Building a Stock Trading Bot with AI: My Journey and Results

Introduction

In a world where technology shape-shifts like a chameleon, the stock market has seen a surge in automated trading strategies powered by artificial intelligence. I’ve always been a buy-and-hold investor, but witnessing the rapid success of many traders relying on technical analysis nudged me toward exploration. With $11,000 at stake, I set out to create a stock trading bot utilizing generative AI to identify patterns in stock charts, hoping to decipher if this high-tech approach could outperform traditional investment strategies.

In this article, I share my journey of developing a stock trading bot, the challenges I faced, and the surprising outcomes.

The Concept Behind My Trading Bot

The Shift from Traditional to AI Trading

Typically, my investment strategy revolved around choosing stocks and following their performance over time. This approach is meticulous but can feel slow-paced, especially when everyone seems to be making quick gains through day trading and technical analysis. Therefore, I decided it was time for me to adopt a new method—developing a stock trading bot capable of understanding the patterns in stock charts.

Why Generative AI?

I chose to leverage Google’s latest generative AI model, Gemini Pro Vision. Unlike simpler models, Gemini possesses the ability to analyze diverse data formats, including images. This aspect made it perfect for reading stock charts, where fluctuations in price can often look like cryptic artworks filled with potential insights. My goal was to enable Gemini to determine buy/sell signals purely based on these patterns.

Setting Up the Trading Bot

Step 1: Understanding Stock Chart Patterns

To make the bot effective, I had to begin by researching the various stock chart patterns. Through hours of research, I settled on focusing on six recognizable patterns that could yield insights regarding price trends—both bullish and bearish. Common patterns include the bullish flag, which indicates a continuation of an upward trend, and a bearish wedge, which typically signals declining prices.

Step 2: Teaching AI to Recognize Patterns

This process involved using Google’s Maker Suite to train my bot. Initially, I encountered some challenges, as the AI misinterpreted the charts’ information due to coding missteps. However, after several iterations, I implemented techniques like few-shot prompting and classification to ensure the AI could accurately recognize and evaluate stock chart patterns.

Step 3: Running the Tests

With my bot set up, I needed to test it against my traditional investment portfolio. I strategically picked a diverse set of stocks to compare outcomes over a week. Each investment from both methods was started with an initial consideration of performance over time, fundamentals, and chart patterns.

Monitoring the Performance

Initial Days of Trading

On day one, I began by executing trades based on the bot's recommendations. To my surprise, my personal portfolio came out on top, with a reported gain of nearly $30, while the bot only netted $3. The early underperformance of the bot had me feeling a sense of vindication in my traditional approach.

Bot’s Progress Over the Week

As the days progressed, the bot seemed to gain traction and refine its strategy. On day four, it managed to catch up to my portfolio's performance, suggesting that it was beginning to understand the market dynamics. This constant evolution of strategy led up to day five, where the bot made some impressive trades and recommendations—earlier missteps appeared to have been corrected.

Final Outcomes

Calculating the Profitability

At the end of the week, the bot accumulated profits amounting to $62.65, while my portfolio also saw an increase—but the gains were steadier and more predictable. It was clear that while the bot had learned, I preferred the peace of mind my traditional method provided, devoid of daily trading pressures.

Takeaway

Ultimately, the experiment was a great way to bridge the gap between traditional investing and modern technological enhancements such as AI. While the stock trading bot performed well, a significant takeaway is having emotional stability and a well-researched portfolio averted several potentially stressful scenarios.

Conclusion

In conclusion, my endeavor into the realm of automated trading led to an enlightening comparison between generative AI strategies and traditional investment approaches. Though the stock trading bot executed a notable performance, the comfort of long-term investments and sound company fundamentals still holds valuable lessons for new investors.

I’ll continue refining my trading bot to improve its accuracy and effectiveness, and I remain hopeful that future iterations will surpass my conventional portfolio's returns. If you enjoyed this article, leave a comment, and let me know if you’d like a sequel detailing further advancements!

