Skip to content

Neyt/deep-value-hunter

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Deep Value Hunter

A two-stage Claude Code skill for finding genuinely overlooked pockets of value in public equity markets. Not a monolithic DD checklist — an actual discovery pipeline.

What it does

  • Stage 1 (Discovery): Narrows the ~10,000 US-listed ticker universe to ~20 finalists using quantitative filters (Piotroski F-score, Altman Z, Greenblatt Magic Formula, EV/EBIT, FCF yield, insider buying) and "overlooked pocket" heuristics (spinoffs, post-bankruptcy emergers, microcap neglect, 13D clusters).
  • Stage 2 (Deep Dive): Applies a Buffett / Munger / Fisher / Pabrai / Damodaran framework only to survivors, not to every stock the user has ever heard of.

Philosophy

Most "value investing" prompts in the wild are 2,000-word monolithic checklists that assume you already picked the ticker. That's backwards. Real deep value is 80% searching, 20% analyzing — you spend most of your time rejecting candidates, not writing essays about them.

This skill flips the ratio.

Why it's grounded

Built on the academic consensus of what actually works:

Finding Citation
~82% of 452 published anomalies fail the t > 3.0 bar Hou, Xue, Zhang (2020) RFS
Anomaly returns decay ~58% post-publication McLean & Pontiff (2016) JF
Use t > 3.0 not t > 2.0 after multiple-testing correction Harvey, Liu, Zhu (2016) RFS
Four survivor premia: Value, Momentum, Carry, Defensive AQR working papers (various)
Piotroski F-score adds ~7.5% annualized to a simple B/M screen Piotroski (2000) JAR
Greenblatt Magic Formula ~30% CAGR 1988-2004 backtest (degraded since) Greenblatt (2006)

These aren't the only things that work, but they're the things that have survived replication, cost modeling, and multiple-testing correction. Everything else is suspect.

Install as a Claude Code skill

# Clone to your skills directory
git clone https://github.com/Neyt/deep-value-hunter.git ~/.claude/skills/deep-value-hunter

# Or link from this repo
ln -s "$(pwd)" ~/.claude/skills/deep-value-hunter

Then in Claude Code, invoke with:

Find me overlooked value stocks in US microcaps, Pabrai style, 5 names

Usage

Three entry points:

  1. Full pipelinefind overlooked value in [universe]
  2. Stage 1 only (screen)screen for [Piotroski / spinoffs / net-nets / etc.]
  3. Stage 2 only (deep dive)deep dive on $TICKER

Output

  • Markdown report by default
  • .docx auto-generated via pandoc
  • Cites SEC EDGAR primary sources for every number
  • Explicit bull / base / bear scenarios with probability weights
  • Explicit rejection criteria (value trap red flags)

License

MIT. Do whatever you want with it. Nothing here is investment advice.

Not investment advice

This is a research tool. It does not know your tax situation, risk tolerance, time horizon, or liquidity needs. Every number it produces is a starting point for your own independent verification against primary sources. Past replication does not guarantee future returns. Anomaly returns decay. Value traps exist. You can and will lose money if you blindly trust an LLM with your portfolio.

About

Two-stage Claude Code skill for finding overlooked value stocks. Quantitative screens + overlooked-pocket heuristics, then Buffett/Munger deep dive. Grounded in academic evidence.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors