Algo trading in India: a practical guide

What an algo actually is

An algorithm, in trading, is a set of rules precise enough for a computer to follow without a person deciding anything: when to enter, how much, where to exit, and what to do when something goes wrong.

That last part is what separates a system from a script. A scheduler that fires an order at 9:20 is automation. A system knows what it holds, checks that the exchange agrees, notices when its data has gone stale, and stops itself when it should.

• Alerts and semi-automation: software watches the market and a person places the order.
• Rule-based automation: orders go out through a broker's API when conditions are met.
• A full system: strategy, order management, risk limits, reconciliation and monitoring, running unattended.

Is it legal?

Yes. Institutions have traded by algorithm on Indian exchanges since SEBI allowed direct market access in 2008. Individuals have done it through broker APIs for years in a grey zone.

SEBI closed that grey zone with its circular of 4 February 2025 on safer participation of retail investors in algorithmic trading, in force from April 2026. It keeps API trading open to individuals and puts the broker in charge of it. The details are on SEBI's algo trading rules.

Who trades this way

• Proprietary and high-frequency firms, co-located at the exchange, competing on microseconds.
• Mutual funds, insurers and foreign investors, using execution algorithms to work large orders without moving the price.
• Portfolio managers, family offices and HNIs, running systematic strategies in futures and options at ordinary speeds.
• Individuals, through broker APIs and no-code platforms.

Only the first group needs speed. For everyone else the edge, if there is one, is in the rules and in running them without error.

What it takes

• A rule that has been tested honestly, on Indian data, with costs, on periods it was not built on.
• Clean data. Corporate actions, expiry changes and lot-size changes all have to be handled, or the history lies.
• A broker API and an understanding of its limits: orders per second, order types, how sessions expire.
• Infrastructure: a server with a static IP address, logging, monitoring, and an alert that reaches a person.
• Risk controls that do not depend on the strategy being right: a loss limit, position limits, a kill switch.

Algo trading infrastructure goes through each part.

What it costs to trade

Every order pays brokerage, securities transaction tax, exchange and SEBI charges, GST on the charges, and stamp duty. Options sellers also tie up margin.

The cost most often left out of a test is slippage: the difference between the price on the screen and the price the order gets. For a strategy that trades often, slippage and charges together can be larger than the edge.

Why most systems fail

• The backtest was fitted to the past. Enough parameters will make any history look profitable.
• Costs were ignored, or assumed away.
• The system does not know its own state: it believes an order was filled, or cancelled, when the exchange says otherwise.
• Nothing watches the watcher. A feed freezes, the system keeps running on old prices, and nobody is told.
• There is no plan for the bad day: no loss limit, no way to flatten quickly.

The first two are research problems: see backtesting Indian markets. The rest are engineering.

Three ways to begin

A platform. No-code tools let you assemble rules and run them through a broker. Quick to start, limited in what they can express, and you depend on the platform staying up.

Your own code. Write against a broker's API. Full control, and full responsibility for everything listed above.

A bespoke build. Your rules, engineered and tested by people who do it for a living, running on infrastructure set up for you. That is what Aurevian builds.

Keep reading