← 90-Day Plan landing  |  Updates feed

Asset Class 90-Day Plan — EQUITY — 2026-05-15

Senior Quant Audit using money-maker-continual-improve skill (7-step process).
Focus: EQUITY crossed T2 threshold (PF 1.57 / WR 51.9% / n=420) but narrow speculative universe (18 tickers heavy on pennies/memes), limited fundamentals via yfinance wrapper (no primary SEC), strong hidden VIX regime edge in research not wired to prod, backtest/live gaps. Prioritize clean large-cap + regime filter before any sizing expansion. Not system pilot (COMMODITY CT=F leads PF 2.49 / WR 61.5% n=322).
Data sources: audit_dashboard/data/dashboard_data.json (2026-05-15T02:28Z), reports/MASTER_ACTION_PLAN_2026-05-15.md, alpha_engine/config.py (EQUITY_SYMBOLS + CATEGORY_RISK), alpha_engine/equity_strategies.py, alpha_engine/equity_factor_model.py, alpha_engine/non_crypto_quality_gate.py, alpha_engine/scanner.py, reports/asset_class_research_EQUITY_2026_05_12_0438Z.md, reports/equity_vix_regime_breakthrough_20260513.md + equity_momentum_vix_regime_backtest.json, reports/equity_yc_regime_breakthrough_20260513.md, reports/equity_vix_yc_combined_super_breakthrough_20260513.md, reports/fix_EQUITY_20260505T005402Z.md, reports/audit_asset_feedback_2026-05-05T0121Z_EQUITY.md, reports/asset_class_research_EQUITY_2026_05_12_0438Z.md, local git branches (feat/equity-vix-regime-gate-sidecar-2026-05-13), audit_dashboard/data/equity_top*.json, audit_dashboard/data/db_health.json, reports/daily_ideas_synthesis_2026-05-15.md.

Current Charter context (from CLAUDE.md / master plan / dashboard): EQUITY T2-candidate (May-3 snapshot PF 1.41 / WR 52.7 / n=421; today improved PF 1.57 / WR 51.9 / n=420 resolved, total_pnl +385.5%, sizing_allowed=true, stable). by_asset_class shows closed ~1050. Concentration OK: top_symbol AMD 11.65% PnL mass, top_strategy "Classic Momentum" 10.96%. Master plan flags "NEEDS-EVIDENCE-FIRST" for 2026-05-25 institutional (real T2 but require 30d clean rolling + MATCH on verify_system_pf before sizing; repeats claude_gainer_st pattern). Pending: M-009 (PEAD on top-100), M-025 (overnight_intraday_reversal EQUITY), M-026 (day-of-week Tue/Wed tilt). VIX regime sidecar branch exists but not merged/wired. Penny-stock kimi-EQUITY hallucinated names noted in drop list. Hermes unblocked some zero-evidence equities.

Step 1: Establish Brutal Baseline (EQUITY)

Live performance (dashboard_data.json::performance.asset_class_health.EQUITY + by_asset_class.EQUITY + asset_class_concentration.EQUITY):

Symbol universe coverage and quality (large-cap vs mid vs penny/meme):

Earnings, SEC filings, or fundamental data usage:

Outcome tracking / DB reality:

GitHub Actions / Data flow + recent activity:

Summary baseline: PF 1.57 meets T2 floor on modest n=420, driven by Classic Momentum on AMD + factor/PEAD + SPY200+VIX gates. But universe is narrow speculative (18 tickers, 8 pennies/memes), fundamentals shallow (yf only), research on 30 LC + VIX<20 shows Tier-1 PF 5.37/MDD 7.3% (n=88 periods) vs live reality. Backtest optimism (MDD 24% baseline) not reflected in prod. Better than FOREX (PF 0.81), competitive with ETF (1.48), but far from COMMODITY flagship. Hidden edge in regime filter + clean universe expansion is the lever. Risks: penny gaps, concentration, yf fragility, no costs model yet (M-018 pending).

