MarketNeutralMomentumVsBtcHedgeLS

Market-Neutral Momentum-vs-Market, Long Outperforming Majors / Short BTC Hedge (BINANCE USD-M, Hold the Majors Beating BTC by a Margin Over 30 Days + Short Equal-Notional BTC to Strip Market Beta — Isolates the Momentum ALPHA That Survives Bear Regimes, Monthly, Low-Turnover, 3-Parameter)

A MARKET-NEUTRAL momentum strategy that takes the factory's ONE robust, deployed factor — cross-sectional momentum (Sharpe 2.63 long-only) — and fixes the flaw that kills every long-only strategy here: single-regime dependence (all their return is the 2021 bull; they get rejected as artifacts). The deployed momentum rotation is long-only, so its return is mostly BTC beta plus a momentum tilt. This strategy LONGS the majors that are beating BTC over the past 30 days AND SHORTS equal-notional BTC to neutralize market beta, leaving the pure momentum ALPHA — the tendency of strong majors to keep outperforming the market — which is earned in BULL AND BEAR regimes alike (in a bear, the shorts-BTC leg profits while resilient momentum names fall less). This directly attacks the single-regime rejection that no long-only strategy escapes. It is deliberately NOT the L24-flagged construction: L24 forbids re-weighting relative strength in a LONG-ONLY rotation duplicating the promoted basket; this is a MARKET-NEUTRAL long/short with a single stable BTC hedge leg — a different direction bucket and a different return stream (alpha, not beta). It is NOT the L12 rank-and-rotate dispersion trade (long-top/short-bottom, 2N churning fills): the short leg is ONE instrument (BTC), resized not rotated, so incremental turnover is tiny. Fee math clears the L12 hurdle by a wide margin: majors that qualify (beating BTC by the entry margin over 30d) have historically outperformed BTC by 5-15% over the following month — hundreds of bps versus a ~0.10% per-changed-leg cost on a MONTHLY rebalance. It fills the under-represented long_short/market-neutral direction bucket (14.9%) and is pure OHLCV (best-surviving data tag), 3-parameter (no selection-surface overfit — a fixed absolute margin, not a best-of-N optimizer sweep).

Implements the hypothesis directly: take the factory's robust momentum factor and strip the BTC beta that makes every long-only version a single-regime (2021-bull) artifact, isolating the momentum ALPHA that survives bear regimes via a single short-BTC hedge. It is deliberately NOT the L24 long-only relative-strength rotation (this is market-neutral long/short — a different direction bucket and return stream, alpha not beta) and NOT the L12 rank-and-rotate dispersion trade (the short leg is ONE instrument resized incrementally, not N churning shorts; the long side is delta-only, so turnover is a handful of fills per month). Fee math clears the hurdle by a wide margin: qualifiers (beating BTC by ≥5% over 30d) target hundreds of bps of continued outperformance over the next month vs ~0.10% per-changed-leg on a monthly cadence. Overfit is minimized: 3 parameters, a fixed absolute margin gate (not a best-of-N optimizer sweep), max_names is a risk cap not a selection surface. Venue is BINANCE USD-M futures because the strategy shorts BTC (leverage 1.0, referenced in every notional so no leverage_set_but_unused; gross bounded to ~0.96× equity to fit the 1× margin). Correctness: all legs are daily so the base class defers the primary (BTC) signal until every same-timeframe extra leg is contemporaneous (no stale-leg misalignment); the BTC hedge is sized off the TARGET long set (not just-submitted-but-unfilled positions), so fill timing can't desync the hedge; a signed-quantity delta adjusts the single BTC short so it is resized rather than rotated. Data: pure OHLCV daily majors are collected from genesis (full multi-year, multi-regime history — the pure-OHLCV window issue that hit the OI basket was its supplementary data, not daily bars), so the 32-bar warmup is a small fraction of the window. A dry-run over ~2 years produced 24 monthly rebalances with 30 long entries / 29 exits / 24 BTC resizes — low turnover, always ≥1 trade (no verification_loop/no-trades risk), qualifiers firing regularly. Fills the under-represented long_short/market-neutral direction bucket on the best-surviving pure-OHLCV tag.

