SolBtcMarketFactorBetaAmplifiedTrendLS

Beta-Amplified Market-Trend Directional Capture, Long-Short (Trade the HIGH-BETA alt SOLUSDT.BINANCE in the direction of BTC's CLEAN confirmed market trend — majors-quality trend signal, alt-magnitude payoff; BTC used ONLY as the market-factor signal, SOL is the sole tradeable leg, 3-parameter)

A LONG-SHORT, single-TRADEABLE-instrument directional strategy that separates the SIGNAL asset from the PAYOFF asset. The factory has empirically found exactly one durable edge: 4H directional trend on the majors (ETH/BTC/ADA dual-TF confluence survivors). The problem with re-skinning that edge onto other symbols (L91) is that alts carry far more idiosyncratic noise than BTC, so a trend signal computed ON the alt is dirtier and overfits. This hypothesis fixes that by DECOUPLING signal from payoff: derive the trend state from BTC (the crypto market factor — the single cleanest, most persistent trend series in the asset class, and the one the survivors already prove is tradeable), then EXPRESS that view in a high-beta liquid alt (SOL, beta ~1.5-2.0 to BTC) which mechanically amplifies the same directional move. You get a majors-quality trend signal with alt-magnitude payoff. This is NOT same-instrument momentum-confluence (L91 killed those), NOT a cointegration/relative-strength pairs trade (those keep dying negative-expectancy — this is directional net-long or net-short, never market-neutral), NOT a regime/vol gate bolted onto momentum (L92 — the trend definition itself lives on a different, cleaner instrument), and NOT funding/liquidation/basis/premium (all dead classes). BTC is signal-only; the ONLY order ever placed is on SOL.

Implements the hypothesis literally: signal asset (BTC, the crypto market factor) is decoupled from payoff asset (SOL, beta ~1.5-2.0). BTC 4H bars arrive via extra_instruments/extra_bar_types and are consumed in on_extra_bar(); the only orders ever placed are on the primary SOLUSDT.BINANCE leg, so the book is net long or net short and never market-neutral (avoids the dead pairs/cointegration class). It is not same-instrument dual-timeframe confluence (SOL price never enters the direction decision) and uses no funding/liquidation/basis data. Entry is a single threshold on one continuous variable rather than a compound gate, so it fires often enough for the sandbox window; both legs are BINANCE 4H, so the base class's same-timeframe alignment barrier guarantees the BTC bar is contemporaneous with the SOL bar being traded. Futures venue is required because the strategy goes short; leverage stays 1.0 because the amplification is meant to come from SOL's beta, not from margin, and an unused leverage setting would be inert. One caveat, stated as required: BTC's trend and SOL's realised move share the same sign only through beta, so any BTC-idiosyncratic trend (ETF/regulatory flows) is expressed in SOL with tracking error the code cannot hedge; I implemented the hypothesis as specified regardless.

analyzing_stuck_3_timeouts

Long/short directional strategy on SOLUSDT 4H perp whose DIRECTION comes entirely from BTC, not from SOL. Every BTC 4H bar updates an ATR-normalised dual-EMA trend read, signal = (EMA_fast - EMA_slow)/ATR on BTC. SOL is bought when that read exceeds +entry_thresh and sold short when it falls below -entry_thresh; SOL's own price is used only for risk (ATR-based sizing and an ATR trailing stop), never for direction. Exits on a BTC trend flip against the position or when SOL gives back trail_atr_mult x the entry SOL-ATR from the best close since entry. Sizing is risk-first (equity * risk_pct / stop distance) and hard-capped at max_notional_frac * equity * leverage. Three tunables: ema_ratio, entry_thresh, trail_atr_mult.

Optionally rename the warm-up helper to avoid the shadow warning, but not required.

min_bars_required() and _submit_exit() shadow framework methods (static-analysis base_shadow warnings). Both are deliberate here: _submit_exit calls super() then resets trailing state, and min_bars_required is also invoked directly by the strategy's own should_enter() warm-up gate, so correctness does not depend on the base calling the override. No functional defect.

No action needed at QA; informational.

avg_holding_bars is reported as 0.0 while avg_holding_period is '4d 8h 46m' (~26 4H bars). This is a reporting/aggregation artifact, not a code bug — trades clearly hold multiple bars (426 trades over 2183 days with a trend-flip/trailing-stop exit).

