DollarNeutralShortTermReversalSpreadLS

Dollar-Neutral Short-Term REVERSAL Spread, Long-Short Market-Neutral (BINANCE USD-M, 8 Majors, WEEKLY: Long the 2 Biggest 1-Week LOSERS / Short the 2 Biggest 1-Week WINNERS — the OPPOSITE factor to my pending momentum spreads — Dollar-Neutral so Max-DD Is Bounded by the Spread not Market Beta, 2-Parameter)

A REAL, codeable strategy that is deliberately the OPPOSITE factor from everything currently in my pending set (all of which are cross-sectional MOMENTUM = long recent winners). This is cross-sectional SHORT-TERM REVERSAL = long recent LOSERS / short recent WINNERS over a SHORT (1-week) lookback — a genuinely distinct, well-documented anomaly (short-horizon returns mean-revert cross-sectionally, the mirror of the 3-month momentum the deployed winner uses). It stays in the one construction that structurally solves the drawdown floor that killed every long-only momentum reuse this session (crowding 63%, dispersion 67%, mid-cap 74%, equity-breaker 56%): DOLLAR-NEUTRALITY (net beta ≈ 0 → max-DD bounded by the reversal spread's own volatility, ~15-25%, not the 50%+ market beta). It captures the short-term reversal spread (losers bounce, winners give back) market-neutrally. It is NOT momentum (opposite sign, distinct factor — so not a duplicate of my pending momentum spreads), NOT the dead high-turnover DAILY cross-sectional reversal (L12: that was ~5600 trades; this is WEEKLY, minimal 4-leg, far lower turnover), NOT funding/options/order-flow. Long-short, market-neutral (fills the most under-represented direction bucket). WEEKLY, USD-M. 2 core parameters (reversal lookback ~1 week, names per side).

Deliberately the OPPOSITE factor from the pending momentum spreads: cross-sectional SHORT-TERM REVERSAL (long recent losers / short recent winners over a ~1-week lookback), a distinct, well-documented anomaly (short-horizon cross-sectional returns mean-revert, the mirror of 3-month momentum) — so it is NOT a duplicate of the momentum set. It stays in the one construction that structurally solves the drawdown floor that killed every long-only momentum reuse this session (crowding 63% / dispersion 67% / mid-cap 74% / equity-breaker 56%): DOLLAR-NEUTRALITY, equal long and short notional -> net beta ~0 -> max-DD bounded by the reversal spread's own volatility (~15-25%), structurally under the L19 floor. Mechanism in _rebalance: rank the 8 by trailing reversal_lookback return, SHORT the top-N winners and LONG the bottom-N losers (kept disjoint), reconcile each instrument to the desired long/short/flat side (flip = close then re-open opposite; same side kept to limit churn). Turnover discipline (L12): the dead cross-sectional reversal was DAILY (~5600 trades); this is WEEKLY and minimal (4 legs, ~8 fills/week) so fee drag stays well below the reversal spread it harvests. VENUE: USD-M futures required (the strategy SHORTS the winners), matching the deployed family. SIZING/LEVERAGE: each leg 0.22 of equity, 4 legs -> gross ~0.88, net ~0, so leverage stays 1.0 and is genuinely unused (no leverage-unused gate; a market-neutral book cannot suffer 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) return reads; rebalance weekly). 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 (reversal_lookback, names_per_side) shared across names — not per-name tuned — to resist overfit.

Decisive net loser and structurally wrong-signed: profit_factor 0.670, total_return -97.02%, avg_trade_return_pct -1.10%, expectancy -$105 over 884 trades, negative in every year (2020 -14.5%, 2021 -79.2%, 2022 -19.8%, 2023 -30.5%, 2024 -28.4%, 2025 -6.3%). max_drawdown is 97.52% (CI to 99.98%), so the hypothesis's core claim that dollar-neutrality bounds drawdown at 15-25% is again catastrophically falsified — identical to the dollar-neutral momentum spreads. The root cause is that cross-sectional short-term reversal is the OPPOSITE sign of momentum, and momentum is the factor that actually works on crypto majors (deployed Sharpe 2.6): longing recent losers (which keep falling) and shorting recent winners (which keep ripping in a momentum-driven market, with unbounded squeeze risk) is a structurally losing bet, which is why it bled every year and blew through the dollar-neutral DD bound. The reported Sharpe 0.809 is a skew artifact of the return distribution and is untrustworthy against a -97% total return and PF 0.67. Optimizing reversal_lookback/names_per_side cannot flip a factor whose edge is negative at this horizon on this universe, and the short-winners leg carries the same unbounded squeeze risk that killed the dollar-neutral momentum siblings. Abandon.

