Cross-Asset Quant Strategy — Pre / During / Post Framework

The playbook that wires this project’s skills + API clients into one process across equities, forex (research-only), commodities, options, and thematic plays.

This framework is research-and-decision-support only. All execution stays manual — every artifact carries manual_review_required: true and a data_gaps[] array. You always pull the trigger.

Table of contents

The spine — applies to every asset class

Three loops run continuously:

Loop Cadence Output
Macro regime Weekly Risk-on / risk-off / transitioning posture
Idea generation Daily Ranked candidates with hypothesis cards
Position lifecycle Per trade Pre → During → Post

The 7-layer signal stack

A trade only earns capital if layers 1–5 align:

1. Macro regime          → BISClient, BLSClient, BEAClient, EIAClient
2. Theme / sector        → theme-detector, sector-analyst
3. Asset-class screener  → VCP / CANSLIM / PEAD / Dividend / Parabolic
4. Setup confirmation    → technical-analyst (chart), breakout-trade-planner
5. What's priced in      → PolymarketClient, NewsClient, FMP consensus
                          (see references/what-is-priced-in-framework.md)
6. Sizing                → position-sizer (R-multiples), exposure-coach
7. Postmortem            → signal-postmortem, trader-memory-core

The kill rule

Every trade must have a written thesis before entry, a kill criterion (what makes this wrong), and a journal entry after exit. No exceptions. This is the manual review gate.

1. Equities — Industry / Single Stock / Sector

Scope: US-listed stocks + sector/thematic ETFs. Swing (3–30 days) or position (1–6 months).

PRE

Step Tool Output
1. Idea source vcp-screener (bullish swing), canslim-screener (growth), pead-screener (earnings drift), parabolic-short-trade-planner (mean-reversion shorts), dividend-growth-pullback-screener (quality dip) Ranked candidates
2. Theme context theme-detector Filter to top-3 accelerating themes
3. Chart confirmation technical-analyst (chart image) Yes/no on visual pattern
4. Trigger levels breakout-trade-planner 5-min ORL, breakout extension, invalidation level
5. Earnings reaction history earnings-trade-analyzer Don’t fight buy-rumor-sell-news patterns
6. What’s priced in PolymarketClient, NewsClient, FMP consensus vs your estimate Gap = (your view − consensus) × reaction function
7. Hypothesis card trade-hypothesis-ideator → trader-memory-core IDEA → ENTRY_READY Thesis + kill criterion + R-target + time-stop
8. Sizing position-sizer (stop-loss or Kelly) Shares + dollar risk
9. Portfolio check exposure-coach Sector cap, breadth posture, ceiling %

DURING

  • Daily morning: run market-regime-daily workflow (breadth + uptrend + exposure). If composite drops a tier → tighten stops, no new entries.
  • News monitor: NewsClient.get_market_news(tickers=[...], days=1) — any thesis-killing headline.
  • PEAD plays: watch SIGNAL_READY → BREAKOUT transitions.
  • Earnings during hold: size down to ⅓ or close before print, unless earnings is the thesis.

POST

  • trader-memory-core → CLOSED with realized R, MAE/MFE.
  • signal-postmortem: what worked, what didn’t, was the kill criterion respected, was sizing correct.
  • Roll lessons into monthly-performance-review.

2. Forex — Research Only

Scope: Directional bias for USD/JPY, EUR/USD, GBP/USD, AUD/USD, USD/CAD. Output is a research artifact, not orders. Execution lives in a separate project; this project must never import from it.

PRE (research)

Step Tool Signal
1. Rate differential BISClient.rate_differential("US", "JP") Carry tailwind / headwind in pp
2. US macro BLSClient.get_named("unemployment_rate"), BLSClient.get_named("cpi_core"), BEAClient.real_gdp_growth() Fed hawkish/dovish bias
3. Counterparty macro EStatClient.cpi_national() for JPY; commodity for AUD/CAD Cross-country surprise potential
4. Commodity beta EIAClient.natural_gas_spot(), CommodityClient.latest(["WTI", "GOLD"]) AUD = iron-ore + copper; CAD = WTI; gold = USD inverse
5. Catalyst pricing PolymarketClient.search_markets("Fed cut") Implied probability of policy move
6. Research artifact Markdown report with manual_review_required: true + data_gaps[] Directional bias score (−5 … +5), no trade ticket

DURING (monitoring research)

  • Watch BLS NFP, CPI prints (Finnhub econ calendar)
  • BIS rate revisions monthly
  • News via NewsClient

POST (research calibration)

  • Did the rate-differential thesis play out?
  • Calibrate: which BIS countries / BLS series have the highest hit rate for predicting direction?
  • Log to trader-memory-core with research_only: true

Explicitly OUT

Order placement, stops, broker code — lives in the separate forex project. Release gate enforces this.

3. Commodities

Scope: Energy (WTI, Brent, nat gas), precious (gold, silver), industrial (copper). Trade via equity proxy ETFs/stocks (XLE, USO, GLD, GDX, FCX) — futures are out of scope.

