| name | narrative-reversal-screen |
| description | Screens for "narrative reversal" candidates — stocks down 30%+ from 52W high with concrete catalyst still intact, worst-case priced in, early reversal signal (first higher low, 50DMA cross, insider buying after capitulation). Returns top 3 with entry plan. Triggers in English ("beaten-down stocks with thesis", "find reversal plays", "stocks at bottom that can recover", "fallen angel screen", "comeback candidates") or Chinese ("找暴跌反转股", "找回归类股票", "ORCL 那种反转", "已经跌透的好股", "底部反弹候选"). |
Narrative Reversal Screen — Buy The Beaten Down
🔍 Pre-flight checklist — reversals fail spectacularly in the wrong regime
A "down 50% with intact catalyst" looks like value until the broader market sells off another 20% and your "value" becomes -70%. Required checks:
- Macro regime gate — trigger
macro-warning. Reversals work in 🟢 GREEN (recovering market lifts beaten-down names) and mid-late 🟡 YELLOW (capitulation phase). 🔴 RED → reversals are still falling knives. Don't recommend reversal buys in RED; flag as "watch list only" instead.
- Insider must be slowing sells or starting to buy —
$(ls ~/.claude/{skills,plugins/claude-investment-skills}/review-investment-screenshot/scripts/insider_ratio.py 2>/dev/null | head -1) TICKER --window 180 (180d for reversals — see the trend over 2 quarters, not just 90d). The ORCL signal was: pre-crash 8 sells / 2 buys → post-crash 2 sells / 5 buys. Form 4 code "P" only.
- First higher low CONFIRMED, not predicted — Don't pre-empt. Wait for: (a) 50DMA crossover OR (b) confirmed higher low on weekly. Pre-empting "looks bottomed" is how -30% becomes -50%.
- 3-tier entry — T1 = first confirmed higher low; T2 = retest of capitulation low; T3 = full revisit (if it happens). Never "buy at market on a chart that looks bottomed."
- Sizing 2-3% per pick — reversals are higher-risk than momentum names; don't size like core holdings.
- Tax framing — reversal entries are positions, not swings; assume 12+ month hold. Frame return targets after-tax: +50% gross = +40-43% after LTCG.
"Look carefully" rule: most "beaten-down with intact catalyst" stocks are beaten down for a REASON the bull case denies. Always answer: "what would have to be true for the bear thesis to be correct?" If you can't articulate it, you don't understand the trade. Verified false reversals from 2026: WBA -60% (mgmt still selling), DLTR -45% (insiders also exited), KSS -50% (PE walked away).
See README's Hard Rules for the full anti-pattern list.
Goal
Find stocks where:
- Worst-case is already priced in (down 30%+ from ATH)
- Concrete catalyst still intact (real contracts, not just hope)
- Reversal early signal (50DMA up-cross or first higher low)
- Insider not capitulating (sells slowing or buying starting)
- Forward valuation cheap (P/E < pre-crash 50%)
The ORCL pattern that delivered $172 → $185 reversal: down 50% from $345 + OpenAI $300B intact + insider selling slowing + 50DMA crossover.
Why this matters
Most "value" looks dead. But true narrative reversals (where the bear story is wrong) outperform:
- ORCL 2025/9 → 2026/5: -50% then +35% rebound
- META 2022 → 2023: -76% then +200%
- NFLX 2022 → 2023: -75% then +180%
- CRM 2022 → 2023: -55% then +90%
These all share: panic priced in maximum darkness then truth comes out.
