What is algo trading? A plain explanation for Indian traders

Algo trading means writing your trading rules down precisely enough that a computer can follow them without you.

That is the whole idea. Everything else is detail.

A worked example

Suppose you trade this: sell a Nifty straddle at 9:20, close it at 3:20, and get out early if the loss hits 30% of what you collected.

That is already a strategy. You could run it by hand. But “by hand” means being at your screen at 9:20 every day, watching the position all day, and reacting correctly when it moves against you — at exactly the moment it is most uncomfortable to do so.

Written as an algo, the same thing becomes:

ENTRY   09:20  sell ATM call + sell ATM put
EXIT    15:20  close both legs
STOP    loss >= 30% of premium collected  ->  close both legs

Nothing is smarter here. It is the same strategy. What changed is that the rules are now explicit and get followed whether or not you feel like following them.

What algo trading is not

It is not a money machine. An algo executes your edge. If your rules lose money slowly by hand, they will lose money faster automatically.

It is not prediction or AI. The overwhelming majority of retail algo trading in India is simple conditional logic: if this price, at this time, then this order. No machine learning involved.

It is not high-frequency trading. HFT is a different business with different infrastructure and budgets. SEBI’s retail framework sets its threshold at 10 orders per second — and almost every retail strategy sits far below it, placing a handful of orders a day.

It is not illegal. Automating your own strategy is explicitly permitted. See our breakdown of the rules for the specifics.

The real benefit, which is not speed

People assume the advantage is reaction time. For most retail strategies it is not — it is consistency.

A rule executed by a computer is executed the same way on the day it is comfortable and the day it is not. You do not widen a stop because you are convinced it will come back. You do not skip an entry because last week was bad. You do not forget.

The second benefit is that a written-down rule can be tested. “I generally sell straddles on expiry day” cannot be checked against history. sell ATM straddle at 09:20 on expiry day, stop at 30% can be — on years of real data, in minutes.

That is usually where the value actually appears: not in the automation, but in discovering whether the idea ever worked.

What it takes to start

To test ideas: nothing but a browser. You do not need a broker API, a server, or any code. This is where everyone should begin, and many people never need to go further — finding out that an idea does not survive costs is itself the result.

To run it automatically, you need four things:

1. A broker with an API. Angel One, Dhan, Fyers and Upstox offer one free. Zerodha’s Kite Connect is ₹500/month.
2. A static IP. Required by SEBI for self-run algos. Usually means a small cloud server — why, and what it costs.
3. A machine that stays on through market hours. Your laptop is not it.
4. LIMIT orders. Market and IOC orders are not permitted for algos.

The honest order of operations

Most people do this backwards. They buy a platform subscription, connect a broker, and automate a strategy they have never tested.

A better order:

1. Write the rules down. If you cannot state your strategy in three lines, you do not have one yet.
2. Test it on real history — including brokerage, STT and slippage. Costs are what kill most intraday options strategies, and a backtest that omits them is a fantasy.
3. Check it survives being nudged. Change the entry time by ten minutes. If the result collapses, you found a coincidence, not an edge.
4. Paper trade it for a few weeks.
5. Then think about automation.

Steps 1 to 4 need no broker connection, no server, and no money. Step 5 is the only part that needs any of it — and by then you will know whether it is worth the trouble.

Where to start today

If you have never tested a strategy against real data, start there. Our backtest tool runs in your browser on real NSE data, with every cost applied, and nothing you type is sent anywhere. The guide walks through reading a result without fooling yourself, which is the harder skill.