LLM-Trading-Lab - GitLite

KO | EN

gitlite — search

#javascript #python #react #ai #typescript #rust #linux #java #nodejs #cli #pytorch #mcp #ios #open-source

LLM-Trading-Lab

(★ 7,497)

This repo powers my experiment where ChatGPT manages a real-money micro-cap stock portfolio.

파일 탐색기

최종 버전 다운로드 (.zip)

- - __init__.py
  - workflow.py
- AUTOMATION_README.md
- CODE_OF_CONDUCT.md
- CONTRIBUTING.md
- ignore_list.gitignore
- License.txt
.gitignore
Makefile
README.md
requirements.txt

# 설치 가이드

1. 코드 내려받기

git clone https://github.com/LuckyOne7777/LLM-Trading-Lab

깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.

cd LLM-Trading-Lab

방금 내려받은 프로젝트 폴더 안으로 이동합니다.

2. Python

쉬움 추천

사전 준비물

Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
Python 3 설치 시 'Add Python to PATH' 옵션을 꼭 체크하세요 (Windows).

pip install -r requirements.txt

requirements.txt 등에 명시된 파이썬 라이브러리를 설치합니다.

python <실행할 파일명>.py # README에서 정확한 실행 파일명을 확인하세요

파이썬 스크립트(또는 모듈)를 실행합니다.

✅ 에러 메시지 없이 실행되고 터미널에 안내 문구가 출력되면 정상입니다.

3. Make

보통

사전 준비물

Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
Make Linux/macOS는 보통 기본 설치되어 있습니다. Windows는 별도 설치(예: MSYS2, WSL)가 필요합니다.

make

생성된 빌드 설정을 바탕으로 실제 컴파일을 진행해 실행 파일을 만듭니다.

✅ 에러 없이 끝나면 성공입니다. 생성된 실행 파일을 직접 실행해보세요.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# LLM Trading Lab
This repository started as a **6-month live micro-cap trading experiment** in which a large language model (ChatGPT) manages a real-money portfolio under strict, predefined rules.

What began as a single experiment has evolved into a **baseline framework** for studying how large language models behave as portfolio decision-makers.  
All historical data, research artifacts, and logs are preserved for transparency and auditability.

**Full research evaluation out now: [Evaluating ChatGPT as a Portfolio Decision-Maker in Micro-Cap Equities](Experiments/chatgpt_micro-cap/evaluation/paper.pdf)**

---

## Running Your Own Experiment

If you want to run your own AI-managed trading experiment, check out this framework I created for LLM research: [LLM Investor Behavior Benchmark - LIBB](https://github.com/LuckyOne7777/LLM-Investor-Behavior-Benchmark)

## Repository Purpose

This repository serves two primary purposes:

1. A **complete, forward-only record** of a live AI-managed trading experiment  
2. A **reusable foundation** for future AI-driven trading experiments built on the same structure

Historical artifacts remain unchanged. New experiments, analyses, and methodologies are layered on top without rewriting past results.

---

```text
ChatGPT-Micro-Cap-Experiment/
│
├─ README.md
├─ requirements.txt
├─ Makefile
│
├─ experiments/
│  └─ chatgpt_micro_cap/
│     │
│     ├─ trading_script.py
│     │
|     ├─ graphing/
|     │  ├─ daily_returns.py
|     │  ├─ drawdown.py
|     │  └─ ...
|     │ 
│     ├─ csv_files/
│     │  ├─ Daily_Updates.csv
│     │  └─ Trade_Log.csv
│     │
│     ├─ evaluation/
│     │  ├─ evaluation_report.md
│     │  └─ paper.pdf
│     │
│     ├─ collected_artifacts/
│     │  ├─ deep_research_index.md
│     │  ├─ chats.md
│     │  │
│     │  ├─ Weekly_Deep_Research_MD/
│     │  │  ├─ Week_01_Summary.md
│     │  │  ├─ Week_02_Summary.md
│     │  │  └─ ...
│     │  │
│     │  └─ Weekly_Deep_Research_PDF/
│     │     ├─ Starting_Research.pdf
│     │     ├─ Week_01.pdf
│     │     ├─ Week_02.pdf
│     │     └─ ...
│     │
│     ├─ images/
│     │  ├─ equity_vs_baseline.png
│     │  ├─ repeated_exposure.png
│     │  └─ ...
│     │
│     ├─ tables/
│     │  └─ metrics.txt
│     │
│     ├─ metrics/
│     │  ├─ load_dataV3.py
│     │  └─ episode_pcr.py
│     │
│     └─ processing/
│        ├─ ProcessPortfolio.py
|
│
├─

```
---

## The Concept

Every day, I kept seeing the same ad about having some A.I. pick undervalued stocks. It was obvious it was trying to get me to subscribe to some garbage, so I just rolled my eyes. 
Then I started wondering, "How well would that actually work?"

So, starting with just $100, I wanted to answer a simple but powerful question: **Can powerful large language models like ChatGPT actually generate alpha (or at least make smart trading decisions) using real-time data?**

Today, this repo has evolved into so much more than simply chasing alpha.

---

## Why This Matters

AI is being aggressively marketed as a replacement for human decision-making across industries.  
Trading is a domain where mistakes are measurable, irreversible, and costly.

This platform tests those claims using:

- Forward-only decisions
- Full transparency  
- Publicly logged results

---

## Research & Documentation

Here are the artifacts links for the Micro-Cap Experiment:

- **Research Index:** [Deep Research Index](Experiments/chatgpt_micro-cap/collected_artifacts/deep_research_index.md)

- **Decision Logs / Chats:** [Chats](Experiments/chatgpt_micro-cap/collected_artifacts/chats.md)

---

## Features of This Repository

- 40 page PDF evaluation over results
- Live trading engine used in production  
- LLM-driven trade selection under hard constraints  
- Daily CSV-based portfolio accounting  
- Automated stop-loss enforcement  
- Benchmark comparisons (S&P 500, Russell 2000)  
- CAPM, Sharpe, Sortino, and drawdown analytics  
- Full trade and decision logs

---

## Tech Stack

- Python 3.11+  
- pandas  
- yfinance (primary data source)  
- Stooq (fallback data source)  
- Matplotlib

---

## Future Work

I am currently designing the future experiment over newly listed IPOs with monthly analysis on my [Substack](https://nathanbsmith729.substack.com/).

Also, I developing the general experimental framework I created for LLM research [LIBB](https://github.com/LuckyOne7777/LLM-Investor-Behavior-Benchmark) for the upcoming and all future experiments.

---

## Contributing

Contributions are welcome.

- Issues: bugs, edge cases, or design critiques  
- Pull Requests: improvements, refactors, or extensions  
- Collaboration: high-quality contributors may be invited to help maintain future experiments

Contributing guide:  
https://github.com/LuckyOne7777/ChatGPT-Micro-Cap-Experiment/blob/main/Other/CONTRIBUTING.md

---

## Contact

All my links can be found on my profile, feel free to reach out anywhere!