Earnings Prep

Earnings Prep — Pre-Earnings Decision Framework

🔍 Pre-flight checklist — run BEFORE building scenarios or making recommendations

Earnings reactions are ±8-20%. That's not where to skip the discipline layers.

1. Macro regime — trigger macro-warning skill first. In 🔴 RED regime, even a 20% beat can sell off (AMD 2/3/2026 beat 23%, dropped -17.31%). RED regime → recommend trim/hedge only, no adds regardless of expected reaction.
2. Insider 60d window — $(ls ~/.claude/{skills,plugins/claude-investment-skills}/review-investment-screenshot/scripts/insider_ratio.py 2>/dev/null | head -1) TICKER --window 60 (60d catches pre-earnings ad-hoc activity that 90d dilutes). Pre-earnings C-suite distribution = they know something. Never trust yfinance "% Net Shares" headline (RSU pollutes it). Form 4 code "P" only counts as buy.
3. Position context first — if user already holds, lead with their position size + cost basis + holding period. Generic "buy/sell" without position context is useless.
4. 3-tier entry if recommending add — Don't say "buy after earnings". Map T1 = post-earnings reaction level, T2 = 50DMA support, T3 = 200DMA / pre-earnings level. NEVER "buy at market" on earnings hype.
5. Tax for trims — If trimming a position held < 12 months, run tax-optimize for STCG vs LTCG impact. STCG ~25-37% federal vs LTCG 15-20%. Often changes the recommendation (hold 30 more days → save 10%).
6. Sizing — ≤ 10% single name; if recommending add post-earnings, current allocation must allow it.

"Look carefully" rule: the most common pre-earnings mistake is treating "beat consensus" as automatic upside. Always answer "what's already priced in" (Step 4 below) — if Forward P/E expanded 30% in 3 months, the beat is already discounted.

See README's Hard Rules for the full anti-pattern list.

Goal

Answer 3 questions:

1. What's the market expecting? (consensus + implied move)
2. What's the historical pattern? (last 8 quarters reactions)
3. What should I do with my position? (hold / trim / hedge / add)

The 5-Step Workflow

Step 1 — Pull consensus + guidance

Tools:

• WebSearch: "[ticker] Q[X] FY[YEAR] earnings consensus revenue EPS"
• mcp__yfmcp__yfinance_get_ticker_info for current price + earnings date

Capture:

Step 2 — Calculate implied move

Method 1 (preferred): ATM straddle

• Pull option chain for first expiry after earnings
• Find ATM call + ATM put
• Implied move % = (call mid + put mid) / current price × 100
• Convert to $ range: current ± implied move

Method 2: WebSearch fallback

• WebSearch: "[ticker] earnings expected move implied volatility"

Output:

Current: $XXX
Implied move: ±X.X%
Expected range: $YYY - $ZZZ

Step 3 — Historical 8-quarter reactions

Pull last 8 quarter reactions. Use optionsai.com, optionslam.com, or WebSearch.

Output table:

Key statistics:

• Average absolute move: X.X%
• Win rate: Y/8 (Z%)
• Largest up: +X% / Largest down: -X%
• Recent trend: Last 3-4 quarters skew (often more important than 8q average)

Step 4 — Predictability assessment (NEW)

Before scenario building, assess earnings predictability based on growth model:

For each model, watch for these red flags pre-earnings:

• Capacity-bottlenecked: "缺料" / component shortage rumors → likely miss on operational metrics even if EPS beats (FN pattern 5/4/2026)
• Demand-elastic: "guide in-line" without raise → "beat but priced in" (AMD pattern)
• Cyclical: Inventory buildup at customers → next-quarter weak guide
• Recurring: Customer churn news → revenue miss

Step 5 — What's priced in vs not?

Critical analysis — this is what separates pros from amateurs.

Question 1: What's already in the price?

• Current Forward P/E vs historical avg
• Distance from 50DMA / 200DMA
• 1Y price change
• If +50% in 3 months → expectations极高

Question 2: What's NOT in the price?

• Specific business optionalities
• Geographic recovery (e.g., NVDA China = $0 priced in)
• New product cycles
• M&A rumors

Examples:

• NVDA 5/2026: China = $0 priced in → any China good news = pure upside
• ORCL 6/2026: Capex concerns priced in → any FCF improvement = upside
• AMD 5/2026: MI400 + OpenAI已 priced in → 需要"上修指引"才能涨

Step 5 — Build 4 scenarios with weighting

Weighted average price = Σ(probability × target) If weighted < current → 期望值是负的，市场已 priced in 利好

Position-Specific Recommendations

Based on the 4 scenarios, give SPECIFIC action:

Already long the stock?

Want to make a directional bet?

Want to hedge existing long position?

Output format

# [TICKER] Earnings Prep — [Date]
**Earnings: [Date BMO/AMC] | Days away: N**

## TL;DR
[One paragraph: what to do with current position + size]

## Consensus
| Revenue | EPS | YoY |

## Implied Move
- ATM straddle: ±X.X%
- Range: $YYY - $ZZZ

## Historical Pattern (Last 8 Quarters)
[Table]
- Avg abs move: X.X%
- Win rate: Y/8

## What's Priced In
- ✅ Already in price: [list]
- ❌ NOT priced in: [list — these are upside catalysts]

## 4 Scenarios
| Scenario | Prob | Reaction | Target |
| Mega-bull | X% | +X% | $Y |
| Bull | X% | +X% | $Y |
| Flat | X% | -X% | $Y |
| Bear | X% | -X% | $Y |
**Weighted avg target: $X (vs current $Y) → +/-Z%**

## My Recommendation
- **If I'm long [N shares]**: [trim X / hold / add Y]
- **If I'm not in**: [skip / play call spread / wait]
- **Specific orders**: [exact ticker, qty, price, type]

## Watch list (next 24h)
- [Pre-market reaction signals]
- [Key levels: $X = breakout, $Y = breakdown]

Hard rules

1. Pre-flight before recommending. Skip the checklist at top = incomplete call. Macro + insider 60d + position context are not optional.
2. Never recommend "hold through earnings" without setting a stop. Implied move is real money.
3. Compare last 3-4 quarters more heavily than 8q avg. Pattern recently changed = more relevant.
4. Big beats sometimes lead to big drops. AMD 2/3/2026: EPS beat 23% → -17.31%. Always check "is this priced in".
5. In 🔴 RED macro regime, no new longs regardless of earnings setup — trim/hedge only. Document the regime call in the output.
6. IV crush is real. Buy options 5+ days out, sell into earnings hype.
7. Never sell naked puts on stocks you wouldn't own. Sell put = "I want to buy at $X". If $X isn't a level you'd actually buy at, don't sell.
8. Cite implied move source. Either ATM straddle calculation or WebSearch with link.
9. When user is already long: lead with their position, not generic analysis.
10. 3-tier entry for any add recommendation. "Buy at market post-earnings" = wrong answer. Always map T1/T2/T3.
11. Tax-aware trims. If position held < 12 months, surface STCG impact in the recommendation, not as footnote.

Common pitfalls

Tool cheat-sheet

When to invoke

• User asks: "Should I hold X through earnings?"
• User asks: "X reports tomorrow, what do I do?"
• User asks: "Should I sell put before earnings?"
• Auto: 7 days before any earnings of a stock the user holds
• Auto: Day-of for any stock with options position expiring within 1 month of earnings