Real Money — Tools, Gates, and Readiness

⚠ NFA — Not Financial Advice. No live capital sizing without explicit user greenlight AND all readiness gates clear AND DSR≥0.95 on the candidate strategy/symbol. Paper-pilot graduation gate (cot_paper_pilot_status.json) is the canonical SHADOW → LIVE_ELIGIBLE bar.

10-step readiness gate (Lopez de Prado AFML + Codex governance)

Pre-register — entry signal + exit rule + sizing committed in writing before any data look
In-sample backtest — PF≥1.5, WR≥55, MDD≤10 on the candidate
Walk-forward analysis — OOS WR within ±10pp of in-sample, WFE>60%
CPCV (combinatorial purged cross-validation) — mean OOS WR≥75% across 10 folds, worst fold≥60%
DSR (deflated Sharpe ratio) — tools/anti_overfit_audit_sidecar.py → DSR≥0.95 at conservative n_trials
Break-tests — pre/post-2020 regimes, low-vol vs high-vol regimes, single asset removal
Sensitivity sweep — parameter robustness within ±20% of optimal
Transaction-cost overlay — net per-trade economics survive realistic slippage
3-6 month paper-pilot — paper_pilot.html tracker, WR within ±10pp of backtest
Graduated sizing — 5% → 25% → 100% over 3+ months, halted on any decay alert

Canonical reference: SUPREME EDGE master plan + reports/db_evidence_graded_final_2026-05-08.md. Codex state machine: REHAB → OOS_READY → SHADOW → LIVE_ELIGIBLE. Single-class deviation (cot_positioning + CT=F) accepted by user 2026-05-12.

Tools shipped this session (2026-05-12)

COT Paper Pilot Tracker SHADOW

SHADOW-state ledger for the DSR=1.0000 verified cot_positioning + CT=F edge (n=100, WR 90%, Sharpe +1.377). 1-contract paper sizing, 4-week graduation gate, per-trade P&L vs expected $3.40-$13.40 band.

View paper_pilot.html → · alpha_engine/strategies/cot_paper_pilot.py · cron: hourly via audit-dashboard.yml

Anti-Overfit Audit (DSR) RESEARCH

Deflated Sharpe Ratio per strategy via Lopez de Prado AFML eq 14.5. Surfaces strategies passing DSR≥0.95 ("publishable confidence" per Bailey-Lopez de Prado) and overfit-likely rejections. Hourly cron.

View anti_overfit.html → · tools/anti_overfit_audit_sidecar.py

Research Sidecars (Agent C) RESEARCH

4 academically-backed sidecars: commodity carry+momo (Miffre 2010), sector dual-momentum (Antonacci GEM), UST term-structure momentum, overnight-intraday reversal. Hourly cron.

View research_sidecars.html → · tools/research/*.py

Top-N Rank Backtest NEW

Hindsight replay: "if I bought the top-10 EQUITY picks ranked by score T days ago, would I have been profitable?" Aggregates today / yesterday / day_before / 7d / 30d / 90d windows. Per-day detail + cumulative P&L.

tools/top_n_rank_backtest.py · audit_dashboard/data/top_n_rank_backtest.json · cron: hourly

DB Health Panel REFRESHED

6 truth-layer metrics: PnL Integrity (sampled), Ghost Rows (constant pnl_pct cohorts), Forward Validator Freshness, Phantom EXPIRED rows, Raw-Pick Outcome Coverage, WON-vs-PnL contradiction. Per-metric "Action required" remediation banner now reflects shipped fixes.

tools/db_health_check.py · View on /audit → · remediation status →

MONEY READY filter SHIPPED

Strictest preset on the main /audit picks grid: only DSR-verified Tier-1 sleeves visible. Confidence ≥70% (normalized), score ≥ per-class floor, no decay-alert, asset_class not blocked. Each surviving pick gets a justification tooltip.

audit_dashboard/money_ready_filter.js · Toggle via btn-money-ready on /audit

Ghost-Row Symbol-Triple Block SHIPPED

New BLOCKED_ASSET_STRATEGY_SYMBOL_TRIPLES axis in quality_gates.py with 5 documented cohorts (quan_engine MATIC, KIMI_signal_tracker ETH/BTC, irb_hoffman ADA, funding_rate_carry ROBO). Enforced at passes_active_gate AND _is_historical_blocked_pick.

commit 597819d79c7 · expected ~220k+ rows excluded from CRYPTO aggregates

WON-vs-PnL Sign-Coherence Guard SHIPPED

When source dict supplies (exit_reason='TP', pnl_pct<0) or (SL, pnl>0), writer trusts pnl sign and logs won_pnl_contradiction: WARNING. Applied to both atomic writers (resolver + mysql_close_trade).

commit 22b677c1167 · alpha_engine/outcome_resolver.py:1670, audit_trail/mysql_client.py:628

Commodity Tooltips NEW

Friendly-name + contract-spec hover tooltips for 40+ yfinance =F tickers: CT=F → Cotton Futures, ES=F → S&P E-mini, ZN=F → 10Y T-Note, GC=F → Gold, etc. Auto-attached on /audit load.

audit_dashboard/dashboard_enhancements.js · function attachCommodityTooltips()

Cotton Sizing Brief RESEARCH

Per-trade economics for cot_positioning + CT=F at 1-contract notional. Margin tiers, expected net per-trade $3.40-$13.40, risk-of-ruin math at $5k/$10k/$25k starter, micro-contract status (none — full-size only).

