SpotMajorsMomentumRotationLong

Crypto-Majors SPOT Cross-Sectional Momentum Rotation, Long-Only & UNLEVERED (BINANCE_SPOT, 6 Deep Majors, Weekly Hold Top-3 by Relative Strength — the DEPLOYED USD-M Momentum Winner Ported to the CASH SPOT Venue: No Leverage → No Liquidation, Fills the Under-Represented Spot Venue, Same Proven Relative-Strength Signal, 2-Parameter)

A REAL, codeable strategy that ports the single strongest LIVE, non-decayed, non-options deployed winner — cross-sectional relative-strength momentum rotation (USD-M majors version, Sharpe 2.6) — to the BINANCE_SPOT cash venue, which is genuinely different in risk profile and is under-represented in the book (11.5% vs the venue-quota target). The signal is the proven one (no L30 new-indicator risk — it reuses a deployed winner, not a re-labeled sort); the genuine differentiators are: (1) VENUE — BINANCE_SPOT CASH account, so the strategy is structurally UNLEVERED and CANNOT be liquidated (the −100% blowups that killed leveraged siblings are impossible), which the prompt notes is the highest-survival venue; (2) it fills the spot venue-diversity gap and reduces the book's heavy USD-M concentration. Mechanism: weekly, rank 6 deep-history spot majors (BTCUSDT, ETHUSDT, SOLUSDT, BNBUSDT, XRPUSDT, ADAUSDT) by trailing relative strength and hold the top-3 equal-weight, fully unlevered. It is distinct from my three pending momentum reuses (crowding-filter=funding, dispersion-gate=dispersion, mid-cap=universe) — this one changes the VENUE/leverage profile, not the signal or filter. NOT funding/order-flow/options/reversion (all dead), NOT a new indicator. Long-only, diversified, weekly, spot. 2 core parameters (momentum lookback, number held).

Ports the deployed cross-sectional relative-strength momentum rotation (USD-M majors, Sharpe 2.6) to the BINANCE_SPOT cash venue, keeping the PROVEN signal unchanged (no L30 new-indicator risk). The genuine differentiators vs the deployed version and my three pending momentum reuses (funding crowding-filter, dispersion gate, mid-cap universe) are VENUE and leverage profile: BINANCE_SPOT is a CASH account, so the strategy is structurally unlevered and CANNOT be liquidated — the -100% liquidation blowups that killed leveraged siblings are impossible here — and it fills the under-represented spot venue, reducing the book's heavy USD-M concentration. Mechanism in _rebalance: rank the 6 majors by momentum, hold equal-weight top-3, but only fill a slot if that name's own momentum is positive (absolute-momentum cash gate), so the book steps aside to cash rather than holding falling knives. SPOT COMPLIANCE: the venue is a CASH account — the strategy is long-only (only BUY entries; exits use close_position on existing longs, never a short), which the base also enforces via its spot-short guard; leverage is pinned to 1.0 (spot hard-caps at 1x). Purely price-driven (no supplementary or tick data) so it avoids cross-instrument-supp and tick-timeout pitfalls and is inherently timeout-safe (bounded per-bar work: 6 O(1) momentum reads; rebalance only weekly). All trading across the 6 legs (including the primary) happens in the rebalance routine 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 the primary. All 6 majors are deep-history, highly-liquid BINANCE_SPOT symbols, so every leg has data (no data_unavailable). Risk is bounded by construction: cash-funded, long-only, <=3 names x 0.30 = 0.90 of equity deployed, no leverage, no compounding-liquidation path. 2 core parameters (momentum_lookback, top_k) shared across names — not per-name tuned — to resist overfit. Note on fees: spot taker fees (0.10%) are higher than USD-M, but weekly low-turnover rotation keeps fee drag modest while gaining the unliquidatable cash-venue survival profile the prompt flags as highest-survival.