PRE

Step Tool Signal
1. Fundamental driver EIAClient.electricity_demand("PJM") (AI power demand), EIAClient.natural_gas_spot() (gas), EIAClient.power_demand_yoy() YoY inflection?
2. Spot prices CommodityClient.latest(["BRENT", "GOLD", "COPPER"]) Where are we vs 30-day range
3. Series trend CommodityClient.time_series("BRENT", start, end) Direction over last 30d
4. Theme context theme-detector — Oil & Gas, Gold & Precious Metals, Power Infrastructure Confirm theme heat ≥ 60, lifecycle = Accelerating
5. Spark spread (for IPPs) (power_price) − (gas_price × heat_rate) using EIA data Widening = bullish VST/NRG/TLN; compressing = bearish
6. Equity expression Pick proxies — energy long: XLE/XOP/OIH or VST/CEG/EQT; gold: GDX/GLD/NEM; copper: FCX One equity per thesis
7. What’s priced in Polymarket OPEC + Fed cut markets, implied vol of commodity ETF Surprise potential

DURING

  • Spark spread on a weekly watch — widening confirms thesis on IPPs
  • News for geopolitical (Mid-East, Russia, OPEC) via NewsClient
  • EIA inventory prints (weekly)

POST

  • Did EIA prints confirm the demand thesis?
  • Calibrate your power-demand model vs actual
  • Roll to trader-memory-core

4. Options

Scope: Defined-risk strategies on liquid US tickers. No naked options (uncapped risk).

PRE

1. Underlying must already be a §1 high-conviction equity setup. Options are an expression, not a thesis source.

2. Strategy selection — decision tree:

Bias IV regime Strategy
Bullish High IV Bull put spread (sell premium, defined risk)
Bullish Low IV Bull call spread or long call (buy cheap directional)
Bullish + own stock any Covered call (yield enhancement)
Range-bound High IV Iron condor
Bearish High IV Bear call spread
Bearish Low IV Bear put spread or long put

3. Validation: options-strategy-advisor for Black-Scholes pricing + Greeks + scenario analysis. Probability of profit ≥ 60% as the floor.

4. Sizing: max loss per trade = position-sizer R amount. Net debit OR max-loss of credit spread ≤ R.

5. What’s priced in: IV percentile vs 1Y range (premium-rich vs cheap). At earnings: straddle price = market’s expected move — compare to your own move estimate.

DURING

  • Monitor delta (directional risk) and theta (decay benefit)
  • Close winners at 50% max profit (don’t squeeze last 25%)
  • Adjust on tested short strike: roll out + down/up
  • Earnings during hold: close or roll. Never hold short-vol through binary event.

POST

  • Realized P&L vs theoretical max
  • Greek attribution: how much P&L came from delta vs theta vs vega
  • signal-postmortem with IV regime context

5. Thematic / Sector Rotation

Scope: Multi-week thematic plays via ETFs or basket of top constituents.

PRE

Step Tool Output
1. Theme scan theme-detector --dynamic-stocks Ranked themes by heat + lifecycle
2. Lifecycle filter Avoid Exhausting, favor Emerging / Accelerating Late-cycle is crowded trade risk
3. Constituent picks Top 3-5 stocks per theme by relative strength Equal-weight basket
4. ETF expression Proxy ETF from skills/theme-detector/references/cross_sector_themes.md Single-line expression
5. Industry confirmation Industry rank top quartile on FINVIZ Cross-check on theme detector output
6. Volume confirmation PolygonClient.get_grouped_daily() Real institutional flow

DURING

  • Daily breadth check (market-breadth-analyzer) — themes are first to roll when breadth contracts
  • Theme lifecycle drift watch: exit before Exhausting, not after

POST

  • Did the proxy ETF outperform SPY by ≥ 5% over the hold?
  • Update theme model with realized success/failure

Cadence — when each routine runs

When Routine Skills / Workflows
15 min before US open Daily regime check market-regime-daily workflow
Daily, post-close News scan, position monitor NewsClient.get_market_news(tickers=open_positions)
Daily (post-regime) Multi-asset opportunity scan multi-asset-opportunity-daily workflow (this playbook in flow form)
Weekly (Sunday) Fresh candidates + macro refresh swing-opportunity-daily, theme-detector, BIS/BLS/BEA refresh
Monthly Performance + calibration monthly-performance-review, signal-postmortem aggregation
Per trade — Pre Hypothesis card + sizing trade-hypothesis-ideator, position-sizer, exposure-coach, trader-memory-core IDEA → ENTRY_READY
Per trade — During Daily kill check, news monitor trader-memory-core review-due, NewsClient
Per trade — Post Close, MAE/MFE, lessons trader-memory-core CLOSED, signal-postmortem

Toolchain map — every step has exactly one tool

Need Tool
OHLCV (replace yfinance) PolygonClient.get_aggs()
US macro (GDP, savings) BEAClient
US labor + inflation BLSClient
Cross-country rates BISClient.rate_differential()
Energy / power EIAClient
Commodity spot CommodityClient.latest()
Japan macro EStatClient
News + sentiment NewsClient (Marketaux + Newsdata)
Catalyst probability PolymarketClient
Calendars (econ + earnings) FinnhubClient (free) or FMP
Theme detection theme-detector skill
Stock screening VCP / CANSLIM / PEAD / Dividend / Parabolic skills
Hypothesis card trade-hypothesis-ideator
Sizing position-sizer
Posture / exposure exposure-coach
Memory / journal trader-memory-core
Postmortem signal-postmortem
Backtest backtest-expert

Out-of-scope (the manual review gate)

  • Auto-trade / broker execution / order placement
  • OANDA / forex execution (lives in a separate project)
  • Binance / crypto auto-trade
  • Any closed-loop without human sign-off

Every artifact this framework produces carries manual_review_required: true and data_gaps[]. You always pull the trigger.

Reference docs in the project

  • skills/theme-detector/references/cross_sector_themes.md — theme definitions + constituents
  • skills/theme-detector/references/energy-power-market-signals.md — spark spread, capacity auctions, LMP mechanics
  • skills/trade-hypothesis-ideator/references/what-is-priced-in-framework.md — three model-to-trade frameworks
  • scripts/api_clients/README.md — full API client catalog
  • workflows/ — canonical workflow definitions