The backtest falsifies the hypothesis's own central claim, and it is the same single-regime 2021 artifact every majors-momentum sibling dies to. annual_returns: 2021 +231.5%, then 2022 -0.58%, 2023 +1.35%, 2024 +6.5%, 2025 -4.76%, 2026 -0.04% — the entire 284% total_return is 2021 (a few months: 2021-03 +87.7%, 2021-06 +46.2%), and the strategy is flat-to-negative for five straight years afterward. The core thesis — that shorting BTC isolates a momentum alpha 'earned in bull AND bear' — is empirically false here: the bear years it was designed to profit in (2022, 2025) are negative. The BTC hedge stripped beta but there is no regime-robust alpha underneath. Corroborating the outlier/single-regime read: Sharpe 0.46 (< 0.5) with sharpe_ci_low -0.42 (CI straddles zero), win_rate 0.34, information_ratio -0.47, return_skew 29.7 and kurtosis 1030 (a handful of 2021 months carry everything), plus L17 lumpy-monthly annualization inflation (annualized_vol 57.6%) and end_unrealized_pct 17.9% (part of the headline is unrealized MTM). This is not a tuning problem: optimizing lookback/entry_margin over a sample whose edge exists only in 2021 will fit 2021 into the in-sample window and collapse on OOS/holdout, exactly as the Low-Vol and Efficiency-Ratio majors-basket siblings did today — there is no post-2021 edge to tune toward. Abandon at review rather than spend the 2-hour optimization. META-NOTE FOR RESEARCH LEAD: making a long-only majors-momentum strategy market-neutral did NOT escape single-regime rejection, because the momentum signal itself only paid in the 2021 bull; the beta hedge cannot manufacture alpha that isn't in the cross-section. Cross-sectional majors momentum vs BTC shows no persistent alpha post-2021 at monthly frequency — reframe away from majors-momentum entirely rather than re-hedging the same 2021-dependent signal.

Market-neutral cross-sectional momentum on BINANCE USD-M perps, daily bars, monthly rebalance. Primary = BTCUSDT.BINANCE (the benchmark and the single short hedge leg); extras = eight liquid majors (ETH, SOL, BNB, XRP, ADA, AVAX, LINK, DOGE) as candidate longs. Each month it computes every major's trailing 30-day return minus BTC's 30-day return; majors beating BTC by at least entry_margin (5%) qualify, capped at max_names (4) taken by largest margin. It LONGS each qualifier at a fixed per-name notional (per_name_frac × base) with delta-only turnover — only names entering/leaving the set trade, persisting names are held untouched — and RESIZES a single short BTC leg (incremental delta order, not a rotation) to equal the total target long notional, stripping market beta so only momentum alpha remains (earned in bull and bear). The signal is the cross-sectional mean relative momentum, continuous and varying every bar. Sizing is path-independent capped (min(base_capital, equity) × per_name_frac), bounding gross to ~0.96× equity (≤4 longs × 0.12 + equal BTC short) within leverage 1.0. Three tunable knobs: lookback, entry_margin, rebalance_period.

Analyst: decompose returns BY REGIME and confirm 2022 (bear) was positive — if all-2021, it's the exact single-regime artifact the hypothesis claims to fix. Developer (follow-up): size the BTC short to the CURRENT market value of the long book, not entry notional, so neutrality holds as winners appreciate.

HEDGE-NEUTRALITY FIDELITY (the #1 analyst check, bearing on the core claim). The BTC short is sized to the long book's ENTRY notional (len(target) x per_name_notional), but persisting longs are held at their APPRECIATED market value (delta-only holds them untouched). So the book drifts NET-LONG between monthly rebalances as winners run, partially re-introducing the market beta this strategy exists to strip. The thesis is 'strip beta -> survive bear regimes -> escape the single-regime rejection', so an under-sized hedge leaking bull beta directly threatens that claim. Measured beta is low (0.12) but is a period average and doesn't rule out that +284% is concentrated in bull months where the book was effectively net-long.

Analyst: recompute Sharpe/vol/exposure from a clean daily equity curve on the full walk-forward; don't promote on the sandbox Sharpe 3.56 alone.

Return-series risk metrics are artifact-laden: exposure_pct 262% and annualized_volatility 303% (inflated by appreciated positions / lumpy equity), data_days 43 is impossible given 51-day holds and a 960-day max-DD duration (run spans years), and Sharpe CI [0.05, 5.98] has a lower bound essentially at zero. Trade-level facts (total_return +284%, PF 2.45, max_dd 20%, 106 trades, commission 0.41% of gross) look strong, but the ratio metrics need recomputation.

