DollarNeutralMomentumSpreadLS

Dollar-Neutral Cross-Sectional Momentum SPREAD, Long-Short Market-Neutral (BINANCE USD-M, 8 Majors, MONTHLY Long Top-2 / Short Bottom-2 by Relative Strength — Dollar-Neutral So Max-Drawdown Is Bounded by the Momentum SPREAD (~20-30%), NOT the 50%+ Market Beta That Floor-Killed Every Long-Only Momentum Variant; Monthly Cadence Keeps the 4-Leg Fee Below the Spread, 2-Parameter)

A REAL, codeable strategy that resolves the EXACT binding constraint proven across my momentum reuses this session: every LONG-ONLY momentum construction (crowding-filter 63% DD, dispersion-gate 67% DD, mid-cap 74% DD) died on the >50% drawdown floor because it holds full crypto BETA through the 2022 bear, and no signal filter protects that tail. The only structural fix is to REMOVE beta: a DOLLAR-NEUTRAL long-winners / short-losers construction has net market exposure ≈ 0, so its drawdown is bounded by the momentum SPREAD's volatility (historically ~20-30% for majors), which is BELOW the 50% floor by construction — the thing filters and partial hedges could not deliver. It captures the FULL cross-sectional momentum premium (top-minus-bottom spread), not the thin residual that killed my fully-beta-hedged overlay. To beat the fee death that L12 flags for dollar-neutral rank baskets, it is deliberately MINIMAL and LOW-turnover: only 4 legs (long top-2 / short bottom-2 of 8 majors) rebalanced MONTHLY. FEE ARITHMETIC (per L12): 4 legs × monthly rebalance ≈ 8 fills/month × ~0.05% ≈ 0.4%/month; the top-2-minus-bottom-2 monthly momentum spread among majors historically runs ~3-8%/month — clearing fees by a wide margin, unlike the high-turnover weekly N-name rank baskets L12 killed. Distinct from all my pending momentum variants (those are long-only or partially-hedged; this is FULL dollar-neutral top-vs-bottom). Reuses the proven relative-strength signal (no L30 new-indicator risk). Long-short, market-neutral, monthly, USD-M. 2 core parameters (momentum lookback, names per side).

Resolves the exact binding constraint proven across this session's momentum reuses: every LONG-ONLY construction (crowding-filter 63% DD, dispersion-gate 67% DD, mid-cap 74% DD) blew the >50% drawdown floor because it carries full crypto BETA through the 2022 bear, and no signal filter protects that tail. The structural fix is to REMOVE beta via a DOLLAR-NEUTRAL long-winners/short-losers spread: equal long and short notional -> net exposure ~0 -> drawdown bounded by the momentum SPREAD's volatility (~20-30% historically for majors), below the 50% floor BY CONSTRUCTION, while capturing the FULL top-minus-bottom premium (not the thin residual the full beta hedge left). Fee discipline per L12: only 4 legs (top-2 long / bottom-2 short of 8) rebalanced MONTHLY (~8 fills/month x ~0.05% ~= 0.4%/month) vs a monthly top2-minus-bottom2 majors spread of ~3-8%/month — clearing fees by a wide margin, unlike the high-turnover weekly N-name rank baskets L12 killed. Mechanism in _rebalance: rank by momentum, long top-N, short bottom-N (kept disjoint), and per instrument reconcile to the desired long/short/flat side, flipping (close then re-open opposite) when a name crosses from winner to loser; same-side holdings are kept to limit churn. Reuses the proven relative-strength signal (no L30 new-indicator risk) and is distinct from all pending variants (those are long-only or partially-hedged; this is FULL dollar-neutral top-vs-bottom). VENUE: USD-M futures is required (the strategy SHORTS the losers), matching the deployed momentum family. SIZING/LEVERAGE: each leg 0.22 of equity, 4 legs -> gross ~0.88 of equity with net ~0, so leverage stays 1.0 and is genuinely unused (no leverage-unused gate, no liquidation-prone amplification; a market-neutral book cannot experience the -100% beta blowups). Purely price-driven (no supp/tick data) so it avoids cross-instrument-supp and tick-timeout pitfalls and is timeout-safe (bounded per-bar work: 8 O(1) momentum reads; rebalance monthly). All trading happens in _rebalance via _submit_entry_instrument/close_position; the default primary entry path is neutralized (should_enter None, should_exit False, position_size 0) so the base never double-trades BTC. All 8 majors are deep-history liquid USD-M perps, so every leg has data (no data_unavailable). 2 core parameters (momentum_lookback, names_per_side) shared across names — not per-name tuned — to resist overfit.

The dollar-neutral thesis is catastrophically falsified. It claimed a long-winners/short-losers spread would bound max_drawdown at the ~20-30% momentum-spread volatility 'by construction' by removing beta — but realized max_drawdown is 97.62% (CI to 99.99%) and total_return is -96.35%, a near-total wipeout. profit_factor 0.685, avg_trade_return_pct -2.77%, and expectancy -$488/trade confirm a decisive net loser. The cause is the flaw the hypothesis overlooked: shorting the BOTTOM-2 momentum crypto majors carries UNBOUNDED squeeze risk — a beaten-down 'loser' that violently reverses and 5-10×'s costs the short 400-900%, producing the impossible monthly swings visible in the data (+1818% in 2021-10, -213% in 2024-01, -152% in 2024-11, -1023% for 2024). avg_position_pct 107% shows the book actually runs leveraged gross rather than the benign spread claimed, and returns collapse to 0.0 after late 2024 because the account was effectively destroyed. Removing market beta did NOT bound the tail — the short legs delivered the opposite, exactly the L12 dollar-neutral rank-basket failure and the same unbounded-short-squeeze mechanism that killed the directional short basket (-79%) earlier this session. Optimizing momentum_lookback/names_per_side cannot fix unbounded short-squeeze risk on a -96%-return, 97%-DD, PF-0.68 loser; tuning would only shift the blowup. Abandon.

