Decoded: Sebi's proposed rules to regulate algo trading by retail traders

Mutual fund small and mid-cap stress test results to be announced on March 15, Madhabi Buch, Chairperson, SEBI said on the sidelines of an AMFI event felicitating women fund managers.

Last month, market regulator Sebi proposed that brokers should only offer open Application Programming Interfaces (APIs) to retail clients if they are aware of whether the API is being used for algo trades or manual trades. If the broker is not certain, the client should not be given access to the API.

Through this explainer, we will try to decode why Sebi is putting safeguards in place for algo trading by retail investors.

It's the automated use of algorithms to generate trading signals and execute buy or sell orders with the broker.

'Buy 100 shares of SBI if the stock makes a new 52-week high.'

No. There can be multiple rules built in. Like, for instance, buy 100 shares of SBI when it makes a 52-week high and if trading volumes are at least twice the daily average, and if the Bank Nifty index is trading at least 0.5 percent over the previous close.

Brokers need the approval of exchanges to offer algo trading. They need to inform exchanges about the algo strategy and any changes to them.

All algo orders must be routed through broker servers located in India. Also, all algo orders must be tagged with a unique identifier provided by the stock exchange in order to establish an audit trail. This allows the exchange to know if an order is algorithmic one non-algorithmic.

By using the open Application Programming Interface (API) offered by brokers to their clients.

It is a set of programming codes that analyses data and sends instructions between one software platform and another. In this case, it will be between the broker’s software and that being used by the client.

When clients put in orders through the API, neither exchanges nor brokers are able to identify if they are algo trades or non-algo trades.

A handful of traders who know how to write software code devise their own strategies. But a vast majority buy them from third party service providers.

In three steps:

1. The client hands over his login credentials of his broking account to the algo developer, who then does the trades through the client’s trading account.

2. The algo developer offers a desktop application, which the retail trader instals in his system.

3. The retail trader links his broking account to the developer’s site and chooses from the strategies offered by the developer. In this case, the client does not have to share his login credentials with the developer.

1.  Template ready made algos available on platform offered by the algo developer.

2. ⁠Algos offered by different third parties on a common platform

3. ⁠algos custom made by client

Sebi has two major concerns. One is that many of these algo developers promise lofty returns to gullible traders. Two, algo developers are bypassing rules applicable to registered investment advisors (RIA) as mandated by the rmegulator. The developers so far have tried to stay out of Sebi's reach saying they need not be registered because it is the software that is generating the signals.

No, because it is the developers who are creating the rules which the software executes. Also, it is the developers making tall promises to clients, which is a clear violation of Sebi rules.

Misselling too is a big problem, say industry people. Many algo developers are selling basic strategies to retail traders while giving them the impression that it is something exotic, and charging them hefty fees. There are also allegations of backtesting results

Simply put, what is being packaged ‘savoury harmonious medley of aromatic rice and lentils with a garnish of cilantro’ is nothing but plain dal khichri.

There are also allegations that many algo developers back-tested results showing good returns are unreliable. That is because the developer may have cherry-picked certain time periods and also fine-tuned the trading strategy such that it fits the historical price pattern.

Many algo developers have side deals with brokers. So, the developers tell clients: “If you only buy my services, I will charge you x for every recommendation. But if you trade through the broker of my choice, I will charge you one-tenth of that.” The developer then takes a cut of the commission that the broker earns from that client.

An algo gone rogue can cause huge price fluctuations. There have been instances of that in India as well as globally. To be fair, even an institution’s algo can go haywire. But these algos are vetted by the exchanges. In the case of algos used by retail clients, majority of them are not vetted by the exchanges because the trades are never disclosed as algo trades.

That brokers offer open API only to clients where the broker can be sure if it is being used for algo trades or non-algo trades. SEBI wants to ensure that brokers are aware of who is using their API and for what purpose.

Brokers will then work with only those developers who will be transparent about their algos. Through this route, the regulator is gradually trying to bring algo developers under its watch. Such a move will also reduce misselling to a significant extent. Also SEBI has proposed that the back testing results will have to be endorsed by an independent performance validation agency and not the algo developers themselves. This will get a better picture about the effectiveness of the strategy.

In April 2016, the US Securities and Exchange Commission (SEC) approved a rule proposed by the Financial Industry Regulatory Authority (FINRA) requiring algorithmic trading developers to register as securities traders. The move was primarily aimed at reducing market manipulation.

In the UK, any market participant must notify the regulator Financial Conduct Authority if it is engaging in algorithmic trading.  A firm must provide the following, at the FCA’s request, within 14 days from receipt of the request:

1. A description of the nature of its algorithmic trading strategies.

2. Details of the trading parameters or limits to which the firm’s system is subject.

3. Any further information about the firm’s algorithmic trading and systems used for that trading.

While it is not clear at this stage, traders using their own strategies too will need to inform their brokers about the algo stratgies. If institutions have to disclose their algos to stock exchanges, there is no reason why retail traders should be treated any differently.

Any investment strategy works well when it is known to minimum number of people. The more well known a strategy becomes, the less effective it will be in generating profits. So mass market strategies rarely deliver good returns consistently.

But the bigger criticism of algo developers is that if they are so convinced about their strategy, why are they giving it out for a pittance. Why don’t they invest their own capital and make even bigger profits?

It largely depends on the strategy being followed. A handful of retail traders may win, but the vast majority are no match for quant firms that spend a fortune on hiring the best talent available and invest heavily in state-of-the-art hardware. As one industry observer put it: “If you are serious about algo trading, you need to have co-location server facility, because in most cases speed is everything. Trying to win the game by trading over the net is like bringing a knife to a gunfight.”