Step 2: Identify the Single Best Pilot Candidate (within EQUITY)

Filters: Live PF>1.5 + WR>50% + n≥100 post-noise; clear explainable edge (academic + regime); feasible infra (no new paid data first); protectable (VIX gate + liquidity).

Gap vs current: Prod uses 18 mixed (pennies drag), no VIX hard filter (only adj + macro), research 30 LC not prod.

Step 3: Gap Analysis (EQUITY)

Gap Type Diagnostic (data-backed) Severity Typical Fix
Symbol universe quality (narrow + penny/meme drag) 18 tickers in config.py (8 pennies/memes: NIO/LCID/RIVN/PLTR/SNDL/GME/AMC heavy beta/gap risk); research uses 30 clean LC (no overlap full); no ADV/liquidity runtime gate; Hermes blocks "zero-evidence"; concentration on AMD 12%; kimi hallucinated names. Live n=420 but from speculative names. P0 Separate LARGE_CAP_EQUITY (20-30 liquid names from research list + ADV>$5M filter) vs PENNY_MEME (quarantine or 0.1% micro only). New list in config.py; gate in scanner.py / equity_strategies.py before emit. Deprecate meme/penny volume from main EQUITY class.
Fundamentals / earnings / SEC yf wrapper only (PE/ROE/surprise in equity_factor_model.py, 6h cache, no PIT guarantee); earnings_gap uses sector ETF proxy not real per-name calendar; no EDGAR/10K/insider/13D in core pipeline (vt_baby only); M-009 PEAD top-100 pending (earnings-calendar partial). P1 Harden yf with failover + validate PIT; add free earnings calendar (e.g. via yf or FMP low-cost); wire real PEAD strategy (M-009) on liquid core only; spike FRED macro (M-032) for value/quality context. No paid SEC APIs first.
Hidden regime / factor not wired VIX<20/22 filter delivers Tier-1 PF 4.5-5.4 / MDD<17% (backtest json + reports/equity_vix_*_20260513.md) on 30 LC; YC weaker; prod only has soft vix_confidence_adj + SPY200 gate in non_crypto_quality_gate.py; regime_filtered_momentum optional in _RAW_EQUITY_STRATEGIES but not default; feat/equity-vix-regime-gate-sidecar-2026-05-13 unmerged. P0 Wire VIX hard filter (opt-in sidecar per Wire-Up) in equity_strategies + scanner / quality_gates.passes_active_gate; backtest on actual 18 + expanded LC; promote to default for momentum strats. Add to equity_macro_gate.
Backtest vs forward / execution realism Research PF 2.8+ / low MDD with filter on clean univ; live 1.57 on narrow+spec; no slippage/position_sizer (M-017/018 scaffolds in PR #1026 not wired for equity); CATEGORY_RISK loose on penny; costs eat edge on volatile names. P1 Wire slippage_validator + vol-target sizer (M-017/018) for EQUITY; add 5-20bp friction to all equity backtests; re-run VIX filter + momentum on prod symbols + costs.
Strategy bloat / thin n ~15-20 equity strats (momentum/penny/meme/quality/connors/rsi/gap/turn_of_month + community); "Classic Momentum" dominant live; many low-n or unproven (earnings_gap on sectors only); n=420 modest vs CRYPTO 8k. P2 Prune via quality_gates (delete low-PF on penny sub if any); promote 4-5 core (momentum + connors + quality + intermarket + vix-filtered); focus volume on liquid pilot.
Data freshness / missing low-cost yf rate limits/fails (failover exists); no FRED yet (M-032 pending); Stooq key-gated; DB checks blocked by pymysql in some envs but ghost=0 good. P2 Cache yf/stooq aggressively; add FRED secret + macro filter (DXY/yield for equity value); extend DB Freshness Guardian to equity tables.

Why PF 1.57 but not higher / more robust? Narrow + speculative universe (pennies/memes inflate vol, gaps, correlation to retail sentiment not factors); research optimism on clean 30 LC + VIX regime not in prod; shallow fundamentals (no SEC depth); no costs model; momentum works but fragile without hard VIX gate (crashes kill it, as in baseline MDD 24%). Factor/PEAD marginal. Matches "data quality > model sophistication".