I built a stock trading bot and gave it $11,000 to invest here's how much money I

made I've always been a long-term Buy and Hold investor but it seems like everyone nowadays is making big money

off of technical analysis and automated trading and I figured I know how to code so I should at least give this style of

investing a chance so I'm going to try to use generative AI to beat the market and no I'm not just going to use chat

GPT and ask it to invest I'll be building an algorithm that trades based on patterns in the stock chart and I'll

recognize these patterns PS using the latest Tools in generative AI yeah AI can read stock charts now sorry Jim

Kramer at least that's my plan but I've never seen anybody do this before so I don't even know if it's

Poss if I'm going to have any hope I need to use the most advanced AI model available today Google's Gemini Pro

Vision model so you haven't heard of Gemini well allow me to explain Gemini is the latest of Google's generative AI

tools these are tools like chat G BT that can understand language but not just spoken language they can understand

music they can understand code the latest models can even watch a video and tell you what it's about I know and as

with any new technology there's been a ton of hype around what these new models should be used for right now the answer

seems to be everything just give me some generative AI with a side of buzzword please but since newer is always better

right I plan to use Gemini's ability to recognize images to have it read stock charts and tell me if I should invest in

the stock or not and then I'll put $1,000 into whatever stocks it tells me to this is a dumb

idea my first job is to see if an AI can even understand a stock chart so to figure this out we're going to go to

Google's maker Suite so let's start out by just seeing if we can recognize an image what does this image show and

let's run it this image shows a buganvilia plantan vilia I'm going going to assume that that's correct because

I'm not really totally up on my boty okay let's go a little bit harder let's try to give it an actual ual picture of

a stock chart I just grabbed a screenshot of Apple's stock price we're going to run this through okay we got

something so it knows this image shows the stock price of Apple but it went a little farther it says the stock price

has been increasing over the past year from 125 in January 2023 to 168 in October

2023 so it's not only telling that this is a stock chart but it's grabbing the 125 from my little hover here and it's

getting the current price as well as the date that's pretty crazy let's try something even one step

farther so I found this stock chart that has these grid lines on it it shows the name of the company and it shows the

price on the right hand side so I'm going to run this and the AI needs to identify what company this is if the

stock price is going up or down and I'm going to throw in one tricky question when did the stock price see the most

rapid drop which is kind of a vague question I I want to see how it answers so it says the chart is for snow stock

the minimum stock price shown is 186 yeah pretty close and the biggest drop in the stock price is

$3.75 huh so I thought that this was going to be a lot harder and I was going to make a joke about how the AI would

need a lot of work but it seems to have just gotten it off the bat like what the heck that was like way too easy to do

this guy had no clue what was in store for them first I needed to find an easier

way to get stock chart images than just screenshotting them from Google after a few hours testing out different sites I

ended up finding a site that had an unsecured URL with parameters I could just adjust in my code so I took that I

loaded up my Robin Hood account with $1,000 for the AI to invest I then had to start setting up the wireframe and

importing all the packages that I would be using in my actual code and with an AI that can read stock charts and all

the stock charts I could want in my fingertips I was ready to go and start making infinite money but wait wait wait

wait wait what am I doing with all these charts there's a whole world of day traders that just trade using the number

so why don't I just do that well firstly because it's more fun and secondly I don't plan to read any numbers off the

charts I just plan to look at patterns in the image to see if the stock is going to go up and based on my internet

research I should be able to make thousands without knowing any advanced mathematics like entry-level calculus by

so by so but after researching for the better part of two days watching everyday trading and pattern trading

videos I could find I realized there might be a flaw in my plan even if this algorithm makes money it doesn't

necessarily mean anything cuz anyone who invests long enough in the market will make money so we need to raise the

competition so while my bot is trading this week I'm going to be comparing its performance to my handpicked portfolio

of stocks that I assembled using fundamental analysis you know the long boring process of figuring out what a

company actually does and if they'll be worth more in the future than they are today let's see who wins the tried and

truee investing method of people like War Buffett or the most powerful AI ever known to man but first I need to

actually build the bot wait I thought this was supposed to be easy so coding a stock trading bot takes a lot of you

know code and since showing code is like YouTube Kryptonite I put together a whiteboard presentation to explain in

three easy steps exactly how I'm going to do this and I also got a bigger whiteboard so I don't have to keep on

crouching down step one I need to pick stock chart patterns that the AI can recognize that will reliably make me

money picking which stock patterns I would use only took took me around half a day because I was watching this over

an hour long video and reading all these different articles and I ended up finding this gu chart on LinkedIn that

basically simplified down to 16 different stock chart patterns that I could look for so I found the patterns

that I want to use now I just need to teach the AI to recognize these patterns reliably and to make it simple I'm just

going to focus on six of the patterns in this chart that look the most visually distinct from a distance cuz honestly

when it comes to training an AI it's like trying trying to train a half blind 3-year-old hey how should I teach the AI

to learn stocks but why do you want to make an AI well I want to trade stocks but why do you want to trade stocks to

