MacroWeeklyCrossSectionalMomentumQuintileNeutralLS

Macro Long-Horizon Cross-Sectional Momentum — WEEKLY Bars, Dollar/Beta-Neutral Long-Short: Rank the Liquid Perp Universe by Trailing 24-Week Return (Skipping the Last 4 Weeks), LONG the Top Quintile / SHORT the Bottom, Rebalance Every 4 Weeks (BINANCE USD-M, Market-Neutral, 3-Parameter)

A MARKET-NEUTRAL, MULTI-INSTRUMENT momentum strategy operating at the MACRO (1-week-bar, ~6-month-lookback) horizon — deliberately filling the single most under-represented bucket in the entire portfolio: the macro_1w_plus horizon at 0.6% (15 of 2541 experiments). It uses the ONE edge class this factory has shown survives (momentum) in its empirically MOST-robust form (long-horizon / 12-month-style momentum, the highest-Sharpe momentum variant in every asset-class study), and is a distinct CONSTRUCTION from my in-pipeline daily/monthly momentum factors: weekly bars, a 24-week lookback, a 4-week skip, and 4-week rebalancing produce a much slower, macro-cycle-driven book with a fraction of the turnover and different holding behavior. It stays outside every graveyard family: NOT single-name absolute trend (L62's 0.003-survival class — this is beta-neutral relative), NOT diversified TSMOM (always-in absolute; this is long-short), NOT the falsified cross-sectional REVERSAL (opposite sign, macro horizon, skips the recent 4 weeks where short-term reversal lives), NOT calendar/seasonality, NOT options (which all die on Layer-3 timeouts), NOT funding/OI, NOT the always-in multi-factor composite of L60 (single mechanism). MECHANISM: at the 3-6 month horizon, crypto assets that have led their peers keep leading (persistent capital rotation, adoption/narrative cycles that play out over quarters), and laggards keep lagging; a dollar- and beta-neutral long-leader / short-laggard book at this slow horizon captures that persistence with minimal fee drag and low crash frequency (long-horizon momentum crashes less often than 1-3 month momentum). DATA (named per L61, all present with multi-year Binance weekly history, TIMEOUT-SAFE): BINANCE USD-M 1-WEEK OHLCV for the universe only — trailing return and beta are O(1)-per-bar incremental statistics, NO full-series rescans, NO supplementary feed, NO options surface. Only 3 tunable parameters (lookback_weeks, quintile_fraction, rebalance_weeks).

Implements the hypothesis exactly at its stated macro horizon. (1) WEEKLY bars, 24-week lookback, 4-week skip, 4-week rebalance - the slow, low-turnover construction the hypothesis specifies, distinct from the factory's daily/monthly momentum factors; a quarter-long holding period means the ~0.10% USD-M round trip is a small fraction of the per-trade edge, and only legs whose quintile membership actually changes trade. (2) The 4-week skip removes the short-horizon reversal window, isolating 3-6 month leadership persistence - the empirically most robust momentum variant - and keeps the strategy sign-opposite to the falsified cross-sectional reversal family. (3) Rebalance timing is derived from the bar's OWN timestamp (ISO week index = ts//86400e9//7, fire when week_index % rebalance_weeks == 0), never from a process-relative bar counter, so the rotation grid is identical across restarts, backtests and paper replay. (4) Dollar/beta neutrality: equal-weight k-long / k-short is dollar-neutral by construction; the beta hedge ratio is applied only as a clamped scale on the short leg so the book is beta-neutral without ever carrying material net directional exposure - and, per the -100% neutral-book failure mode, the gross cap is enforced against actual summed open notional under Nautilus netting (a leg-count cap cannot see a stacked netted position), with an explicit flatten-all circuit breaker. (5) Timeout safety: 15 legs x 52 weekly bars/yr, all statistics O(universe x lookback) and computed only at rebalance dates - no full-series rescans, no supplementary feed, no options chain. (6) Venue: BINANCE USD-M is required because the strategy shorts; leverage stays at 1.0 (gross 0.90 <= 1x equity) so no unused-leverage gate is triggered. Universe verified against the catalog: all 15 symbols have complete 1-WEEK BINANCE history (297-351 bars each, 53 bars in the trailing 365-day Layer-3 sandbox window), so with a 29-bar warmup the sandbox still hits ~6 rebalance dates and trades. One note per instructions: the 4-week rebalance means the sandbox's 365-day window leaves only ~24 tradable weeks, so Layer-3 trade counts will be modest by design - the full-history backtest is where this construction has statistical power.

No measurable edge in a 0/84-survival mechanism class (beta-neutral cross-sectional long-short major-perp basket). Sharpe 0.39 with CI straddling zero (low -0.275) is below the 0.5 OOS floor before optimization even runs; information_ratio -0.549. The 61% headline is regime-concentrated in 2024 (+31.4%) and 2026 (+21.3%), with 2022/2023/2025 flat-to-negative — most walk-forward OOS windows will be ≤ 0. Max DD 32.7% and return_kurtosis 70.4 confirm a poor risk profile; PF 1.21 is in the class fail band. Not tunable — the weakness is statistical insignificance + regime dependence in a graveyard class, so optimization would curve-fit the 2024/2026 spikes and die overfit at holdout. Failure pattern: cross_sectional_ls_basket_regime_concentration (class-prior no-edge, Sharpe CI straddles zero, spike-driven returns).