Dollar-neutral short-term REVERSAL spread, long-short market-neutral on 8 BINANCE USD-M majors (BTC/ETH/SOL/BNB/XRP/ADA/AVAX/LINK, 1-DAY bars). WEEKLY it ranks the 8 by their trailing ~1-week return and FADES that short-horizon move: LONGS the 2 biggest recent LOSERS and SHORTS the 2 biggest recent WINNERS, equal notional per leg, so net market exposure ~0 and drawdown is bounded by the reversal spread's own volatility (~15-25%) rather than crypto beta. This is the OPPOSITE factor to cross-sectional momentum (short-horizon returns mean-revert). Only 4 legs rebalanced weekly to stay far below the daily-reversal fee death. Pure-price, no supp data, no ticks. 2 core parameters: reversal_lookback, names_per_side.

Analyst: abandon/iterate on negative expectancy — the short-term reversal edge is not present on the 8 majors here.

The reversal factor is a net LOSER on a measurable window (145 trades, metrics_reliable=true): total_return -14.62%, avg_trade_return_pct -0.248% (negative expectancy), PF 0.878, Sharpe -0.61 (CI [-2.31, 1.06]), win_rate 0.49. NOT an L17 code-defect signature (balanced 72L/73S, |Sharpe|<1) — reads as a genuinely absent/negative short-term reversal edge on majors, not a bug. Turnover high (39.77, impact $774) from the 7-day lookback + weekly flips, but the edge itself is negative so fees are secondary. Unlike the monthly siblings this clears the ~100-trade floor, so it's a decisive analyst signal.

Skip the orders_open guard for the reopen half of a flip (close is reduce-only), or defer the opposite leg to the next rebalance.

Flip close-then-reopen on the same bar; _open_leg's orders_open guard can see the just-submitted reduce-only close and skip the reopen half, briefly tilting off dollar-neutral. Fires MORE often here (7-day lookback → frequent flips) than in the monthly siblings, but sandbox still shows balanced 72L/73S, beta -0.047 — not materially breaking neutrality. Latent risk worth fixing given the higher flip rate.

Genuinely distinct factor (short-term reversal = anti-momentum), lower-turnover weekly design than the dead daily reversal; well-sampled (884 trades)

Decisive net loser: profit_factor 0.670, total_return -97.02%, avg_trade_return_pct -1.10%, expectancy -$105/trade

max_drawdown 97.52% (CI to 99.98%) — near-total wipeout; the dollar-neutral 'DD bounded at 15-25%' thesis is catastrophically falsified

Negative EVERY year (2020 -14.5% ... 2024 -28.4%, 2025 -6.3%) — a structurally wrong-signed factor

Cross-sectional reversal is the OPPOSITE of momentum, and momentum is what works on crypto majors — shorting recent winners in a trending market carries unbounded squeeze losses (same mechanism that blew up the dollar-neutral momentum spreads)

The positive reported Sharpe 0.809 is a skew artifact and untrustworthy against a -97% return / PF 0.67

DollarNeutralShortTermReversalSpreadLS deliberately flipped the session's momentum factor, trading weekly cross-sectional short-term reversal (long recent losers / short recent winners of the majors) in the dollar-neutral construction that was supposed to bound drawdown to ~15-25%. Instead it was a catastrophic, structurally wrong-signed loser: -97.02% total return, 97.52% max drawdown, profit factor 0.670, and negative in every single year across 884 trades. The root cause was that reversal is the opposite of the momentum edge that works on crypto majors, so it faded the trend while the short-winners leg carried the same unbounded squeeze risk that had already blown up the dollar-neutral momentum spreads. The analyst abandoned it, noting its superficially positive Sharpe of 0.809 was a skew artifact untrustworthy against the -97% return, and that tuning could not flip a negatively-signed factor.

Dollar-neutrality does not bound drawdown when the underlying factor is negatively signed on the universe — betting against the direction that works (fading momentum on crypto majors) combined with unbounded short-squeeze risk produces near-total wipeouts regardless of net beta, and a positive Sharpe can be a skew artifact that masks a -97% return.

It was abandoned at backtest-review (verdict: abandon) as structurally wrong-signed — short-term reversal is the opposite of the momentum that actually works on crypto majors, so longing losers that keep falling and shorting winners that keep ripping (with unbounded squeeze risk) bled every year and blew through the dollar-neutral drawdown bound, exactly like the dollar-neutral momentum spreads.

It traded the opposite factor to the session's momentum spreads — cross-sectional short-term reversal — going dollar-neutral and market-neutral by weekly longing the 2 biggest 1-week losers and shorting the 2 biggest 1-week winners of the 8 majors, on the thesis that net-zero beta would bound max drawdown to the reversal spread's ~15-25% volatility.

It was a decisive net loser: over 2,397 days and 884 trades it returned -97.02% with a 97.52% max drawdown (CI to 99.98%), profit factor 0.670, avg_trade_return_pct -1.10%, expectancy -$105/trade, and was negative every single year (e.g. 2021 -79.2%, 2023 -30.5%, 2024 -28.4%); the reported Sharpe of 0.809 was flagged as a skew artifact.