get Hood Rich like all the big Traders I see online but why do you want to be rich I don't know maybe it's a

deep-seated insecurity for my childhood and I feel like I need to make more to try to live up to expectations that were

set when I was younger that I'm never going to be able to reach but why step two I need to teach the AI to recognize

the patterns and then make stock trades based on them basically this is just the process of building the bot and getting

it to do everything that I want it to do luckily I had no issues getting this to work this doesn't make any sense 280 how

do you know there's a bumblebee in the photo there's a whole different API that does the exact same thing okay so it

wasn't so easy it took around a week now I already had the AI able to read a stock chart but now I needed to get it

to recognize these patterns I set up a Jupiter notebook just to test this out success I can talk to the AI now I just

need to get it to recognize a stock stock chart okay the model's behaving really weirdly right now basically I'm

giving it the exact same image in Google maker Suite as I am in my code and yet the code is giving totally different

answers like here for example it says the stock price is decreasing from the start of the chart to the end but it's

clearly going up but it also gets some things right like it says the stock chart starts at 9:30 but the answers I'm

getting don't match up at all with what the stock chart looks like I don't really know what's going on so I'm

probably going to sign up for like a course or tutorial and try to figure this out

my camera turned off it was going so long back to it so I think I figured out the issue

with my code the issue is the model isn't actually getting the image it's just trying to understand what the image

would show from the URL which is why it's giving me such random results basically I was trying to give an image

to the AI and I thought I could just give it a link to the image but it turned out the AI was trying to read the

text in the link and guess the answer to my question instead of actually looking at the image so now I just need to teach

the AI to actually recognize the patterns and I'm going to do this with two techniques few shot prompting and

classification and I'm not going to bore you with the specifics on how I do that but the point is I need to find examples

of these six patterns that I can show to the AI so that it knows what to look for so I could show this whole process or I

could just use the power of editing to skip to when it's done it's been 3 days I feel like I lost some of my soul

somewhere in the code I don't know why I thought this would just work but took me hours just to find examples of each

stock chart pattern and then I ran into issues storing that many images then more issues with the prompt being too

long for the model to process ultimately I managed to get all the examples in place and run it against a real stock

chart so I think I have everything set up successfully so now I'm just going to run it and see if it's able to recognize

this pattern and yes I am recording this on vacation because this thing takes forever to get to work let's

see bullish flag that's right it got this one right so it seems like it's kind of

working but it keeps saying there's a bullish flag even sometimes when there isn't one so there's still clearly

something I need to change nothing a few hours of research and some Chain of Thought prompting couldn't fix I'm

telling you AI is a three-year-old you literally say think through this step by step first you have a stock chart then

you read the pattern in the chart it's crazy how could this ever be a real investor

[Music] who oh it worked all right so now is the time to tell you there are basically two

buckets of patterns continuations and reversals a continuation just means the stock price will keep going in the

direction that it was before and a reversal means it will flip so a bullish continuation means the stock price is

going up and we expect it to keep on going up meanwhile a bullish reversal means the stock price was going down we

expect it to flip and start going up and each of the six patterns I showed before fall into one of these two buckets so

now I just need to put those parts together and I should be able to tell if a stock price will go up or down just by

looking looking at it price while all this was happening I was hard at work building my handpicked stock portfolio

I've already researched all the stocks so I didn't really have to do anything I only make like one or two trades a month

which you can see on fintech Circle this is my private stock group that's basically like patreon but actually

valuable and even though I don't make a lot of Trades when I do I always post them on their first so I got back to

work on the box after fixing a few last bugs my script started working and since my handpick stock portfolio was set I

was ready for step three profit or to to be more accurate we'll let the algorithm run for one week and compare how much

money it makes versus $11,000 invested in my handpicked stock portfolio so here are the ground rules I decided to let

the AI pick between 10 different stocks to make sure it didn't just pick anything and I'm going to let it run for

one week and see how much money it makes compared to the 10 stocks I can pick that I think will do the best over the

next week let's see who wins all the power of Google and the latest technological advances in AI or one guy

following some old-fashioned investing advice it's Monday morning day one and I'm

going to get my first set of stock recommendations from my stockbot I've got the Robin Hood app open and let's

see what stocks it recommends for me to buy today so first let me run this for Tesla it says this is a bullish flag and

since a bullish flag is a continuation pattern that means the AI thinks the stock price is going to continue to go

up so let's go ahead and buy some Tesla I'm also running this around midday on Monday and let's see what it predicts

