Algorithmic trading
Also written Algo trading · Automated trading
Any order generated by automated execution logic rather than typed in by a person — software that pushes buy and sell orders into the exchange once its parameters are met.
In plain language
The regulatory definition is deliberately wide and deliberately unglamorous: any order that is generated using automated execution logic is algorithmic trading. It does not matter how clever the logic is, how fast it runs, or whether a hedge fund or a retail investor wrote it.
The companion definition of automated trading spells out the mechanism — any software or facility by which, on the fulfilment of specified parameters and without manual entry of orders, buy and sell orders are automatically generated and pushed into the exchange's trading system for matching.
Everything that follows in the rulebook is a consequence of removing the human from the loop. A person who fat-fingers an order places one bad order. A misconfigured algorithm places bad orders until something stops it.
How it works
SEBI's framework (circular CIR/MRD/DP/09/2012 dated 30 March 2012 and the later master circular) builds the controls in layers. All algorithmic orders must be routed through broker servers located in India. The exchange must run order-level risk controls — price check, quantity limit check, order value check. A broker may offer the facility only with the prior permission of the stock exchange, and must satisfy the exchange that it has the minimum risk controls in place at its own end.
Then there is an audit obligation that catches candidates out: a broker providing algorithmic trading must subject its algo trading system to a system audit every six months, carried out by a certified system auditor. Ordinary system audits are less frequent; the algo one is half-yearly.
SEBI's circular of 4 February 2025 on safer participation of retail investors added the API layer. Where algo trading is offered through an Application Programming Interface, the broker acts as the principal and the algo provider, fintech or vendor acts as the broker's agent. An algo written by a tech-savvy retail investor using their own programming knowledge must be registered with the exchange through their broker only if it exceeds the specified orders-per-second (OPS) threshold, and brokers must build systems to detect and categorise all orders above that threshold as algo orders. Algo providers are not directly regulated by SEBI, but must be empanelled with the exchanges.
A worked example
Two brokers apply for membership on the same day.
| Broker P | Broker Q | |
|---|---|---|
| Trades for clients | Yes | Yes |
| Proprietary trading | Yes | Yes |
| Offers algo trading | No | Yes |
| Base Minimum Capital deposit | Rs 25 lakh | Rs 50 lakh |
| System audit of algo system | Not applicable | Every six months |
Switching on algorithmic trading doubles Broker Q's BMC deposit, from Rs 25 lakh to Rs 50 lakh, and Rs 25 lakh of that must be in cash and cash equivalents because at least 50 per cent of the deposit has to be. No exposure is granted against any of it. That is before the cost of the half-yearly certified system audit, the exchange permission, and the India-located servers.
For Broker Q's retail client running her own strategy through the broker's API: below the exchange's OPS threshold she registers nothing; cross it, and the strategy must be registered with the exchange through Broker Q.
Why NISM asks about it
Chapter 3 (section 3.1.1) carries the definitions of algorithmic trading and high frequency trading, Chapter 3 section 3.4.5 carries the system audit obligation, and Chapter 4 links it to money through Table 4.3, where "All Brokers with Algo" is a distinct and much more expensive BMC category. Expect a definition question, a question on the six-monthly audit, and a question that pairs an algo broker with its Rs 50 lakh BMC. With 25 per cent negative marking against a 50 per cent pass mark, guessing between "six months" and "one year" on the audit interval costs you a quarter of a mark on top of the one you did not earn — leave it blank if you do not know it.
Common exam traps
- HFT is a subset of algo trading, not a synonym. High frequency trading is the latency-sensitive type, characterised by high daily portfolio turnover and a high order-to-trade ratio. All HFT is algo; most algo is not HFT.
- The algo system audit is every six months, by a certified system auditor. Do not answer "annually".
- Retail algos are not all registrable. Registration through the broker is triggered by exceeding the exchange's specified orders-per-second threshold, not by writing an algo.
- Principal and agent, in that order. Under the API framework the broker is the principal and the algo provider is the agent. Reversing them is a favourite distractor.
- Algo providers are empanelled with exchanges, not registered with SEBI. They are outside SEBI's direct registration net.
- Prior exchange permission is required before a broker offers the facility at all — it is not a notification after the fact.
Where this is taught
- Series XVI · Chapter 6: Trading Mechanismintroduced here
- Series VII · Chapter 3: Introduction to Securities Broking Operationsintroduced here
- Series III-A · Chapter 12: SEBI (Stock Brokers) Regulations, 2025introduced here
- Series IV · Chapter 6: Trading Mechanism in Exchange Traded IRDintroduced here
- Series I · Chapter 6: Trading Mechanism in Exchange Traded Currency Derivativesintroduced here
Related terms
- Base Minimum CapitalThe deposit every trading member must keep with the exchange purely to meet contingencies — it earns the member no trading exposure at all, and its size depends on what kind of trading the member does.
- Co-locationA facility offered by exchanges allowing brokers' and data vendors' systems to be located within or near the exchange premises and connected through a direct and private network.
- Direct Market AccessA facility allowing brokers to offer clients direct access to the exchange trading system through the broker's infrastructure without manual intervention.
- High Frequency TradingThe part of algorithmic trading using latency-sensitive strategies and high speed networks.
- Trader WorkstationThe terminal from which a member accesses the trading system, with a unique Trading Member ID and User ID, showing both the member's own transaction information and market information.
- Kill switchA pre-trade facility that lets a trading member cancel every one of its outstanding orders with a single command — the emergency brake for a malfunctioning terminal or algorithm.