- The Explorer
- Posts
- My AI Stock Experiment
In this edition:
Screenshots Galore
Scrape Anything Fast and Easily
AI Stock Picking
My ChatGPT/Claude Stock Experiment
I want to thank you for subscribing, and now reading, my newsletter! It would help me out if you would click on a link (any link will do). This tells Gmail and other programs that we are not spam. FYI - none of my links are affiliate links.
1. Screenshots Galore
If you’re like me, you are screen-shooting lots of content. The problem is how to find that screenshot a month later.
Trickle.so is the answer. Just send your screenshots to Trickle and it will extract the text, summarize it, and store the data. You can later ask Trickle a question and it will search all your screenshots to find an answer. Pretty cool!
The first 7 days are free then $10/mo. MyMind.com is similar and what I’ve been using. Includes a free plan.
2. Scape Anything Fast and Easily
Scraping is using software to extract data from a website or other data source. It can be a very effective and cheap way to get data but it can be a real pain to set up.
AI-scraper seeks to heal your pain.
Nothing to download or install because it runs in the cloud. Includes a built-in playground to test your scraping. Sets up as a simple API to scrape data to your own app. Easy peasy so give it a try.
3. AI Stock Picking
Who couldn’t use a little higher return on their investments?! I know I could (so my wife could quit one of her night jobs).
Each day danelfin ranks stocks with an “AI Score” with a 10 being the best. I have not followed this closely yet so if anyone has experience please let me (and the rest of us) know. I will do some paper trading stocks based on their recommendations and let you know how it goes.
My concern is that investing is a lot more than just picking a stock. There is a risk/reward analysis along with an exit plan and risk management that factor in heavily too. I’m open-minded though. They do offer a Free Forever Plan if interested.
4. My ChatGPT/Claude Stock Experiment
So I do a little investing and figured why not leverage some AI to make some extra cabbage. I asked ChatGPT if I uploaded some historical stock data, if it could help me find an investment strategy. So I downloaded 1 year of historical data for Tesla from Yahoo Finance and I uploaded it to ChatGPT.
I then told it that I look for a breakout above a resistance level to buy and typically set a stop 2-5% below the purchase price for swing trading (holding the stock 1-4 weeks) and added this…
Here’s the reply….
And it concluded this strategy would give me a 147% return on investment in comparison to the 6% loss for a buy-and-hold strategy. WOW!
So I started investing this way and am now rich!! Just kidding. I did verify that TSLA dropped 6% over that year so the buy-and-hold ROI was correct. But before investing I wanted to test the strategy.
I tried backtesting on TradingView using Pinescript code that chatgpt gave me. It didn’t work correctly.
I tried a backtesting platform and it didn’t work correctly there either.
So I started asking chatGPT more questions and noticed some discrepancies.
The more I asked the worse things got. In short, it kept trying to re-write the code to test the analysis but couldn’t get it to work correctly. In summary, the initial strategy it recommended was flawed and it could never identify why. And it could never fix the code to get a real ROI.
I then tried the same analysis on Claude and got essentially the same results. I noticed discrepancies and it couldn’t identify why or fix it.
In short, I used a full day to learn that these AI systems cannot yet reliably do basic spreadsheet analysis. Great for text, but not yet great for this. I did have the Wolfram (math) plugin active, which should allow it to do math, so I think it was an error in the scripts it kept trying to write.
I plan to experiment a little more with something a little simpler. Will keep you posted.
In the News…
AI Meme of the week…
Thanks so much for reading my newsletter. If you have any questions or suggestions, just reply to this email.
How was today's newsletter?Feedback helps me improve. |
Finally, if you know of anyone that might enjoy my newsletter, please forward this along using the link below. Thank you!
Your pal,