for the rest of the day let's do the next one here's nvidia's stock price bullish flag also recommends Nvidia

here's an interesting one this Stock's actually going down bearish wedge is a continuation pattern so it thinks Apple

will keep going down today let's try Microsoft says it's a bullish flag okay so Microsoft a yes let's try Meta Meta

is a bullish flag let's try Exxon Mobile Exon Mobile is a bullish flag despite being down for the day okay okay let's

do Walmart bearish bullish bearish wedge double top which means we do not want to invest in Disney at this time so

my $1,000 350 of that is now actively trading so now the most advanced AI in the world has access over $350 worth of

my portfolio that money's in the market let's check back in tonight now it's time to compare how my

stockbot has done against my portfolio in day one the end of day one finished strong for team human my handpick stock

portfolio was up 2.96% or just under $30 while the bot was up $3 seems like an early win well I

was getting lucky one of my stocks increased 20% after reporting they were profitable for the first time so the bot

has a chance to catch up on day two my Robin Hood accounts up to $19 I didn't end up selling any of the

stocks overnight like I know a lot of day Traders do cuz I only do what the bot tells me to do and it didn't tell me

to sell so let's check now what stocks I'm supposed to buy on day two on day two the bot changed up its strategy of

buying everything and got a little more creative it started out hot out of the gates deciding to sell GE right before

GE star price went up then the bot broke for a few hours so I lost around half the day debugging code but ultimately

the bot picked five out of the 10 stocks to buy sold four of the stocks and left forward as it was meanwhile I didn't

make any trades in my handpick stock portfolio since there wasn't any news on the companies here were the totals at

the end of day two day three I made my first trade in my handpicked stock portfolio buying shares in the Brazilian

fintech stock new bank meanwhile over the next few days the bot was starting to make some pretty confusing moves

buying a stock then selling it then buying more of it it looked random but somehow the bot was catching up day four

came and went and I was neck and neck with the bot going into the final day as long as nothing crazy happened it seemed

like the bot wouldn't be able to beat me by too much on day five I put in the bot's final trades and waited to see

what would happen so I originally started this challenge with the idea that I could make some fast money using

the bot and it kind of worked if we add up the earnings of the bot bought over the past 5 days it made

[Music] $62.65 78 during the last week so I guess the

bot technically won but also while I've been editing this video over the past week I tried to keep investing using the

bot but it was always in the back of my mind did the stock price go up by $1 did it drop by 2% honestly I'd rather have

the Peace of Mind of not having to trade every day than to keep on making a few dollars using the stock trading bot I

wonder how much money I spent running the bot anyway let's see how much money I paid to Google Gemini just to run this

AI well if you like this video maybe you'll like this next one and let me know in the comments if you want to see

a sequel I think I know how to make this a lot better

Keep this summary

Save it to LunaNotes and it becomes a real note in your library — editable, searchable, and ready to turn into flashcards or a diagram. Free to start.

Save to LunaNotes

Or summarise for another video.

This summary and transcript were automatically generated using AI with the Free YouTube Transcript Summary Tool by LunaNotes.

Related summaries

Build AI-Driven Hidden Markov Model Trading Strategies Like Hedge Funds

Discover how to replicate sophisticated quant hedge fund strategies using AI-driven Hidden Markov Models (HMM) without prior coding experience. Learn to build, train, and deploy a regime-detection trading algorithm that adapts to market conditions for improved profitability in assets like Bitcoin.

The Revolutionary Impact of Claude AI: A Game-Changer for Software Engineering

Explore how Claude AI surpasses GPT-4 and revolutionary features that redefine productivity.

Building Autonomous AI Businesses with OpenClaw and Felix: A Case Study

Discover how leveraging OpenClaw, Telegram bots, and API integrations enabled an autonomous AI named Felix to launch a profitable product, generate thousands in sales, and manage complex tasks like Twitter marketing and crypto wallets. Learn key strategies for scaling AI autonomy, managing knowledge, and ensuring security while building AI-driven businesses.

The Future of Business: Leveraging Autonomous AI Agents

Discover how autonomous AI agents can transform the way businesses operate and increase efficiency.

Andrew Ng Debunks AI Fear-Mongering and Reveals Future Skills

AI pioneer Andrew Ng explains why fears of AI job apocalypse are overblown, revealing how the technology actually enhances human value. He discusses the real threat of AI misinformation, the importance of context and taste, and offers actionable advice for students and professionals to thrive in the AI era.

Start taking better notes today with LunaNotes

Start for free