failed holdout and deflated Sharpe: the holdout collapsed to Sharpe -1.98 (ratio -0.891, well below the 0.7 bar) — the in-sample-strong config actually LOST money out-of-sample — while DSR=0.0 over 225 trials (expected-max 9.44) means the selected Sharpe is indistinguishable from best-of-N noise. This is a decisive, unwaivable triple hard failure: holdout failed, walk-forward is_overfitted=true (IS avg 12.72 vs OOS avg 2.23, with degenerate 11-15 trial Sharpes), and PBO 0.538 > 0.5, so promote is impossible. The optimizer also collapsed the strategy to just 21 trades concentrated in 2020-2021 (optimized annual returns show only 2020 +24% and 2021 +400%) and the sensitivity analysis found TWO cliff parameters (top_k, per_name_frac jump from ~3.5 to ~9.2 Sharpe on small moves) — a textbook overfit surface. I flagged this exact risk at backtest review ('if the OOS Sharpe in 2025-2026 collapses, abandon rather than promote on a 2021-concentrated headline'), and it collapsed. This is the overfit signature (high IS, negative holdout, DSR 0.0, cliff params), which per policy is abandon not iterate — the edge is genuinely concentrated in the 2021 bull and does not generalize, so a second optimization pass would only re-select the same overfit. The calibration is unambiguous: every near-sibling momentum variant this session (mid-cap, spot dual-momentum, absolute-momentum timer, dispersion-gate, capitulation basket) has already been abandoned overfit. Abandon.

Long-only, structurally UNLEVERED weekly cross-sectional momentum rotation over 6 deep-history BINANCE_SPOT majors (BTC/ETH/SOL/BNB/XRP/ADA, 1-DAY bars). Weekly it ranks the names by trailing relative strength (return over momentum_lookback days) and holds an equal-weight top-3, with an absolute-momentum cash gate (only positive-momentum names are held; the rest stay cash) to de-risk out of falling majors in bear phases. Runs on the BINANCE_SPOT CASH account, so it is long-only and cannot be liquidated. Pure-price, no supp data, no ticks. 2 core parameters: momentum_lookback, top_k.

Analyst: abandon candidate on negative expectancy. Do not iterate the code — the implementation is correct.

PERFORMANCE FLAG FOR BACKTEST_REVIEW (not a QA code defect). Sandbox negative: total_return -18.6%, max_drawdown 27.1%, profit_factor 0.24, win_rate 0.36, Sharpe -0.84, avg_trade_return_pct -5.25%. Large-move directional losses, not a fee/measurability artifact — the same crypto-majors momentum edge failing this window that the two sibling rotations showed. The spot port removes liquidation risk but not directional-edge risk. Analyst abandon call, not a code fix.

If alignment isn't guaranteed by the base, gate the snapshot on all legs sharing ts_event before computing momenta.

Leg-alignment assumption: calculate_signal snapshots each leg's latest close assuming an alignment barrier buffered all 6 same-tf legs for that timestamp. If the base doesn't strictly guarantee same-ts_event alignment, a non-primary leg's _latest_bar could be one bar stale, adding a consistent 1-bar momentum lag. Would not flip polarity or break trading.

First momentum-reuse this session to keep max_drawdown UNDER 50% (38.1%) — unlevered spot + absolute-momentum cash gate genuinely capped the tail (2022 only -30.9%)

Functions correctly on BINANCE_SPOT (full 2020-2026 daily return series, 192 trades) — avoids the degenerate/freeze infrastructure failure of the buy-and-hold spot baskets, because the cash gate realizes PnL frequently

Genuine edge: profit_factor 1.578, avg_trade_return_pct 13.8%, Sharpe 0.769 (CI low 0.20 positive); reuses the deployed proven momentum signal; fills the under-represented spot venue

Single-regime concentration: 2021 +553% (and 2024 +88%) carry the +993% headline; 2026 -26.9%, rolling Sharpe decayed from 9+ to ~-2.8 — high deflated-Sharpe/PBO risk

