Model Card — Layered Covered Call Income Cycle

Strategy ID: lcc-income-cycle Status: Research — not yet cleared for productization. Walk-forward validation required. Intended Consumer: feature-developer (Raptor/Antlers implementation), software-architect (service design) Date: 2026-05-05

1. What This Strategy Does

The Layered Covered Call Income Cycle is a paper-trading strategy module that manages a share position paired with weekly covered calls distributed across multiple strike zones. The strategy tracks premium income, assignment events, tax-lot selection, and re-entry decisions as explicit, rule-based state transitions.

It is NOT a prediction engine. It does NOT recommend trades. It enforces a pre-committed rule set against a paper-trading position and records what happened.

2. Inputs Required

2.1 Per-Cycle Inputs (provided by the trader, pre-market)

{
  "sentiment_label": "string (Neutral | Bullish | Bearish | HighUncertainty)",
  "ticker": "string",
  "options_chain": {
    "expiration_date": "string (ISO 8601 date)",
    "strikes": [
      {
        "strike": "number",
        "bid": "number",
        "ask": "number",
        "delta": "number",
        "extrinsic": "number"
      }
    ]
  },
  "underlying_open_price": "number",
  "underlying_prior_close": "number"
}

2.2 System State Inputs (maintained by paper-trading engine)

• Current position record (shares, lots, open calls, state)
• Earnings calendar for the next 5 trading days
• VIX daily close (prior day) for gap classification proxy
• Commission config from SimulationConfig

2.3 Historical Data (for backtest only)

• AMZN daily OHLCV from Alpaca Market Data
• Historical options chain: ORATS or Tradier (confirmed license required)
• CBOE VIX daily series (public)

3. Outputs

3.1 Per-Cycle Outputs

{
  "cycle_id": "string (uuid4)",
  "proposed_call_structure": [
    {"zone": "string (ITM|ATM|OTM)", "contracts": "integer", "target_delta_range": "string"}
  ],
  "opening_sell_timing": {
    "market_condition": "string",
    "recommended_delay_minutes": "integer",
    "reason": "string"
  },
  "alerts": ["array of Alert objects"],
  "re_entry_decision": "ReentryDecision object or null"
}

3.2 Ledger Records (written per event)

• EquityLot — on every share acquisition
• ShortCall — on every call sell, update at close
• AssignmentEvent — on every assignment
• Cycle — one per income cycle, closed at expiration
• ReentryDecision — one per re-entry evaluation
• Alert — every alert fired

3.3 Dashboard Metrics (computed on demand from ledger)

See metrics.py for implementations of all 14 metrics.

4. What Can Fail

5. What to Monitor in Production (Phase 3+)

6. Compute and Storage Budget

Storage is append-only ledger records. Historical backtest artifacts are stored separately under docs/data-science/backtests/lcc-income-cycle/ and not part of the per-user runtime storage.

7. Broker Agnosticism

This module is broker-agnostic. It operates on: - Share counts and lot records (internal state) - Options chain data (passed in from whatever data source the platform uses) - Fill confirmation (paper fills provided by the MBT paper-trading engine)

No brokerage vendor name appears in user-facing copy. The broker is plumbing. Raxx is the strategy layer.

For live trading (Phase 4, after paper validation): the paper-trading engine's fill confirmations are replaced by actual broker fill events. The strategy module itself does not change.

8. Regulatory Notes

• This module produces rule-based, pre-committed operational outputs (structured automation) — not personalized investment advice. See feedback_deterministic_execution_ai_augments.md.
• All AI-augmented surfaces (context on prior cycles, pattern flags) are retrospective descriptions of what happened on the user's own simulation data. They do not predict future outcomes. See feedback_no_forward_looking_framing.md.
• Tax-lot simulation is illustrative. The system does not calculate taxes owed. A CPA disclaimer must appear on any UI surface that shows lot-method selection or assignment P/L.
• Options trading involves risk of loss. Paper-trading results do not guarantee live-trading outcomes. This disclaimer must appear in the dashboard.

9. Implementation Scope for v1 (Phase 1 + 2)

Phase 1 — Manual Paper Sim (minimal Raptor work): - Data schema (Postgres tables matching data-schema.md) - State machine service (Python, deterministic, no ML) - Alert log table + basic alert surfacing in Antlers dashboard - AMZN single-ticker support only

Phase 2 — Chain Ingestion + Delta Filters: - Options chain ingestion from Alpaca options API - Delta-based strike selection (pull nearest strike in each zone's delta range) - Earnings calendar check (Alpaca corporate events) - Dashboard metrics (all 14, computed from ledger)

Phase 3 — Recommendations + Alerts + Backtest: - Automated opening bell rule (condition classification, delay logic) - Full alert routing (push notifications for WARNING+) - Backtest runner (Python script, not real-time) - Walk-forward validation report

Phase 4 — Live Broker Integration: - After Phase 3 walk-forward passes - Alpaca live options order submission (or BYOB equivalent) - This phase is out of scope for the current research handoff

10. Open Questions Before Phase 1 Can Start

1. Max debit on roll? (OQ-1 in failure-modes.md)
2. Early assignment workflow identical to standard? (OQ-2)
3. "Uncovered" means no call or far-OTM allowed? (OQ-3)
4. Sentiment label staleness threshold? (OQ-4, proposed: 24h)
5. LIFO default per account type or globally? (OQ-5)

11. Dependencies

This model card is a research handoff document. It does not constitute investment advice. The strategy described is a paper-trading simulation. Real-world outcomes will differ.