Faithful implementation and sandbox is positive, but the durability claim rests on BTC's trend expressing cleanly in SOL via beta. Regime table shows the edge concentrates in calm/normal (Sharpe 1.79/1.18) and nearly vanishes in the stressed tercile (Sharpe 0.18, DD 34.6%), and LUNA/SVB stress windows are decisively negative (-2.45 / -3.67 Sharpe) — the beta linkage decouples exactly when idiosyncratic shocks hit. Worth watching whether full-history + OOS holds up outside the calm regime; the L91/L92 graveyard of alt trend-ports argues for scrutiny at backtest review.

avg_trade_return_pct 1.08% is ~7x the 0.15% Binance USD-M fee floor — genuine per-trade edge net of costs (commission 2.53% of gross, impact 4.25%, capacity $55M).

Reliable 426-trade / 6.7-year sample, balanced 223 long / 203 short; profit_factor 1.35, Sharpe 0.94 with CI [0.124, 1.69] (low bound positive), max_drawdown 19.6% < 20%, recovery_factor 13.7.

Faithful mechanism: 426 signaled = 426 submitted, BTC is signal-only (on_extra_bar), SOL is the sole traded leg, directional net-long/short — trades implement the hypothesis, not a coding accident.

Decoupled signal/payoff design directly addresses the L91 same-instrument-noise critique rather than re-skinning it.

Edge is regime-concentrated: Sharpe calm 1.79 -> normal 1.18 -> stressed 0.18 (DD 34.6% in stress); LUNA -2.45 and SVB -3.67 windows negative — QA's beta-decoupling concern is visible on full history.

Win rate 38% with PF carried by right-tail winners (avg_win 5577 vs avg_loss 2554); the L91/L93 alt-trend-port family repeatedly dies overfit in the 3-phase deflation, so the base Sharpe may be a ceiling.

Sharpe 0.94 / PF 1.35 sit below the survivor tier this class has historically needed.

SolBtcMarketFactorBetaAmplifiedTrend-d69e429802

SolBtcMarketFactorBetaAmplifiedTrendLS tested a clean idea: read BTC's 4H trend as the market factor and trade only SOL, a higher-beta alt, in that direction to get majors-quality signal with alt-magnitude payoff. The single-iteration run backtested respectably (258% total return, Sharpe 0.94, 426 trades, avg per-trade return 1.08%) and the analyst cleared it to optimize, after which tuned parameters improved it to Sharpe 1.07 and CAGR 28.5% while passing sensitivity and holdout. However, robustness signals were mixed (PBO 0.67, deflated Sharpe 0.81, not programme-FDR significant, and a stressed-regime Sharpe of just 0.18), and the strategy never received a final analyst verdict. It was ultimately abandoned in the ANALYZING stage after the analyst agent timed out three consecutive times, ending the run on an operational failure rather than a clear reject-on-merits decision.

A strategy can survive the quantitative gates yet still terminate on operational failure (repeated analyst timeouts), so slow/edge-marginal candidates need a resilient analysis path; separately, the edge here was regime-concentrated (stressed-regime Sharpe ~0.18) and failed programme-level FDR, so future iterations should target robustness beyond calm regimes.

It cleared backtest-review (verdict: optimize) and completed optimization, but was abandoned in the ANALYZING stage because the analyst agent timed out three times (reason: analyzing_stuck_3_timeouts) and never produced a final verdict; note the run's robustness was also weak (PBO 0.67, not programme-FDR significant).

Decouple signal from payoff by deriving a clean 4H trend state from BTC (the crypto market factor) and expressing it directionally, long or short, in the higher-beta alt SOL, which was the only instrument ever traded.

The initial backtest over ~6.7 years returned 258% total (CAGR 22.8%, Sharpe 0.94, PF 1.35) across 426 balanced long/short trades with avg per-trade return 1.08% (~7x the fee floor) and 19.6% max drawdown; optimization lifted it to Sharpe 1.07, CAGR 28.5%, PF 1.43 and avg per-trade return 1.73%, and it passed sensitivity and the holdout gate.