The 5-Step Workflow
Step 1 — Define hunting ground
Common buckets where reversals happen:
- AI Cloud (ORCL after OpenAI doubts)
- Big Tech (META 2022 metaverse fear)
- Beaten retailers (NFLX 2022 streaming saturation fear)
- Software (CRM, ZM, DOCU after 2022 reset)
- Healthcare/Biotech after FDA rejection
- China after regulatory crackdown
- Auto after EV demand reset
- Banks after credit fear
Step 2 — Apply hard filters
Pull via mcp__yfmcp__yfinance_get_ticker_info for candidate list:
Hard cuts (must pass ALL):
| Filter | Threshold | Why |
|---|
| Distance from ATH | -30% to -65% | Need real darkness, but not complete failure |
| Forward P/E | <30 (cheap for sector) | Avoid "down but still expensive" |
| Market cap | >$10B | Liquidity for institutional re-entry |
| Revenue growth (latest) | >0% (positive) | Don't catch a falling knife |
| Cash flow | Positive OR within 1 quarter | Survives long enough for reversal |
Optional but strong:
- Recent earnings beat (sentiment shift)
- New CEO or strategy reset
- Activist investor involvement
- Big contract / partnership announcement
Step 3 — Look for reversal signals
For each survivor, check:
| Signal | What to look for |
|---|
| 50DMA cross | Stock has crossed above 50DMA (or testing) |
| First higher low | Recent low > previous low (bottoming) |
| Volume divergence | Down moves on lower volume vs up moves |
| RSI bullish divergence | Lower price, higher RSI |
| Insider buying | First insider buys in 6+ months (BIG signal) |
| Short interest declining | Shorts covering |
Step 4 — Verify catalyst is intact (CRITICAL)
The hardest part. Use WebSearch and earnings transcripts to verify:
Questions to ask:
- Is the original thesis still true? (e.g., ORCL OpenAI $300B contract still active)
- What's been added since the crash? (new contracts, customers)
- Is management same/changed? (new CEO often = catalyst)
- Have analysts started raising targets? (early signs of recognition)
- Are insider buying? (ultimate signal)
Reject if catalyst is broken:
- Major customer lost
- Management change without coherent strategy
- Regulatory headwinds increased
- Competitive moat eroded
Step 5 — Check insider for capitulation/buying
Run insider_ratio.py "TICKER" --window 90 (or wider window covering capitulation date). Uses openinsider primary; only Form 4 code "P" counts as buy.
- 🟢🟢🟢 RECENT CLUSTER BUY (3+ insiders, $500K+ each): Bottom signal — most reversal candidates have this. Real 2026 example: NKE (CEO Hill + Tim Cook + 2 Directors, $3.7M, 7 days, after -46% drop).
- 🟢 STRONG BUY: 1-2 senior insiders buying after period of selling
- 🟡 OK: Selling slowed / stopped + sells are 10b5-1 trusts (verify at secform4.com)
- 🔴 AVOID: CEO/CFO ad-hoc selling (not 10b5-1) = haven't capitulated
Reminder: News articles' "cluster buy" claims are often false positives (RSU/DSU grants). Always verify Form 4 code = "P" at openinsider.com/[TICKER].
Especially good: CEO/CFO/Director cluster buy after stock down 50% (CEVA 2026 / NKE 2026 / AMKR Kim family / LSCC new CEO patterns).
Output format
# Narrative Reversal Screen — [Date]
## TL;DR
Found [N] candidates with intact catalyst + reversal signal. Top 3:
1. [TICKER] — [tagline]
2. [TICKER] — [tagline]
3. [TICKER] — [tagline]
## Screen Results
### Top 3 Deep Dive
#### #1 [TICKER]
**The Crash**: ATH $X (Date) → Low $Y (Date) = -Z%
**Why it crashed**: [1 sentence]
**The Recovery**: Current $A, +B% from low
**Catalyst Status**:
- Original thesis: [check if still true]
- Concrete evidence: [contracts, customers]
- New positives since crash: [list]
**Reversal signals**:
- 50DMA: [above/below + how recently crossed]
- First higher low: [yes/no, when]
- Insider: [buy ratio + key buyers]
- Analyst: [recent upgrades/raises]
**Valuation**:
- Pre-crash P/E: 35x
- Current P/E: 21x
- 50% multiple compression already
- Peer P/E: 28x → upside in re-rating
**Entry Plan**:
- Now: $A (try 30%)
- Pullback to 50DMA $B: add 30%
- Pullback to 200DMA $C: add 40%
**12-month target**: $D (+E%)
**Risk**: [main risk that would invalidate]
## Watch list (not yet entry signal but close)
| Ticker | Reason waiting |
## Rejected
| Ticker | Why rejected |
| TICKER | Catalyst actually broken |
| TICKER | Insider still capitulating |
| TICKER | Down 30% but PE still 60x |
Hard rules
- Verify catalyst is INTACT. Don't catch falling knives. ORCL had OpenAI $300B verifiable; some falling stocks don't have anything.
