Can AI do better trading than humans?With Ai every industry is ripe for a foundational shift, where does the stock market stand?
I have been trading for more than a decade now. Initially I made nothing but losses. I tried everything: options, candlestick patterns, technicals, fundamentals, theta, gamma and all. It took me years to find that trading requires discipline more than anything. Even now, I make losses here and there.
Now, keeping all these aside, can AI be a better trader than me? Most probably. So how do we make this happen?
The precursor to AI trading is algo trading. But trading algos are mostly static in the sense that we define an entry point and exit point for a stock along with parameters that define where and when.
Take for example a pattern like the Golden Cross where the 50-day moving average crosses the 200-day MA. If we put this into an algo, it would enter the trade every single time. It wouldn’t care about the news around the stock or major economic news around the corner (like a Fed meeting) and institutional data (hedge funds and shoe sellers). You can keep adding stuff to your algo, but you can only take it so far.
No amount of backtesting or algorithmic analysis would prepare you for a news-based event like April 2025’s tariff-related plunge.
Can AI do that? It should, right? May be?!
So a team dedicated to testing this , built an app that tracks a $100k investment. They gave the amount to major LLMs (OpenAI, Claude, DeepSeek, etc.) and let it trade based on available info. Each of them took a plunge during April ($100k to $80k and then to $120k by November).
What this teaches us is that AI-based trading is not there yet. It has to account for:
- News-based events
- Statistical data (historical)
- Multi-nodal forecasting
- Option data
- Economic datas
And much more. While a text-based LLM is absolutely capable of taking in all this data and generate profitable trades, it currently doesn’t have the shrewdness to nail a trade regardless of any condition. It should be as ruthless as a hedge fund and be a short seller based on publicly available inside information (oh please stop , if Congress can trade in a stock that they regulate, so should we).
There is a saying in Dalal Street (Indian Wall Street) that “even a dog can make a profit during a bull run. But only a real trader makes profit in a bear market.”
AI will eventually trade better than all of us combined, but that doesn’t mean we can’t use it today to do what we want it to do, even if it comes with limitations.
The future of trading will be an algo that uses multi-nodal LLMs / Frontier model trained solely based on financial data (Mistral-type smaller ones). And this is where I intend to start my journey with paper trading (with no real money of course) - a true frontier stock ai trader ! Isn’t it fun?!
Note 1: This is not for investing where the horizon is in years, just for trading (intra day / t+2 delivery based trading)
Note 2: I am not an SEC-approved broker so this is just my thoughts and not advice.
Note 3: Invest — don’t trade 😜
