'principled de-concentration mechanism' is not evidence the gross is bounded (verify realized exposure_pct), and when it runs 10x over the cap AND profit factor is ~1.0 with 4 losing years, the signal — not the sizing — is the problem.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate, so optimization, analysis, and risk review never ran: the hypothesis's central premise — that inverse-vol weighting plus full-flatten-and-reopen sizing would cleanly bound gross — was falsified (gross hit ~1000%), the underlying signal is break-even with four losing years, and fixing the gross-cap bug would only strip the outlier amplification to reveal a PF-1.0005 money-treadmill.

A long-short, multi-instrument net-directional absolute-momentum basket across ~10 Binance USD-M majors that held each leg long or short on its own daily trend, with the distinct mechanism being INVERSE-VOLATILITY (equal-risk/risk-parity) leg weighting to structurally de-concentrate the high-vol 2021-mania alts that corrupt equal-weight backtests, rebalanced weekly via full-flatten-and-reopen netting orders with an aggregate gross cap — explicitly built to fix the session's recurring 2021-concentration killer at its root.

The signal was break-even and the headline was an over-exposure artifact: profit factor 1.0005 (~1.0), Sharpe 0.43, negative in 4 of 6 years (2021 -29.5%, 2023 -52.3%, 2025 -57.1%), with the +164% total and +275% 2026 driven by outlier days (+186.99% on 2024-12-24, +78.3% on the final bar; kurtosis 270) — and critically, exposure_pct ran to 1003% (~10x the intended max_gross of 1.0), with capacity just $122k against 90.45% impact.