Macro-horizon (WEEKLY-bar) cross-sectional momentum on a 15-symbol universe of liquid, currently-listed Binance USD-M perps (BTC, ETH, BNB, XRP, ADA, SOL, DOGE, LINK, LTC, DOT, AVAX, ATOM, ETC, TRX, NEAR). On a calendar-anchored grid every 4 weeks it ranks every leg by trailing 24-week total return with the most recent 4 weeks SKIPPED (mom = C[t-4]/C[t-28] - 1), then holds an equal-weighted LONG book of the top quintile (3 names) and SHORT book of the bottom quintile (3 names); the middle 9 are flat. The short leg is scaled by a tightly-clamped beta hedge ratio (mean beta of longs / mean beta of shorts vs the equal-weight universe, estimated over the same 24-week window, clamped to [0.80, 1.25]) so the book is beta-neutral while staying within ~11% of gross in net dollar terms. Gross notional is hard-capped at 90% of equity, measured on REALIZED summed open notional Sum(|qty*px|) rather than a leg count, with oversized/stacked/wrong-side legs flattened and a flatten-all safety at 1.25x the cap. Pure OHLCV, no supplementary data, leverage 1.0.

Judge on the full-history backtest, not the sandbox. Require OOS trade count >~100 and reject if walk-forward OOS Sharpe collapses.

Statistically thin Layer-3 sandbox: only 7 trades over 41 data_days, Sharpe 95% CI -2.72..+6.62 (straddles zero), PF 11.86 driven by a handful of fills. Headline metrics are not yet meaningful. Expected for a 4-week-rebalance macro strategy in a 365-day window, and NOT a code-defect signature (healthy Sharpe/PF/win-rate rule out L17 bugs). Full-history weekly backtest (~67 rebalances, a few hundred leg round-trips) is where power lies.

Watch sensitivity/walk-forward for a flat/negative OOS surface like the daily sibling; abandon if the macro horizon does not deliver better OOS robustness.

Pure-OHLCV momentum has low historical survival (L7) and cross-sectional momentum siblings were recently falsified at the daily/monthly horizon. This is a genuinely distinct bucket (weekly bars, macro horizon, low turnover) and momentum is the surviving edge class, so not a re-parameterization — but requires strong, robust OOS proof.

Clean, timeout-safe implementation with careful realized-gross accounting and genuine beta-neutrality (beta -0.017)

Fills a genuinely under-represented horizon bucket; developer was honest that Sharpe ~0.4 is suggestive not established

Sharpe 0.39 with sharpe_ci_low -0.275 — CI straddles zero, already below the 0.5 OOS floor before optimization

information_ratio -0.549: negative risk-adjusted active return vs its own basket benchmark

Regime-concentrated: 2022/2023/2025 flat-to-negative; entire 61% return comes from 2024 (+31.4%) and 2026 (+21.3%) spikes

Max drawdown 32.7% (CI high 65.7%), return_kurtosis 70.4, Sortino 0.44 < Sharpe — fat left tails

Beta-neutral long-short major-perp basket class is 0/84 (multi_instrument) / ~0.01 (market_neutral); PF 1.21 sits in the class fail band

MacroWeeklyCrossSectionalMomentumQuintileNeutralLS aimed to occupy the factory's most under-represented horizon (macro, 1-week bars, ~6-month lookback) by running long-horizon cross-sectional momentum — long the top-quintile leaders, short the bottom-quintile laggards — dollar- and beta-neutral over 15 weekly Binance perps. The implementation was clean and genuinely beta-neutral (beta -0.017), and it posted a 61.4% headline return, but Sharpe was only 0.39 with a CI (low -0.275) straddling zero, information_ratio was -0.549, and max drawdown hit 32.7%. The analyst abandoned it on the first iteration at the backtest_review gate, noting the return was concentrated in 2024 and 2026 spikes while other years were flat-to-negative, and that the beta-neutral long-short major-perp basket class has 0/84 prior survival, so optimization would merely curve-fit those spikes and die overfit at holdout. It never advanced to optimization, analysis, or risk review.

Filling an under-represented horizon bucket with the most robust momentum variant does not rescue the beta-neutral cross-sectional long-short basket class — a 61% headline built from two regime spikes still leaves Sharpe statistically indistinguishable from zero and a negative information ratio against its own benchmark.

It was abandoned at the pre-optimization backtest_review gate on iteration 1: the analyst issued an 'abandon' verdict because Sharpe 0.39 sits below the 0.5 OOS floor with a CI straddling zero, information_ratio is negative, max drawdown fails the 25% limit, and the beta-neutral long-short major-perp basket class has 0/84 prior survival — a graveyard class where optimization would only curve-fit the 2024/2026 spikes.

A market-neutral, macro-horizon cross-sectional momentum factor on weekly BINANCE USD-M perp bars over a 15-name universe, ranking legs by trailing 24-week return (skipping the last 4 weeks) and going long the top quintile / short the bottom quintile with a clamped beta hedge, rebalanced every 4 weeks to fill the under-represented macro_1w_plus horizon bucket.

The backtest returned a headline 61.4% total return with Sharpe 0.39 (CI low -0.275), profit_factor 1.21, win_rate 0.55, information_ratio -0.549, and max_drawdown 32.7% (CI high 65.7%, return_kurtosis 70.4) across 143 trades, and the book was genuinely beta-neutral (beta -0.017). The returns were regime-concentrated, coming almost entirely from 2024 (+31.4%) and 2026 (+21.3%) spikes while 2022/2023/2025 were flat-to-negative.