Step 4: Decision Framework for Strategies (live >60d)

Strategies to delete/invert/promote (file paths): - Quarantine/delete: speculative penny/meme emitters if PF<1.2 30d; hallucinated kimi names → alpha_engine/strategy_blocklist.py, config.py (move to PENNY_MEME_SYMBOLS separate class), audit_trail/quality_gates.py (new per-class quarantine). - Promote/wire: VIX regime filter (from feat/equity-vix-regime-gate-sidecar + backtest tool) → alpha_engine/equity_strategies.py + non_crypto_quality_gate.py (extend equity_macro_gate) + scanner.py; equity_factor_model.py already good. - Invert/test: RSI2 overbought short on high-VIX (branch). - Reference: equity_strategies.py:1232 (_RAW + wrap), non_crypto_quality_gate.py:136 (macro + vix_adj), config.py:587 (universe).

Step 5: Leverage AI Keys Intelligently (High-ROI Prompts)

(Per skill: use for feature ideation, deconstruction, data discovery — not production code gen without verify.)

Feature Ideation (EQUITY):

For EQUITIES (large-cap US), give 8-10 high-signal, low-cost features from peer-reviewed papers 2018-2025 (e.g. momentum, value, quality, PEAD, volatility timing, yield-curve). For each: exact free data source (FRED series, Yahoo, CBOE VIX, earnings calendar), statistical validation test (e.g. t-stat, DSR), and realistic implementation cost (API calls/day).

Strategy Deconstruction:

Summarize core mechanics, entry/exit, risk controls, failure modes, and tx cost assumptions (5-20bp) of: (1) 12-1 momentum (Jegadeesh-Titman) on S&P100, (2) Connors RSI-2 on SPY/QQQ, (3) VIX<22 regime overlay on momentum. Include walk-forward decay expectations.

Data Source Discovery:

What free/near-free 2026 sources for US equity: real-time earnings calendar + surprise (beyond yf), full SEC EDGAR 10-Q text embeddings (free tier?), point-in-time Russell 1000 constituents, FRED macro (ISM, GDP, rates) for factor overlay? Quant shop scale.

Pruning / Inversion:

Given EQUITY strategy "penny_volume_breakout" on NIO/LCID (high beta, n=XX, realized PF 0.9 WR 44% last 60d), should we delete, invert to short, or tighten (liquidity gate + VIX<25 only)? 3 concrete recs with turnover/edge impact.

Step 6: Low-Cost Data & Validation Stack

Step 7: Produce a Focused Output (90-day plan)

Pilot Asset Class rec: VIX-regime + clean Large-Cap Momentum (not broad EQUITY). COMMODITY remains system pilot.

Top 5 gaps (P0/P1): 1. P0: Narrow/speculative universe (18 mixed tickers) vs research 30 LC — primary drag. 2. P0: VIX regime Tier-1 edge (PF 5+ / MDD 7%) discovered May-13 not wired (sidecar branch exists). 3. P1: Shallow fundamentals (yf wrapper, no SEC/earnings calendar primary). 4. P1: No execution realism (slippage/costs not in equity path; M-017/018 pending). 5. P2: Strategy bloat on pennies + modest n=420; DSR/effective n visibility low.

Specific actionable fixes (files): - Universe split + LC list + ADV gate: alpha_engine/config.py (new LARGE_CAP_EQUITY_SYMBOLS), alpha_engine/equity_strategies.py (filter in each fn), alpha_engine/scanner.py. - Wire VIX hard filter sidecar: alpha_engine/equity_strategies.py (new vix_regime_momentum wrapper), non_crypto_quality_gate.py (extend gate), audit_trail/quality_gates.py; merge from feat/equity-vix-regime-gate-sidecar-2026-05-13 with Wiring Plan. - PEAD real + earnings: alpha_engine/equity_factor_model.py (add calendar source), new alpha_engine/strategies/pead_equity.py (M-009). - Costs + sizer: wire PR #1026 scaffolds into equity path. - Prune: audit_trail/quality_gates.py (EQUITY-specific quarantine for penny sub-PF).

