{"abstain":false,"candidates":[{"label":"robust","params":{},"researchEvidence":"fee 10bps, dataset ohlcv-sha256-9f8ce49cc04885b6 full 0-6569: SMA200 +171.68% DD45.48% (95 trades) vs BH +136.43% DD77.04% ratio 3.77 vs 1.77. Train 0-4598: +137.37% DD45.48% vs BH +86.38% DD73.62%. Holdout 4599-6569 flat-start: +23.66% DD19.57% (33 trades) vs BH +59.61% DD21.19% (sweep holdout BH +26.94%), positive. P1 bull 0-2189: +121.04% vs BH +284.28%. P2 bear 2190-4379: -20.40% DD28.17% vs BH -65.18% DD73.62% (loses 1/3). P3 recovery 4380-6569: +19.58% DD19.57% vs BH +38.62% DD26.68%. Plateau full: SMA100 +87.77%, SMA150 +156.29%, SMA200 +171.68%, SMA250 +305.66%, SMA300 +226% all >0, 150-300 beat BH absolute and risk-adjusted. Sweep 50 variants exits SL/TP/TO with holdout 30%: train winner +69.87% train / +20.73% holdout, exits hurt vs let-run 137% train, so no exits chosen. Turtle20 -16.01% rejected.","script":"strategy \"Cash under SMA200\" version 1.0\nasset: BTC\ntimeframe: 4h\ncapital: 10000\nentry:\n    condition: price > SMA(200)\n    size: 100%\nexit:\n    # let winners run — no take-profit, no timeout\n"}],"model":"muse-spark-1.3-contributor-free","primaryIndex":0,"reasoning":"Chose the iconic 1-condition trend filter price>SMA(200) let-run, 100% size, no SL/TP/timeout to avoid exit overfit. Not the in-sample best (SMA250 +305% beats it) to avoid picking the peak; 200 is central in a broad 100-300 all-positive plateau where 150-300 beat buy-hold. Validated with fee 10bps (eval fee), 70/30 train-holdout split via backtests plus one run_sandbox_sweep holdoutPercent=30, three sub-periods covering bull/bear/recovery, and neighbor lengths. Loses less in bear (-20% vs BH -65%), wins full on absolute and return/DD, positive in train, holdout and 2/3 regimes with lower DD throughout. Simple, low-turnover, lets winners run, suited to survive unseen 2023-2026 bull/bear.","runId":"muse-1","toolCallsUsed":34}