← Back to Updates · Comparison page →

Money-Maker Plan — Claude Code (Opus 4.7)

RESEARCH SURFACE — NOT FINANCIAL ADVICE
Generated 2026-05-11 · Plan mode output · Skill: .claude/skills/money-maker-ready/SKILL.md v1.0 · Plan file: C:\Users\zerou\.claude\plans\follow-that-skill-and-joyful-lemon.md
⚠ Self-correction posted in comparison page: the §1 "all classes n=0" finding below was an Explore-agent read-error, not a data bug. Cursor + the peer 2026-05-11 combined plan both read the same payload and got real n values (COMMODITY n=408, EQUITY n=443, CRYPTO n=7875, etc). Treat §1's "STRUCTURAL BUG" and §2's "MODULE BROKEN" with skepticism and verify both against the canonical reads before acting on the P0 actions tied to them.

Headline verdict

NOT READY. Three data-layer CRITICALs allegedly block any real-money sign-off:

  1. Dashboard 40.7h stale (vs 2h SLA)
  2. All asset_class_health.n=0 (now known to be read-error; see correction above)
  3. Walk-forward module emits 0 folds across every class (likely also read-error)

Holding survivors of the cross-check: drift alert TRUE but KS_D uncomputed; baby_strats 12-flag overfit; system draggers (kimi_signal_tracking -954%, mercury2_fast, ml_bg_system_b); claude_gainer_st winner-vs-blacklist contradiction (in BLACKLISTED_STRATEGIES at alpha_engine/config.py:216 but tops systems with PF 6.12).

System winners (Tier-2-MDD verified) — 6 pass gate, all unverified

RankSystemPFWR %nMDD %Flag
1multi_asset_cot19.1986.913017.8implausible PF — verify or quarantine
2aggregated_picks6.4277.344114.4aggregator artifact suspect
3claude_gainer_st6.1278.5347214.0in BLACKLISTED — contradiction
4ml_crypto_pred_v122.5355.612311.0likely DEAD (no last_signal)
5signal_validation2.2255.050214.1validator-as-pick-source meta-eval risk
6copy_trader_intel1.8450.07392.23MDD 2.23% looks synthetic

All 6 lack last_signal_date. Treat as DEAD until proven live.

System draggers

SystemPnL %PF
kimi_signal_tracking-954.90.26
multi_asset-146.00.31
mercury2_fast-139.50.07
alpha_engine_fast-127.60.62
paper_trading-103.50.35
copy_trader_highscore-81.80.76
ml_bg_system_b-54.70.02
crypto_winners-52.40.36

Top P0-P2 actions (post-correction)

PriorityActionEffortRisk
P0Verify multi_asset_cot PF=19.19 via DB query against ejaguiar1_stocks1hLow
P0Reconcile claude_gainer_st winner-vs-blacklist contradiction; confirm enforcement at exec gate (memory feedback_gate_at_execution_not_generation)1-2hLow
P0Refresh stale dashboard_data.json; diagnose refresh-cron failure via check-gh-actions0.5hLow
P1Execute existing baby_strats 12-flag quarantine proposal1hLow
P1Fix drift detector — KS_D uncomputed-zero + hf_stats 19d stale3-4hMed
P1Investigation-gate top draggers per MUTATION_THREE_AXIS_PROTOCOL.md4-6hMed
P1Add last_signal_date to systems payload1-2hLow
P2Audit HC filter against confidence-inversion on ETF/CRYPTO (memory project_performance_reality)2-3hLow
P2Re-run cross-permutations; verdict on per-class edge existence1hLow

External-data wires considered

Ranked by impact/effort: Riskfolio-Lib (CVaR/HRP), QuantStats (reporting), pandas-ta (130+ indicators), FRED (macro filter — fred_data_fetcher.py exists, needs key), VectorBT (50-100× backtests), Kalshi (Polymarket pairwise), PyPortfolioOpt, Glassnode/Coinglass widen, Freqtrade, FinRL.

All subject to CLAUDE.md Wire-Up Rule — production caller required or explicit opt-in sidecar with Wiring Plan section.

Full plan file

Plan file persisted at C:\Users\zerou\.claude\plans\follow-that-skill-and-joyful-lemon.md (developer workstation, not committed). Contains:

  1. Freshness preflight (FAIL — 40.7h stale)
  2. Per-class baseline (now corrected — see top of page)
  3. Walk-forward verification
  4. System winners + draggers
  5. Backtest-overfit detector
  6. Drift state
  7. UI / filter audit (HC filter, hyrotrader)
  8. External data integrations
  9. Top statistical edges per asset class
  10. P0-P5 ranked recommendations
  11. Verifiable claims log