Explaining the latest SEBI algo trading regulations

SEBI put out a framework to regulate the participation of retail investors in the algorithmic trading space. Now that we have a set framework, we thought it will be appropriate to put out an explainer on this given the various modalities involved. We’ve also made a video explaining the regulations.

Before we begin, here’s an ELI5 version of algo trading. Skip this section if you already know what algo trading is.

What is algorithmic (or algo) trading?

Algorithmic trading (or algo trading) is when you program computers to automatically buy and sell stocks using pre-set rules. Instead of a human clicking “buy” or “sell”, a computer program executes trades at high speeds, following strategies designed to spot opportunities and manage risks. These strategies can be split into two main types based on how often they trade:

• Low-frequency strategies make a few trades over hours or days. For example:
  • Trend-following algos spot market movements (like buying when NIFTY 50 crosses 20,000)
  • Arbitrage algos find price differences between exchanges (buying low on NSE, selling high on BSE)
  • Mean reversion algos look for stocks that have dropped unusually fast, betting they’ll bounce back
• High-frequency strategies (HFT) execute several thousands of trades per millisecond. For example:
  • Market-making algos constantly place buy and sell orders, profiting from tiny price differences
  • Statistical arbitrage algos spot and exploit price gaps between futures and spot markets
  • Latency arbitrage algos take advantage of their speed, reacting to market moves before others can.

What are the regulations? What changed?

For individuals

Until now, there was no method recognised by regulators that retail traders could use to automate their order placement. The only way to do any “algo trading” legally was to get a co-located (hosted in the exchange premises) setup at the exchanges. This, as you might imagine, is quite expensive and only made sense for algo trading firms set up for the express purpose of running high-frequency trading strategies.

Of course, there were always ways for retail traders to automate their trades using things like Excel macros, scripting actions on the web platform, and, more recently, broker APIs. However, there was a lack of clarity from regulators on the regulations surrounding such practices.

The latest circular provides clear guidelines on API-based trading, recognizing it as a legitimate practice and defining what brokers and traders can and can’t do. The regulations assume that most retail traders who are automating their trades are only doing so as a quality of life enhancement and not running complex, high-frequency, latency-sensitive strategies. This means it is safe to assume that the order rate (number of orders placed per second) for such traders won’t be too high. So the regulations require exchanges to prescribe an order rate below which traders needn’t register their strategies with the exchange. Anyone wanting to trade at a higher frequency will be required to register their strategy with the exchange to do so. At Zerodha, we already apply a rate limit of 10 orders per second on our APIs as well as Kite. We are guessing the exchange limits will be above our own pre-existing rate limits so life will change very little for our users.

While retail traders won’t need to register their strategies when training via APIs, they will be required to get a static IP and only use the APIs from this static IP whitelisted by the broker. This is to prevent unregistered (and disallowed) money management services where one person places orders for many people via shared account passwords. Not only are such services not legal, but sharing account credentials also severely compromises the security of trading accounts.

For algo providers

There have always been businesses (outside of the regulatory ambit) that sell strategies and algos to traders. Such service providers exist to bridge the gap between traders who have a strategy but don’t know how to code and developers who can help them program their strategies. Now, this circular also recognises these businesses and prescribes the way they can operate. Algo service providers can now act as agents of stock brokers, offering services to traders through a partnership with brokers. Brokers will need to empanel such services with the exchanges, not unlike how Authorised Persons are registered with the exchange. Any non-compliance on the part of these businesses will lead back to the broker, effectively ensuring brokers don’t partner with unscrupulous actors.

Algos sold by such providers (irrespective of the order frequency) will be classified into two groups:

• White box – Algos where the logic is disclosed and easily replicable.
• Black box – Strategies where the logic is not known to the user.

White box algos will need to be registered with the exchanges once and can be offered to any trader once registered. To be able to offer black box algos, the algo provider will have to necessarily get a Research Analyst (RA) license from SEBI and comply with all the additional compliance that comes with the RA role. This includes maintaining and publishing a report of the strategy periodically and reporting any change made in the strategy to the exchange before making it available to customers. All orders from these algos will also be tagged with a unique identifier issued by the exchange, so there is a record of what algo placed which order (irrespective of the order rate) with the exchanges.

The exchanges are yet to set up a framework for all such registrations. The SEBI circular gives them up to April 1, 2025 to set this up. We’ll post a further update with details once this is done.

About marketplaces and revenue sharing

Algo platforms have developed marketplaces where traders can share their strategies, allowing other users to copy or consume them. However, it is now evident that this model cannot continue, as individuals are permitted to share their strategies only with family members. Meanwhile, algo platforms can publish their own strategies, provided they hold a Research Analyst license, and the algo is registered with the exchange. The role of the upcoming Performance Validation Agency for Algos will also be crucial in regulating this space.

Additionally, the new regulation explicitly permits the sharing of brokerage revenue and subscription charges between algo providers and stockbrokers. However, the same line of the regulation also says the broker has to ensure there is no conflict of interest in such arrangements. This seems to be self-contradictory and we’ll have to wait for more clarity on this.

TLDR

• If you’re a retail trader using broker APIs to automate trades, you can continue to do so as long as your order frequency is below an exchange-prescribed threshold.
• If you’re selling algos or strategies to other traders, you’ll need to partner with a broker and get yourself registered with the exchanges to to so.
• Marketplaces where people share algos for a fee can’t publish strategies without exchange registration as well as an RA license.