- Don't buy at exactly the bottom. Wait for first higher low + 50DMA cross.
- Insider matters MORE here. In growth stocks insider is noise; in reversals it's THE signal (CEVA pattern).
- Re-rating > earnings growth. ORCL went from PE 35 → 21 = -40%. Going back to 28-30 = +35% just on multiple, before earnings.
- Cap position size at 5-7% (still risky, R/R asymmetric).
- Set stop at recent low. If thesis breaks, exit fast.
Common pitfalls
| Pitfall | Example | Lesson |
|---|
| Catching falling knife | Buying ORCL at $200 (still falling) | Wait for higher low |
| Anchoring to ATH | "ORCL was $345, must go back" | ATH is meaningless; intrinsic value matters |
| Ignoring why it crashed | "Stock down 50%, must be good" | Some crashes are correct (BBBY, WW) |
| Buying first reversal signal | Bought 50DMA cross, then -10% retest | Wait for confirmation (volume + 2nd higher low) |
| Selling too early on bounce | Bought $172, sold $185 (+8%) | Multi-month moves can return 30-50% |
Templates: Common reversal patterns
Pattern A: The "Worst-Case Priced In"
- ORCL 2026/5 (China=0 priced in for NVDA, OpenAI doubt for ORCL)
- META 2022 (metaverse + ad recession fears)
- Sign: forward P/E at multi-year low, but earnings stable
Pattern B: The "Activist or New CEO"
- INTC under Lip-Bu Tan (2026 turnaround attempt)
- DIS under Iger return (2022)
- Sign: new strategy = new buying
Pattern C: The "Sector Rotation Reversal"
- Energy 2020 → 2022 (oil $30 → $130)
- Bank 2023 (post-SVB collapse, then recovery)
- Sign: macro tailwind reasserts
Pattern D: The "Earnings Inflection"
- AMD 2023 (data center inflection)
- NVDA 2023 Q1 ER (AI catalyst surprise)
- Sign: first surprise beat after string of misses
Watch list management
Status categories:
- 🟢 BUY now — all 5 criteria pass
- 🟡 WATCH — 4 of 5 pass, waiting for trigger
- 🟠 STILL FALLING — early, wait for higher low
- 🔴 REJECTED — catalyst broken or value trap
Update cadence: Weekly review of WATCH list, daily for current entries.
When to invoke
- User asks: "Find beaten-down stocks with real story"
- User asks: "ORCL-style plays"
- User asks: "Reversal candidates"
- After SPX -10%+ correction (lots of opportunities open up)
- Quarterly: routine reversal hunting
Tool cheat-sheet
| Need | Tool |
|---|
| Distance from ATH | mcp__yfmcp__yfinance_get_ticker_info → 52WeekHigh, current price |
| Pre-crash valuation | WebSearch: "[ticker] historical P/E 2024 2025" |
| Catalyst verification | WebSearch: "[ticker] [catalyst keyword] update [current month]" |
| Insider | insider_ratio.py "TICKER" --since [crash date] |
| 50DMA cross | mcp__yfmcp__yfinance_get_price_history (3-6mo) |
Companion skills
After this screens 3 candidates:
- Run
analyze-stock for deep dive
- Run
option-wall-analysis for short-term levels
- Run
leaps-screen for LEAPS (LEAPS work GREAT for reversals — high IV after crash)
- Run
tax-optimize if user wants to use as alternative to current holdings
