CrossAssetRiskRegimeQqqTltUseq

Cross-Asset Risk-Regime Allocator on Index ETFs — Long QQQ in Risk-On, Flight-to-Quality Long TLT + Short QQQ in Risk-Off, Daily Bars, Trend + Realized-Vol Gate (USEQ, Long+Short, 2-Parameter)

A CROSS-ASSET, LONG+SHORT risk-regime allocator on US index ETFs — a completely different family from the collision list, which is entirely crypto delta-neutral spot-perp FUNDING CARRY (all abandoned fee_edge/other). No funding, no perps, no crypto, no two-leg basis convergence. The mechanism is flight-to-quality regime switching between US equities and long-duration Treasuries, an archetype the factory explicitly flags as 'the most robustly backtestable strategies in the whole factory' because USEQ daily bars carry DECADES of split-adjusted history — exactly the deep sample the 3-phase deflation (PBO, DSR, programme FDR) needs and that the single-name crypto TSMOM/funding clones never had. Regime is read off the S&P 500 proxy: RISK-ON when SPY closes above its long SMA AND SPY trailing realized vol sits below a high threshold; RISK-OFF otherwise. Risk-on: hold LONG QQQ (equity beta). Risk-off: exit QQQ, rotate LONG TLT (Treasuries rally on the classic flight-to-quality bid when equities de-risk) AND put on a modest SHORT QQQ — so the book carries genuine short exposure, directly attacking the 85.3% long-only concentration rather than adding to it. The edge is a persistent negative equity/bond correlation in stress plus equity trend persistence, not a fitted spread or a scheduled cash flow. Kept to TWO tunables (sma_len, vol_lookback) plus fixed threshold multipliers to resist the overfit deaths that killed the multi-parameter crypto clones.

Implements the hypothesis exactly: SPY trend (close vs long SMA) AND a realized-vol gate define the regime; risk-on = long QQQ; risk-off = long TLT + short QQQ. Only TWO tunables are exposed as the model (sma_len, vol_lookback); every threshold is a fixed multiplier (1.5x median vol, 0.10 penalty coefficient, 1% hysteresis) to resist overfitting, and all clamps are declared in _param_bounds so the optimizer never tests values the code rewrites. The vol threshold is expressed as a ratio to the instrument's OWN trailing median vol rather than an absolute constant, so signal frequency is scale- and regime-invariant. calculate_signal returns the raw continuous score every bar (never gated behind the entry condition) and the entry decision is the hysteresis state machine. Legs are managed manually through _submit_entry_instrument/_submit_exit_position because the base class only auto-manages the untraded primary (SPY); should_enter/should_exit therefore return None/False so the base never submits a competing SPY order. Venue is USEQ with leverage 1.0: the book needs shorts (a MARGIN venue) but no amplified notional — gross stays at 0.90x equity in both states, so no leverage is set that the sizing does not consume. Whole-share sizing respects USEQ size precision 0, and holding periods are weeks-to-months so the ~0.02-0.05% USEQ round-trip cost is a small fraction of per-trade edge.

coding_stuck_3_timeouts

Cross-asset risk-regime allocator on US index ETFs (daily USEQ bars). The regime is read off SPY (primary, signal-only, never traded): a continuous risk score = SPY's percentage deviation from its sma_len SMA minus a penalty for elevated realized volatility (stdev of the last vol_lookback daily log returns divided by the median of its own trailing vol history, penalised only above a fixed 1.5x multiple). RISK-ON (score above +1% hysteresis) holds a LONG QQQ sleeve at 90% of equity. RISK-OFF (score below -1%) rotates to LONG TLT (65% weight) plus a modest SHORT QQQ (35% weight) — a genuine short book that expresses the flight-to-quality bid in Treasuries while equities de-risk. Inside the hysteresis band the previous state is held, so flips are a handful per year. Whole-share sizing, gross notional capped at 90% of equity in both states, entries deferred one session after the exit that frees the capital.

CrossAssetRiskRegimeQqqTltUseq-14ee087930

This candidate proposed a genuinely different family from the abandoned crypto funding-carry clones — a flight-to-quality regime switcher on US index ETFs... never got that far, stalling through three consecutive coding timeouts over two iterations, so the edge remains entirely untested.

A strategy this structurally involved risks exceeding the coder's time budget; a simpler first cut may be needed to get a testable version out of coding.

Abandoned at the coding stage after the developer agent stalled through three consecutive timeouts (coding_stuck_3_timeouts) across 2 iterations; verification, backtesting, and analysis were never reached.

A cross-asset, long+short risk-regime allocator on US index ETFs that reads a trend + realized-vol regime off SPY to hold long QQQ when risk-on and rotate into long TLT plus a short QQQ leg when risk-off, harvesting the flight-to-quality equity/bond correlation with only two tunables.

No performance data exists: the strategy never produced a backtest, optimization, or any stage verdict because it did not clear the coding stage.

sandbox_timeout: Verification failed (Layer 3 — sandbox backtest): smoke test exceeded the 300s wall-clock limit on a ~20000-bar window (2400 days of SPY.USEQ-1-DAY-LAST-EXTERNAL). This almost always means per-bar work that scales with history — e.g. rescanning the full funding/supplementary series, or rebuilding a list and calling min()/sorted() inside calculate_signal()/on_bar() on every bar. Precompute sorted timestamp arrays ONCE in __init__ and use bisect, or cache lookups keyed by timestamp, so per-bar cost is O(log n) not O(n).