Strategies to delete/invert/promote: - Quarantine: penny/meme volume strats if <1.2 PF 30d; old low-n gaps. - Promote: momentum + connors_rsi2 + vix-filtered + factor-wrapped core on LC only. - Invert: overbought short variants on high-VIX.

2-3 AI prompt templates: (see Step 5 above; tailored to VIX timing, PEAD calendar, LC universe expansion).

30/60/90-day execution (focused, 5-7 M-IDs max, Wire-Up + clean PRs only):

30 Days (P0 — Wire regime + clean universe foundation): - Create clean LARGE_CAP_EQUITY_SYMBOLS (20-25 from research 30 LC list + ADV filter) + PENNY_MEME_SYMBOLS (separate or research-only) in alpha_engine/config.py. Add is_liquid_equity() gate in scanner.py + equity_strategies.py (all 15+ fns). - Wire VIX regime hard filter (VIX<22 skip or 0-conf for non-exempt momentum strats) as opt-in sidecar from existing branch + backtest tool; add to non_crypto_quality_gate.equity_macro_gate + vix_confidence_adj. Backtest on new LC list + costs (5-10bp). - Scaffold minimal PEAD (M-009) using yf earnings_dates on LC core only (2d post-earnings window). - Enforce in quality_gates.passes_active_gate + score_booster for EQUITY. - Shadow paper 0.2-0.5% on LC + VIX filter (target n≥20 clean, PF>1.7). - Docs: Update reports/MASTER_ACTION_PLAN_2026-05-15.md (Section 21 + M-009/025/026 status, VIX wire); create this reports/asset_class_90day_plan_EQUITY_2026-05-15.md. Small PRs with Wiring Plan. - Success: Dashboard EQUITY concentration shifts to LC names (>70% PnL from top 10 liquid); VIX filter telemetry live; no new penny emissions in main class.

60 Days (P1 — Validate pilot + harden data): - 30d rolling paper on VIX-filtered LC momentum core: target PF>1.8 / WR>55 / MDD<15 / n≥50 / DSR>0.9 (vs COMMODITY benchmark). Full walk-forward + bootstrap CI. - Real earnings calendar integration (free source spike) + full factor_model validation (PIT check vs yf history). - Wire M-025 (overnight_intraday_reversal) + M-026 (DOW Tue/Wed bias) as opt-in on LC pilot only; test with VIX gate. - Add friction model + slippage_validator to all EQUITY backtests (M-018); re-validate VIX PF lift survives. - External replication: compare pilot to SPY/QQQ momentum public (AQR, Alpha Architect) or simple VIX-timed ETFs. - GHA + DB freshness for equity tables; effective_n / DSR per equity strat. - Success: Paper pilot clean 30d metrics meet T2+; live dashboard shows EQUITY PF lift 0.15-0.3+ with lower vol; M-009/025/026 evidence in master plan.

90 Days (P2 — Scale decision or quarantine further): - If pilot meets gates (30d clean rolling PF>1.8 WR>55 n≥100 post real costs/slippage, DSR≥0.95, corr<0.6 with COMMODITY, no MDD>15% breach): promote to 0.5-1% live micro sizing (skin-in-game prep per M-061), expand carefully to 40-50 LC (point-in-time Russell), full FRED macro overlay (M-032). Update master plan EQUITY institutional date. - Else (likely safer initial): Keep shrink (pennies/memes research-only or 0 vol), double-down on 15-20 LC + VIX/Connors/factor core; further prune bloat; consider "PENNY_STOCK" as distinct low-conviction class. No broad expansion. - Master plan: Close highest-priority (VIX wire, universe split, PEAD) with reproducible paper logs + dashboard delta; add M-0xx only for funding/earnings calendar + ADV gate. Reference Wire-Up Rule, clean branches, docs/MUTATION_THREE_AXIS_PROTOCOL.md (no silent kill). - Overall: EQUITY moves from "T2-candidate on shaky narrow speculative base" to "focused Tier-2 / near-Tier-1 on regime-filtered LC core" or "conservative micro-vol only". Contributes to phenomenal performance across classes. Prioritize COMMODITY to real-money Level 4/5 first.

