| name | money-maker-ready-June112026edition |
| description | The June-2026 EDITION of the money-ready program — executes the self-correcting MASTER LOOP (docs/MONEY_READY_MASTER_LOOP_2026-06.md) that converges toward 2-3 profitable asset classes even from 0/9. Use when the user says "/money-maker-ready-June112026edition", "run the master loop", "weekly money-ready cycle", or at the monthly edition review. Inherits data sources + hard rules from /money-maker-ready and /money-maker-readyv2. Aliases - mmr-june2026, master-loop, money-loop. |
/money-maker-ready-June112026edition — the Master Loop (June 2026 edition)
⚠️ SUPERSEDED 2026-07-11 by money-maker-ready-July112026edition (see reports/edition_review_2026-07.md). The June structural bet — mine the crypto ledger for directional alpha — was EXHAUSTED (no net-of-cost alpha; root cause = entry_price P0 + ~90% unresolved). The July edition pivots the primary track to ETF tactical asset allocation (the one validated find) + treats the crypto ledger as fix-before-mine. Use the July edition. This block below is retained for history.
Canonical plan: docs/MONEY_READY_MASTER_LOOP_2026-06.md — READ IT FIRST; this skill is the executor.
Edition discipline: this is a DATED edition. On the 11th of each month, run the edition review (Section E below) and either re-stamp or supersede with a new edition skill. Editions never silently mutate — improvements are visible diffs.
⛔ MANDATORY — ALL ASSET CLASSES + ALL DATA SOURCES (do NOT tunnel on crypto)
The #1 recurring failure of agents on this repo is looking ONLY at the crypto at_signal_outcomes ledger and concluding "no edge." That ledger is ~93% crypto AND is contaminated (see the entry_price P0 below). Concluding "no edge across all classes" from crypto alone is WRONG and has cost months. Before ANY "no edge" verdict you MUST have surveyed, at minimum:
- All 9 databases (creds
/home/eaguiar2015/dbpasses.txt, convention ejaguiar1_<name> / <name>1234560 @ the tools/db_env.py host — NEVER echo/commit): ejaguiar1_stocks (at_signal_outcomes, daily_prices, crypto_ohlcv, futures_daily_ohlcv, equity_daily_ohlcv), ejaguiar1_backtests (bt_backtest_trades 32.7M rows w/ real entry+exit+TP+SL, bt_backtest_runs 285 aggregated PF/Sharpe), ejaguiar1_memecoin (58 tables: bt100_results, mc_winners…), ejaguiar1_news (sentiment), plus events/deals/favcreators. (Sports ejaguiar1_sportsbet is a separate goal — operator may de-scope.)
- Every asset class: CRYPTO, MEMECOIN, EQUITY, ETF, FOREX, COMMODITY, FUTURES, BOND — AND the cross-cutting SOURCES the operator expects edge from: copytraders / public trades (Hyperliquid
copy_hl, multi_asset_copytrader), stock fundamentals / value, prediction markets (Kalshi / Polymarket copy_pm, prediction_market_consensus). If a class/source has too few resolved rows to judge, say so and check whether the emitter is wired + resolving — do NOT silently omit it.
P0 DATA-INTEGRITY GATE (read reports/DATA_INTEGRITY_entry_price_2026-07-03.md): at_signal_outcomes.entry_price is only ~29% clean (37% >10% off bar, 7% >50%), systematically +1.3% (inflates SHORT / deflates LONG), and intrabar_pnl_pct RIDES it. Prefer bt_backtest_trades (real entry+exit) or re-resolve from bar-aligned NEXT-bar entries. Every candidate must pass the 3 mandatory controls (memory feedback-entry-price-contamination-regime-2026-07-03): (1) entry_price vs OHLCV-bar integrity, (2) regime control vs matched-random entries + check market direction, (3) look-ahead control (shift entry to next bar — signal-bar entry is look-ahead-biased).
What this skill does (one weekly cycle)
- MEASURE — refresh the honest ledger + coverage metrics; run the H1 structural audit:
python3 tools/build_intrabar_truth_by_class.py --stdout
python3 tools/stamp_entry_conditions.py --stdout
python3 tools/check_one_sided_resolution.py
- DIAGNOSE — score H1-H5 per focus class (table in master MD §3). H1 red = halt everything else.
- ACT (parallel) — for each focus class (currently CRYPTO + COMMODITY), run the top remedy:
- Replay-variant batches via the proven harness (mirror
reports/strategy_bt_crypto_2026-06-11.json methodology: entry-anchored first-touch, SL-wins-ties, pre-entry features only, per-symbol-day dedup, net of costs). Pre-register the batch (hypothesis + falsification) BEFORE running; the batch is ONE FDR family; family closes after its registered comparisons.
- Plumbing/data fixes ship same-day with tests (the #129 discipline).
- FORWARD — check the pre-registered checkpoint calendar (master MD §7); promote/kill ONLY at the bars (95% CI lower bound of net PF > 1.15 at n≥80 forward + time-split + concentration<35%).
- RATCHET — commit the weekly scorecard to
reports/weekly_loop_scorecard_<date>.md; file/resolve incidents via tools/audit_pick_funnel/cli_track.py; update the live pages.
Data + credentials (for ANY agent, including brand-new ones)
- Read the orientation in master MD §0. DBs via
tools/db_env.py ONLY; local agents get passwords from /home/eaguiar2015/dbpasses.txt (gitignored — NEVER commit/echo); remote agents ask the operator.
