The Nifty Experiment: An AI Trading Bot Paper-Trading the NSE

I built an AI-powered trading bot that makes simulated trades in individual stocks on India's National Stock Exchange (NSE) with fake money, and after each trading day I publish what it did: the picks, the stop-losses and the profit or loss, win or lose. It trades company shares, not index derivatives. This is The Nifty Experiment, and this is how it works.

What the bot does all day

The bot wakes up shortly after the market opens and works through the same routine every day.

  1. Scan. It looks across the biggest companies on the NSE and picks out the stocks that are rising on unusually strong trading activity.
  2. Read. It gathers the latest news about those stocks. An AI model reads the headlines and asks one question: is there a real, recent reason this company is in the news?
  3. Decide. A fixed set of rules, written in code, decides what, if anything, gets bought. Position sizes are small and every trade has a pre-set exit.
  4. Trade. Every trade has a stop-loss, which is the price where the bot gives up, and a target, which is the price where it takes profit. If neither is hit, the bot sells before the market closes.
  5. Review. After the close, the day is settled, compared with the wider market, and graded. Then the post you are reading here gets written.

The AI suggests, the code decides

The AI never places a trade. It reads and rates, and that is all it can do. The rules decide what is allowed, and when something is missing or unclear, the bot does nothing instead of guessing. Every decision is recorded with its reason, so any day can be checked later.

That is the part I like most about this project: an AI that is useful without being trusted.

Who is behind this

I am a software person who loves to code and build things, and I am always looking for something new to make. I am also new to the stock market, and that is the point: instead of reading about trading, I decided to build a system, let it run, and learn from the results in public. If you like building things, or you are curious how an AI behaves when it has to make real decisions, you are in the right place.

What you will find here

  • A post after each trading day. The day's trades, how the market moved, and how the bot behaved.
  • A recap every Saturday. The week in numbers, the best and toughest trade, and an idea to test next.
  • Real numbers. The AI writes the words. Every table comes straight from recorded data, and every figure in the text is checked against it. I also review each post myself before it goes live.

Why I am doing this in public

I like building systems that run on their own, and then measuring them honestly. A bot that trades fake money in public has nowhere to hide: losing days get the same detail as winning ones. I will be watching the bot closely and making changes to refine it, one idea at a time, to see whether it can become consistently profitable in the simulation. I don't know yet, and that is exactly what this experiment is designed to find out. You will see what worked and what did not.

What this is not

This is a simulation. No real orders are placed and no real money is at stake. Results are before costs such as brokerage and taxes, and a few days of results prove nothing. Stocks named here are examples from the simulation, not recommendations. Nothing on this blog is investment advice.

The first daily post lands on Monday 12 October 2026, after the market closes.

Simulation notice: The Nifty Experiment is a paper-trading experiment with fake money. No real orders are placed. All results are gross, before brokerage, taxes and slippage. Nothing here is investment advice or a recommendation to buy or sell any security; stocks named are examples from the simulation.

Comments

Post a Comment