Weaker duplicate of the deployed winner: same signal at Sharpe 0.77 vs the deployed USD-M version's 2.6; information_ratio -0.56 vs its benchmark (does not beat plain equal-weight momentum)

Sharpe 0.769 below the 1.5 floor

Unlevered spot with an absolute-momentum cash gate kept the un-optimized drawdown under 50% (38%); functions correctly on BINANCE_SPOT

Reuses the deployed, economically-real momentum signal

HARD FAIL: holdout failed catastrophically — holdout_sharpe -1.98, ratio -0.891 (the in-sample-strong config LOST money out-of-sample)

HARD FAIL: walk-forward is_overfitted=true (IS 12.72 vs OOS 2.23); trial Sharpes are degenerate 11-15 artifacts

HARD FAIL: PBO 0.538 > 0.5

deflated_sharpe 0.0, is_significant=false; TWO cliff parameters (top_k, per_name_frac) — overfit surface

Optimizer collapsed to 21 trades concentrated in 2020-2021 (optimized returns only 2020 +24%, 2021 +400%), which is why the later holdout cratered

Every near-sibling momentum variant already abandoned overfit (mid-cap, spot dual-momentum, absolute-momentum timer, dispersion-gate, capitulation)

SpotMajorsMomentumRotationLong was the session's most promising momentum reuse: by porting the deployed signal to the unlevered BINANCE_SPOT venue with an absolute-momentum cash gate, it became the first variant to keep max drawdown under the 50% floor (38.1%) while showing a genuine edge (PF 1.578, avg trade 13.8%), earning an 'optimize' verdict at backtest-review. Optimization then exposed the flaw the reviewer had explicitly flagged: the in-sample-strong config lost money out-of-sample (holdout Sharpe -1.98), walk-forward was overfit (IS 12.72 vs OOS 2.23), PBO 0.538 exceeded 0.5, and deflated Sharpe was 0.0 across 225 trials, with the optimizer collapsing to just 21 trades concentrated in 2020-2021 and two cliff parameters. The analyst abandoned it as a decisive, unwaivable triple hard failure — the edge is genuinely concentrated in the 2021 bull and does not generalize — consistent with every near-sibling momentum variant already abandoned overfit this session.

Unlevered spot plus an absolute-momentum cash gate genuinely solved the drawdown-floor problem (38% DD), but a single-regime edge concentrated in the 2021 bull does not survive holdout — a strong in-sample Sharpe with negative out-of-sample performance, DSR 0.0, and cliff parameters is the overfit signature that warrants abandon, not another optimization pass.

It passed backtest-review (verdict: optimize) but was abandoned at the post-optimization analysis stage on a decisive triple hard failure: the holdout collapsed to Sharpe -1.98 (ratio -0.891, losing money out-of-sample), walk-forward flagged is_overfitted=true (IS 12.72 vs OOS 2.23), and PBO 0.538 > 0.5, with deflated Sharpe 0.0 and two cliff parameters — the optimizer had collapsed the book to 21 trades concentrated in 2020-2021.

It ported the deployed USD-M cross-sectional relative-strength momentum rotation (Sharpe 2.6) to the BINANCE_SPOT cash venue — weekly hold equal-weight top-3 of 6 spot majors by trailing relative strength, fully unlevered with an absolute-momentum cash gate — betting that a cash account (no leverage, no liquidation) would fill the under-represented spot venue and cap the tail that killed leveraged siblings.

On the initial backtest over 3,263 days it made 192 trades with +992.7% return, profit factor 1.578, avg_trade_return_pct 13.8%, Sharpe 0.769, and — notably — a max drawdown of 38.1%, the first momentum reuse this session to stay under the 50% floor; but the headline was single-regime (2021 +553%) and its information ratio was -0.56 versus the equal-weight benchmark.