{"as_of":"2026-06-03T10:47:03Z","honesty_disclosure":null,"kill_policy":null,"re_cast_date":null,"variants":[{"slug":"baseline","display_name":"Baseline","hypothesis":"Status quo works — current shadow config","kill_condition":null,"structural_diffs_from_baseline":null,"honesty_status":"stable","config":{"confidence":0.74,"size":50.0,"paper_balance":1000,"take_profit":0.2,"stop_loss":0.1,"max_session_loss":100.0,"min_price":0.05,"analyzer_model":"claude-sonnet-4-6","review_model":"claude-haiku-4-5"},"stats":{"n_closed":44,"wins":12,"losses":29,"breakevens":3,"hit_rate":0.2927,"hit_rate_ci_lo":0.1761,"hit_rate_ci_hi":0.4448,"total_pnl_usd":1.33,"total_pnl_ci_lo":-271.25,"total_pnl_ci_hi":334.87,"open_positions":5,"latest_trade_at":"2026-05-24 02:40:51","paper_balance":654.82,"session_pnl":-53.86},"kill_progress":{"status":"n/a","progress_pct":null,"note":"Reference variant. Replaced when any other variant beats it on both P&L AND Brier at n>=50."}},{"slug":"highconv","display_name":"High Conviction","hypothesis":"Tighter threshold + larger size — fewer trades, higher expectancy. Tests calibration finding that 0.76–0.80 bucket showed 75% WR / +42% ROI.","kill_condition":null,"structural_diffs_from_baseline":null,"honesty_status":"stable","config":{"confidence":0.78,"size":75.0,"paper_balance":1000,"take_profit":0.2,"stop_loss":0.1,"max_session_loss":100.0,"min_price":0.05,"analyzer_model":"claude-sonnet-4-6","review_model":"claude-haiku-4-5"},"stats":{"n_closed":3,"wins":1,"losses":2,"breakevens":0,"hit_rate":0.3333,"hit_rate_ci_lo":0.0615,"hit_rate_ci_hi":0.7923,"total_pnl_usd":-12.97,"total_pnl_ci_lo":null,"total_pnl_ci_hi":null,"open_positions":1,"latest_trade_at":"2026-05-09 01:17:00","paper_balance":837.025,"session_pnl":4.59},"kill_progress":{"status":"pending","progress_pct":10.0,"note":"n=3/30 needed to evaluate. Hypothesis: tighter threshold + larger size beats Baseline on expectancy."}},{"slug":"widenet","display_name":"Wide Net","hypothesis":"Looser threshold + smaller size — more trades, sample size grows fast. Tests whether the 0.70–0.74 bucket can be tuned via better exit discipline.","kill_condition":null,"structural_diffs_from_baseline":null,"honesty_status":"stable","config":{"confidence":0.7,"size":30.0,"paper_balance":1000,"take_profit":0.2,"stop_loss":0.1,"max_session_loss":100.0,"min_price":0.05,"analyzer_model":"claude-sonnet-4-6","review_model":"claude-haiku-4-5"},"stats":{"n_closed":7,"wins":2,"losses":5,"breakevens":0,"hit_rate":0.2857,"hit_rate_ci_lo":0.0822,"hit_rate_ci_hi":0.6411,"total_pnl_usd":-1.3,"total_pnl_ci_lo":-56.59,"total_pnl_ci_hi":62.66,"open_positions":8,"latest_trade_at":"2026-05-23 19:18:02","paper_balance":608.793814,"session_pnl":0.0438},"kill_progress":{"status":"pending","progress_pct":7.0,"note":"n=7/100 needed. Hypothesis: extra trade volume produces dataset-value-per-dollar > Baseline."}},{"slug":"trenchv2","display_name":"Trench V2","hypothesis":"Best in-corpus config under joint (close-rate, OOS ROI, bootstrap CI) optimization. Identified 2026-05-11 by a 3,600-cell walk-forward sweep + bootstrap-rank. Wider symmetric brackets (0.30/0.30) than baseline keep positions out of noise-stop range; loose entry (conf=0.70, edge=0.03) diversifies the tape. Bootstrap on full 22-day corpus: ROI +1.15%, WR 64%, P(roi>0)=57.3% vs baseline 0.2% (140× lift). 49% close rate keeps the result honest vs the wider-bracket configs that bias high by leaving positions open.","kill_condition":null,"structural_diffs_from_baseline":null,"honesty_status":"stable","config":{"confidence":0.7,"size":30.0,"paper_balance":1000,"take_profit":0.3,"stop_loss":0.3,"max_session_loss":100.0,"min_price":0.05,"min_edge":0.03,"analyzer_model":"claude-sonnet-4-6","review_model":"claude-haiku-4-5"},"stats":{"n_closed":1,"wins":0,"losses":1,"breakevens":0,"hit_rate":0.0,"hit_rate_ci_lo":0.0,"hit_rate_ci_hi":0.7935,"total_pnl_usd":-8.17,"total_pnl_ci_lo":null,"total_pnl_ci_hi":null,"open_positions":9,"latest_trade_at":"2026-05-22 05:06:52","paper_balance":601.33,"session_pnl":-8.17},"kill_progress":{"status":"pending","progress_pct":3.0,"note":"n=1/30 OOS needed. Hypothesis: in-corpus bootstrap (+1.15% ROI) transfers to OOS."}}],"roadmap":[]}