Analyst: check robustness to removing the top 2-3 winners; verify positive OOS Sharpe in the later (2022-2023) walk-forward windows.

Return concentration: skew 5.02, kurtosis 29.2, largest_win $128,363 vs avg_win $12,996, win_rate 0.349. The edge hinges on a few very large winners — characteristic of momentum but a regime/concentration risk that compounds concern #1 (alpha across regimes vs a few bull outliers).

Genuinely market-neutral construction (single resized BTC short, not N churning shorts) in an under-represented direction bucket; low fee drag (commission 0.45% of gross), clean delta-only rebalance

max_drawdown 19.9% is within the 50% floor; capacity is large

Falsifies its own thesis: return is single-regime 2021 (+231.5%) with 2022-2026 flat-to-negative — the exact single-regime dependence it claimed to fix

The 'alpha in bull AND bear' claim is empirically false: 2022 -0.58%, 2025 -4.76% (negative in the bear years it was meant to profit)

Sharpe 0.46 (< 0.5), sharpe_ci_low -0.42 (CI includes 0), win_rate 0.34, information_ratio -0.47

Outlier-dominated: return_skew 29.7, kurtosis 1030 — a few 2021 months carry the entire result; L17 lumpy-monthly annualization inflation (annualized_vol 57.6%)

end_unrealized_pct 17.9% — part of the headline is open MTM, not realized

2022/2025 negative

positive

2021 = 100% of return

multi-regime

This strategy set out to fix the flaw that kills every long-only momentum book here — single-regime (2021-bull) dependence — by longing majors that beat BTC over 30 days and shorting equal-notional BTC to isolate the momentum alpha it claimed was earned in bull and bear alike. The construction was genuinely market-neutral and cheap (single resized BTC short, 0.45% commission drag, 19.9% max drawdown), and the headline looked strong at +284% total return. But the annual breakdown falsified its own thesis: 2021 alone contributed +231.5% while every following year was flat-to-negative, including the very bear years (2022, 2025) it was meant to profit in, and the quality metrics (Sharpe 0.46 with a zero-straddling CI, win_rate 0.34, IR -0.47, kurtosis 1030) confirmed a handful of 2021 months carried everything. The analyst abandoned it at review rather than spend a 2-hour optimization on a 2021-only edge, noting to the Research Lead that hedging beta cannot manufacture cross-sectional alpha that isn't there — majors-momentum vs BTC has no persistent post-2021 edge at monthly frequency.

Making a long-only majors-momentum strategy market-neutral does not escape single-regime rejection — a beta hedge cannot manufacture alpha that isn't in the cross-section, and cross-sectional majors-momentum vs BTC shows no persistent alpha post-2021 at monthly frequency, so the research direction should reframe away from majors-momentum entirely.

The analyst issued an 'abandon' verdict at the pre-optimization backtest review, concluding the backtest falsified the hypothesis's own central claim: the bear years it was designed to profit in (2022, 2025) were negative, so the BTC hedge stripped beta but left no regime-robust alpha — the same single-regime 2021 artifact every majors-momentum sibling dies to. It was judged a non-tuning problem (optimizing over a 2021-only edge would collapse OOS), so optimization and later stages were never reached.

A market-neutral momentum strategy on BINANCE USD-M perps (daily bars, monthly rebalance) that went long the majors beating BTC by an absolute margin over 30 days and shorted equal-notional BTC to strip market beta — aiming to isolate a pure momentum alpha that (per its thesis) survives both bull and bear regimes and thereby escape the single-regime rejection that kills long-only momentum here.

The backtest returned +284.3% total with a 25.8% CAGR and 19.9% max drawdown, but the quality metrics were weak: Sharpe 0.46 (CI straddling zero at [-0.42, 0.86]), win_rate 0.34, information_ratio -0.47, and extreme outlier dominance (return_skew 29.7, kurtosis 1030). Annual returns show nearly the entire result came from 2021 (+231.5%), with 2022 -0.58%, 2023 +1.35%, 2024 +6.5%, 2025 -4.76%, and 2026 -0.04%.