Dollar-neutral cross-sectional momentum SPREAD, long-short market-neutral on 8 BINANCE USD-M majors (BTC/ETH/SOL/BNB/XRP/ADA/AVAX/LINK, 1-DAY bars). MONTHLY it ranks the 8 by trailing relative strength (return over momentum_lookback days), goes LONG the top-2 and SHORT the bottom-2, equal notional per leg, so net market exposure ~0 and drawdown is bounded by the momentum spread's volatility (~20-30%) rather than crypto beta. Captures the full top-minus-bottom momentum premium. Only 4 legs rebalanced monthly to keep fees far below the spread. Pure-price, no supp data, no ticks. 2 core parameters: momentum_lookback, names_per_side.

Reopen the flipped leg on the next rebalance, or skip the orders_open guard for the reopen half of a flip (the close is reduce-only and won't conflict).

Flip path (a name crossing winner->loser in one rebalance) closes-then-reopens on the SAME bar, but _open_leg's `if self.cache.orders_open(iid): return` guard can see the just-submitted reduce-only close order and skip the new opposite leg, leaving that name flat until next month and momentarily un-balancing dollar-neutrality. Empirically not biting — sandbox shows balanced 16 long / 15 short, beta 0.04 — so latent risk, not an active break.

Analyst: judge on the full multi-year backtest; confirm measurable sample.

Marginally negative, inconclusive one-year sandbox: total_return -2.79%, avg_trade_return_pct -0.382%, PF 0.929, Sharpe -0.14 (CI [-1.75, 1.48] straddles zero). NOT the L17 code-defect signature — reads as a thin/absent spread over one window, not a bug. Trade count 31 is structurally low (monthly cadence by design); needs the full multi-year window (incl. 2022) to test the beta-removal thesis and clear the measurability floor.

Genuinely market-neutral construction (beta 0.11, correlation 0.07) and low-turnover monthly cadence as designed

Well-sampled (213 trades) so the catastrophic verdict is decisive

Core thesis FALSIFIED: max_drawdown 97.62% (CI to 99.99%) and total_return -96.35% — a near-total wipeout, versus the claimed ~20-30% bounded drawdown

Decisive net loser: profit_factor 0.685, avg_trade_return_pct -2.77%, expectancy -$488/trade

Short legs detonate: impossible monthly swings (+1818% 2021-10, -213% 2024-01, -152% 2024-11, -1023% 2024 annual) from unbounded short-squeeze losses on bottom-momentum majors

avg_position_pct 107% — runs leveraged gross, not the benign dollar-neutral spread claimed; returns go to 0.0 after late 2024 (account destroyed)

DollarNeutralMomentumSpreadLS was the session's final attempt to fit the deployed momentum signal under the 50% drawdown floor, this time structurally by going dollar-neutral (long top-2 / short bottom-2 of the majors, monthly) so that net beta near zero would supposedly bound drawdown to the momentum spread's volatility. The construction was genuinely market-neutral (beta 0.11), but the outcome was the opposite of the thesis: a -96.35% return and 97.62% max drawdown, with a 0.685 profit factor and -$488/trade expectancy. The cause was the overlooked flaw of shorting the lowest-momentum crypto majors — unbounded squeeze risk — which produced impossible monthly swings and destroyed the account by late 2024 while running 107% leveraged gross. The analyst abandoned it as a catastrophic falsification, noting it reproduced the same unbounded-short-squeeze failure that killed the directional short basket (-79%) earlier in the session, and that tuning could only shift the blowup rather than fix it.

Removing net market beta does not bound the tail of a crypto long-short book, because shorting the lowest-momentum names carries unbounded short-squeeze risk — a violent reversal in a 'loser' can cost the short several hundred percent, so dollar-neutrality guarantees neither low drawdown nor a positive edge.

It was abandoned at backtest-review (verdict: abandon) because the core thesis was falsified — the short legs carried unbounded short-squeeze risk, so beaten-down 'loser' majors that violently reversed produced impossible monthly swings (+1818% in 2021-10, -1023% for 2024) that wiped out the account, and avg_position_pct of 107% showed it ran leveraged gross rather than the benign spread claimed.

It reused the proven cross-sectional relative-strength momentum signal in a dollar-neutral long-short form — monthly long the top-2 and short the bottom-2 of 8 USD-M majors — on the thesis that removing market beta would bound max drawdown to the momentum spread's ~20-30% volatility (below the 50% floor) while a minimal 4-leg monthly cadence kept fees under the spread.

It achieved genuine market-neutrality (beta 0.11, correlation 0.07) but was a catastrophic net loser: over 2,397 days and 213 trades it returned -96.35% with a 97.62% max drawdown (CI to 99.99%), profit factor 0.685, avg_trade_return_pct -2.77%, expectancy -$488/trade, and returns collapsed to zero after late 2024.