reports/cotton_cot_real_money_sizing_2026-05-12.md

7-Step Testing Plan (COT) RESEARCH

Pre-real-money validation pipeline for cot_positioning + CT=F: reproducibility, data-integrity, walk-forward CPCV, conservative DSR, sample-window robustness, 4-week paper-pilot, risk-of-ruin Monte Carlo. Disqualifying conditions enumerated.

reports/cot_paper_pilot_testing_plan_2026-05-12.md

Asset-Class Expansion Report RESEARCH

Scope for filling BOND (n=18), ETF (n=87), and FUTURES (5.9% WR). Key finding: symbol universes are adequate; 206 baby_strategies/ files exist with ZERO wired to production; BOND scanner exists but not in production cron.

reports/asset_class_expansion_2026-05-12.md

Audits & reference reports

DB Evidence-Graded Final (2026-05-08)

4-engine swarm evidence review (deepseek + kilo + claude + gemini) of 20 DB findings. Per-finding evidence verdicts, action priorities, retraction confirmations.

reports/db_evidence_graded_final_2026-05-08.md

SUPREME EDGE Master Plan

Authoritative 5-agent synthesis (Claude Code + Cursor + Copilot + Kimi + Codex) + DB Health remediation status table. The canonical real-money-readiness document.

updates/2026-05-11-money-maker-master-plan.html

Performance Charter v1.0

Canonical KPI/risk-cap/tier-threshold doc (2026-04-28). Tier-2 minimum (PF>1.5/WR>50/MDD<20) for sizing; Tier-1 Renaissance (PF>2/WR>55/MDD<10) is the long-run target.

docs/PERFORMANCE_CHARTER.md

Session Summary 2026-05-12

Finished tasks + remaining P0/P1/P2 + future suggestions + verbatim user-prompt chatlog excerpts.

reports/session_summary_2026-05-12.md

Active hard-blocks (won't emit / won't aggregate)

FOREX class FUTURES class MEMECOIN class

BLOCKED_ASSET_STRATEGY_PAIRS (audit_trail/quality_gates.py:1534): 30+ entries — goldmine 1x/2x/3x/4x/6x consensus across CRYPTO+EQUITY, kimi_signal_tracking, crypto_soc_*, forex_carry_momentum, myfxbook_retail_contrarian, MeanReversionBB, futures_momentum, full MEMECOIN class quarantine (29 entries).

BLOCKED_DIRECTION_TRIPLES (quality_gates.py:1687): quan_engine_swing LONG, ml_crypto_predictor SHORT, crypto_keltner_compression_expansion LONG.

BLOCKED_ASSET_STRATEGY_SYMBOL_TRIPLES (quality_gates.py:1687 precedes, NEW 2026-05-12): quan_engine MATICUSDT, KIMI_signal_tracker ETHUSDT/BTCUSDT, irb_hoffman ADAUSDT, funding_rate_carry ROBOUSDT.

PERMANENTLY_KILLED_STRATEGIES: yahoo_analyst_consensus, cta_tsmom_blend, binance_smart_money, hl_funding_fade, winner_pattern_precursor, + more.

Queued action items

P0 — sync_active_mysql_picks_to_json QUEUED

The missing upstream writer that should read ACTIVE at_raw_picks, detect TP/SL/time-exit per asset class, and feed new entries into closed_picks.json. Root cause of 0.09% raw-pick outcome coverage. Investigator-scoped 2026-05-12 (recommended: inline call from forward_validator.validate_picks(), new module at alpha_engine/active_picks_sync.py).

Estimate 2-3h with tests · reuses existing failover price fetchers

P0 — WON-vs-PnL backfill SQL DRAFT

Sign-coherence guard stops NEW contradiction rows. Existing contradicted rows in trading_picks still poison aggregates. Needs an UPDATE pass that re-computes status from pnl_pct for any (status='WON', pnl_pct<0) or (status='LOST', pnl_pct>0) row.

Requires user sign-off for DB write

P1 — meta_strategy CRYPTO blanket block QUEUED

1.6M template rows across ~140 symbol/dir pairs in bt_backtest_trades. Defer until db_health.json::ghost_rows.top_cohorts repopulates after the CI commit-list fix lands (next 1-2 cron cycles), then decide whether blanket-block or symbol-triple enumeration.

Linked: commit d317560ac9c commit-list fix

P1 — BOND scanner production wiring QUEUED

alpha_engine/bond_scanner.py exists with 3 strategies (yield_momentum, duration_rotation, mean_reversion); not visibly wired into production cron. Wiring should lift BOND n from 18 → 50+ within 2 weeks.

Universe ready (14 symbols at config.py:721)

P1 — baby-strats batch backtest QUEUED

206 files in baby_strategies/, ZERO wired to production. Surface a batch DSR runner (anti_overfit_audit_sidecar.py over baby_strategies/*) to find DSR-real candidates.

Massive untapped pipeline

P2 — COT 7-step testing plan execution PASSIVE

Steps 1-5 (~6h active work) + Step 6 (4-week passive paper-pilot, running) + Step 7 (risk-of-ruin Monte Carlo). Gates the only currently-DSR-verified single-class deviation candidate (cot_positioning + CT=F).

reports/cot_paper_pilot_testing_plan_2026-05-12.md

NFA — Not Financial Advice. This page enumerates research tools and pre-real-money readiness gates. No live capital should be sized from anything listed here without (1) explicit user greenlight, (2) all 10 readiness-gate steps clear on the specific strategy/symbol, (3) DSR≥0.95 verified, (4) paper-pilot graduation gate cleared on paper_pilot.html.

Generated 2026-05-12 by Claude Opus 4.7 (caveman mode). Linked from /audit nav pill. Source: audit_dashboard/real_money.html (DO NOT edit audit_dashboard/index.html — auto-generated).