Risks / Anti-patterns avoided: No spread across all 18+pennies; no trust research backtest (30 LC) without paper on prod symbols; no new complex models before VIX gate + liquidity split + costs; prune > add (delete/quarantine 5+ speculative strats); no production risk/execution code without verification + swarm. Follows "one asset class done properly".

Files to touch (Wire-Up compliant, small focused PRs, opt-in/sidecar preferred): alpha_engine/config.py (universe split + LC list), alpha_engine/equity_strategies.py (VIX wrapper + filters + factor apply), alpha_engine/non_crypto_quality_gate.py (hard VIX gate), alpha_engine/scanner.py + production_scanner.py (liquidity gate), alpha_engine/equity_factor_model.py (earnings calendar), audit_trail/quality_gates.py (EQUITY quarantine + M-004 analog), alpha_engine/strategies/pead_equity.py (new M-009), reports/MASTER_ACTION_PLAN_2026-05-15.md (update only). Reference feat/equity-vix-regime-gate-sidecar-2026-05-13 for cherry-pick. All via clean branches + swarm review per AGENTS.md.

Expected impact: Universe split + VIX filter: +0.3-0.8 PF / -50% MDD on pilot (research validated), shift 60%+ PnL to liquid LC names, lower gap risk; PEAD + costs: +0.1 PF sustainability; n stable or grow on quality. Path to T2+ / near-T1 on focused EQUITY within 90d. Positions for "phenomenal across ALL asset classes" Goal #1. Avoids FOREX-style sub-floor trap.

Next: Spawn deep-dive subagent for specific (e.g. penny autopsy or external replication SPY momentum + VIX timing funds) only if pilot paper fails. Update reports/MASTER_ACTION_PLAN_2026-05-15.md highest-priority only (VIX wire + universe). All changes small PRs + evidence.

References (absolute paths): - Data: /mnt/e/findtorontoevents_antigravity.ca/audit_dashboard/data/dashboard_data.json, db_health.json, equity_momentum_vix_regime_backtest.json, equity_top*.json - Master: /mnt/e/findtorontoevents_antigravity.ca/reports/MASTER_ACTION_PLAN_2026-05-15.md - Code: /mnt/e/findtorontoevents_antigravity.ca/alpha_engine/config.py:587 (EQUITY_SYMBOLS), equity_strategies.py:1232 (_RAW + wrap + earnings_gap), equity_factor_model.py:41 (yf + PEAD), non_crypto_quality_gate.py:136 (equity_macro_gate + vix_adj), scanner.py - Reports: asset_class_research_EQUITY_2026_05_12_0438Z.md, equity_vix_regime_breakthrough_20260513.md, equity_yc_regime_breakthrough_20260513.md, equity_vix_yc_combined_super_breakthrough_20260513.md, fix_EQUITY_20260505T005402Z.md, audit_asset_feedback_2026-05-05T0121Z_EQUITY.md - Branches: feat/equity-vix-regime-gate-sidecar-2026-05-13 (unmerged research)

This plan is ruthless, data-driven, focused on compounding real persistent edge (VIX-timed LC momentum + PEAD) while eliminating waste (penny/meme noise, unproven strats). Quality over quantity for EQUITY. One class done right.

Generated 2026-05-15 per money-maker-continual-improve skill invocation on EQUITY audit. Pilot for system remains COMMODITY (COT); this is EQUITY-specific 90d rescue-to-Tier-2 plan. Follows AGENTS.md / CLAUDE.md / SOUL.md / every-session reads of SOUL/USER/memory/2026-05-15.md + skill.

Source: reports/asset_class_90day_plan_EQUITY_2026-05-15.md · 90-Day Plan — May 15 2026 Edition · generated by tools/generate_90day_plan_pages.py