- Backup to
ejaguiar1_backups before ANY table mutation (tools/db_backup_to_backups.py; ≤64-char table names; FK tables need CREATE-AS-SELECT copy).
- All source-of-truth pointers, tier definitions, mandatory data-integrity filters, and the reject-without-reverify list: inherit from
/money-maker-ready + /money-maker-readyv2 (do not duplicate here).
Hard rules (non-negotiable, inherited + edition-specific)
- Every claim:
(asset_class | n | timeframe). Direct-SQL re-verify any number a subagent/peer/LLM produces.
- Pre-register before backtest (M-107). Tuning families close after their registered comparisons — no variant-fishing.
- Mutate-before-kill for strategies; do-not-relitigate list is binding (master MD §8).
- Promotion is FORWARD-lane only. Replay results select candidates; they never size anything.
- One focus-class slot rotates out after 3 consecutive null weekly cycles.
E. Monthly edition review (run on the 11th)
- Score the month: per focus class — did the CI lower bound improve? checkpoints hit/missed? incidents P0 aging?
- What circled (3-null rotations, refuted batches) — add to do-not-relitigate.
- What the next edition changes (one structural change max — editions evolve, not churn).
- Write
reports/edition_review_<YYYY-MM>.md, supersede or re-stamp the skill, update the master MD header.
- Hand the operator the external-review task spec (master MD §9) if the edition made structural changes.
Failure-hypothesis quick reference (full table: master MD §3)
H1 measurement → halt + fix · H2 backtest-only → shadow + close family · H3 data scarcity → free APIs (FRED/CFTC/EDGAR) + shadow-lane universe widening · H4 external signals → per-source scorecards keep/kill · H5 coverage → extend resolution before judging.
Current state snapshot (2026-07-04 — RE-VERIFY, never trust this block after ~1 week)
VERDICT: no net-of-cost systematic ALPHA exists on the internal ledger / free data (exhaustive, swarm-confirmed). Every crypto-ledger candidate (luxalgo SHORT, rsi5070, mega_mutation, funding, etc.) dissolved under the 3 controls above. Root cause = the entry_price P0 + ~90% of positions never resolving, NOT missing strategies. STOP re-mining the crypto ledger for directional alpha until it is re-resolved from bar-aligned entries.
THE ONE REAL FIND (2026-07-04): asset-class MOMENTUM ROTATION (dual-momentum / TAA). Hold top-5 asset-class ETFs by 9-month momentum, monthly, abs-momentum-filter→bonds. Universe = 14 liquid free ETFs (etf_daily_ohlcv, yfinance). Robust: ALL 16 (top-N×lookback) grid cells cut MaxDD to −12/−17% vs SPY −24%; 9-month region Sharpe 1.09-1.21 / Calmar 0.81-1.10 vs SPY 1.00/0.68; both-halves+. It is smart-beta (better RISK-ADJUSTED return, NOT excess return); in-sample-robust → forward-track, don't over-size. Tool: tools/tactical_rotation_tracker.py. Report: reports/TACTICAL_ROTATION_EDGE_2026-07-04.md.
LIVE PROOF-OF-CONCEPT (measurable, in DB): table ejaguiar1_stocks.poc_picks, poc_id tactical_asset_rotation_v1 — 5 picks EEM/IWM/DBC/QQQ/EFA @20% each, entry 2026-07-04, measurement_date 2026-07-18, benchmark SPY. Entry prices locked. CHECKPOINT 2026-07-18 (cron/reminder set): fetch current prices, UPDATE poc_picks.exit_price/pnl_pct/status, compare basket vs SPY. (2wk is too short for a monthly TAA verdict — it's a liveness/plumbing read; the real gate is 6-12mo forward.)
Also deployed: diversified BETA portfolio (tools/beta_portfolio_tracker.py, real ETF sleeves SPY/DBC/AGG/GLD/BTC, inverse-vol + light crash guard) — the honest floor (beta, not alpha).
ML LAYER AUDIT (2026-07-04, reports/ML_AUDIT_2026-07-04.md): the ML layer is globally HALTED (ml_trading_enabled=False), stale (models last trained 2026-06-03, health check 06-21), and its "elite" WRs were FABRICATED — 4/6 hardcoded strategies have 0-2 resolved trades; ml_enhanced_* aggregate is honest 28% WR / −0.96% net (claimed 87-94%). FIXED: emptied the fabricated ML_PROVEN_STRATEGIES override in elite_scorer.py (was force-scoring losers to elite). OPEN tweaks: purge ml_strategy_reviver hardcoded dict; halt-emission-or-retrain ml_crypto_predictor (still emitting −0.96% picks); wire-or-retire ml_gatekeeper (shadow + 100% synthetic A/B log); restart ml_health monitor. Do NOT trust any ML WR until retrained on clean data + validated look-ahead-free (3 controls).
NEXT STEPS (keep this block updated): (1) 2026-07-18 measure poc_picks; (2) forward-track tactical rotation 6-12mo; (3) re-resolve the crypto ledger from bar-aligned NEXT-bar entries (the only path to trust any ledger candidate) — but note the OPEN backlog is systematic losers (dead end for backlog resolution); (4) mine codebase strategy docs (reports/etf_strategy_catalog.md, high_sharpe_strategies_report.md, academic_trading_strategies.md); (5) ML-algorithm audit in flight (utilization/staleness/fabricated-WR).
