I coded a stock trading Ai. Here's how much money I made.

I Coded a Stock Trading AI. Here's How Much Money I Made.

Introduction

I've always been a long-term buy-and-hold investor, but recently it seems like everyone is making big money off of technical analysis and automated trading. As someone who knows how to code, I decided to leverage generative AI to beat the market, going beyond just asking ChatGPT for investment advice.

This article narrates my journey in building an AI-powered stock trading algorithm using Google's Gemini Pro Vision model, its trials and tribulations, and eventually compares its performance against my handpicked stock portfolio.

Setting the Stage

Initially, I wanted to find out if an AI could even understand stock charts. To do this, I started with Google’s Maker Suite. I tested it with simple images and stock charts. Surprisingly, it recognized the patterns and could identify stock prices and trends from the screenshots.

Building the Stock Trading Bot

With an AI capable of understanding stock charts, I moved on to build the bot:

1. Pick Stock Chart Patterns: I focused on six distinct patterns from a chart I found on LinkedIn. The objective was to make the AI recognize these patterns reliably.
2. Teach the AI to Recognize Patterns: I set up a Jupyter notebook, faced several coding challenges, and finally managed to get the AI to recognize patterns like bullish flags and bearish wedges.
3. Incorporate Trading Strategy: Using continuation and reversal patterns, I developed a script that let the bot predict whether a stock price would go up or down and make trading decisions accordingly.

The Showdown: AI Bot vs. Human Portfolio

To test the effectiveness of my AI, I pitted it against my handpicked stock portfolio over one week. I allocated $ 1,000 for the AI to invest and chose 10 stocks that I thought would perform best.

Day 1

The bot started by recommending stocks like Tesla, Nvidia, and Microsoft. By the end of Day 1, my handpicked portfolio was up 2.96%, while the bot was only up $ 3.

Day 2

The bot showed more creativity in its strategy, buying and selling stocks, but it still had mixed results. Meanwhile, I made no changes to my handpicked portfolio, waiting for concrete news.

Days 3-5

As the week progressed, the bot made several confusing moves, buying and selling stocks frequently. However, it slowly caught up to my handpicked portfolio’s performance. By the end of Day 5, the bot had technically won, making $ 62.65 compared to my $ 78.

Conclusion

Though the AI made some money, the peace of mind from not having to trade every day made me favor the traditional investing approach. While the bot showed potential, it still required considerable oversight and tweaks. It was a fascinating journey, and I learned a lot about the capabilities and limitations of modern AI in stock trading.

[Final thoughts on improving the bot and learning costs]

Keywords

• Stock Trading AI
• Generative AI
• Google Gemini Pro Vision
• Technical Analysis
• Continuation Patterns
• Reversal Patterns
• Robin Hood
• Handpicked Portfolio

FAQ

Q1: What is the main goal of building the AI stock trading bot? A1: The main goal was to leverage generative AI to read stock charts and make profitable trading decisions, aiming to outperform traditional long-term investing methods.

Q2: Which AI model did you use for this experiment? A2: I used Google's Gemini Pro Vision model for this experiment.

Q3: How did the AI perform compared to the handpicked portfolio? A3: Over one week, the AI made $ 62.65, while my handpicked portfolio made $ 78, showing that the bot was competitive but fell short of the human approach.

Q4: What were the main challenges faced while building the bot? A4: The main challenges included getting the AI to reliably recognize stock chart patterns, debugging code issues, and ensuring the AI used the actual images instead of guessing from URLs.

Q5: Will you continue using the AI bot for trading? A5: While the AI bot showed potential, I prefer the peace of mind that comes with traditional investing and do not plan to rely on the bot for daily trades.