The Gateway to Algorithmic and Automated Trading

Automated trading in India

Published in Automated Trader Magazine Issue 41 Q4 2016

India's securities market regulator, the Securities Exchange Board of India (SEBI), has released a discussion paper on proposed regulation in the algorithmic trading/high frequency trading space.

AUTHOR'S BIO

Nitesh Khandelwal Nitesh Khandelwal is a director at QuantInsti , a training and research institute in the quantitative and algorithmic trading space which is head-quartered in Mumbai. He is responsible for the launch of the Executive Programme in Algorithmic Trading (EPAT), an online course on algorithmic trading for working professionals.

Since the wave of economic liberalisation in the early 1990s, India has made a name for itself within both Information Technology (IT) and analytics. When SEBI finally permitted automated trading in 2008, proprietary trading firms across the country were able to draw on the country's large pool of IT talent. Within a few years, automated trading had claimed the majority of the volume going through India's exchanges.

The initial approval of Direct Market Access (DMA) by SEBI in April 2008 was soon followed by additional measures that allowed automated trading to become established in its own right. Co-located servers and high frequency market data were introduced to the marketplace. Technology had begun to play a crucial role in India's financial market and in the minds of its participants. This was similar to the wave of enthusiasm toward e-commerce that had captured the imagination of the whole nation at around the same time.

Automated trading volume versus overall trading volume in the equity derivatives segment on NSE

Figure 01: Automated trading volume versus overall trading volume in the equity derivatives segment on NSE

Current participation by automated traders

Over the past eight years, leading exchanges in India have seen a healthy increase in the proportion of volume traded through automated trading. As per the National Stock Exchange of India's (NSE) monthly report for July 2016 (available on its website), almost 50% of the volume in equities came from algorithms (see Figure 01). And almost a third of the total volume in the stocks originated in servers co-located with the exchange.

For commodity trading, automated strategies account for close to 30% of the overall exchange volumes on Multi Commodity Exchange (MCX), the leading commodity exchange in India. This is shown in Figure 02. Regulations do not yet allow Indian commodities exchanges to offer co-location facilities.

Broad guidelines for automated trading by SEBI

This growth in automated trading volumes did not go unnoticed by SEBI. The regulator has time and again provided and updated its guidelines on automated trading. The first set of broad guidelines was issued in 2012, followed by additional guidelines in 2013. These key guidelines included:

  • Economic disincentives for high order-to-trade ratios
  • The requirement for all automated orders to be routed through broker servers located in India
  • Price and quantity sanity checks
  • Authority for the exchanges to inquire about details of strategies used by automated traders
  • Automated submission of orders only possible upon prior approval from the exchange
  • Annual audits (or more frequent, depending on exchange requirements) of brokers by Certified Information System Auditors (CISA) who are authorised by the stock exchanges.

These guidelines have helped Indian exchanges to establish best practices and to avoid any major mishaps due to the increased use of technology in trading. Interestingly, they have also created a differentiation on how automated trading has evolved in India when compared to other geographies, especially the more developed markets. Most of the exchanges in India require each new strategy (as well as changes to every strategy) to be documented and approved by the exchange before it can go live.

This is different to developed markets, where most exchanges need approval only at the system level and not for each strategy. (RegAT in the US is trying to change this. Interested readers can refer to this article ).

Simply put: a strategy-based approval system gives more power to the exchanges to control which kind of strategies are being traded on the exchange. For instance, in the case of commodities exchanges, algorithms are allowed only for liquidity providing (read: passive) strategies. They disallow any liquidity removing strategies being traded through algorithms.

Furthermore, Indian exchanges also require the exchange members to preserve 'control parameters' of each strategy for a number of years. This helps them in tracing back what parameters were set by the trader in their strategy that led to the generation of a specific order.

On the infrastructure side, any connectivity coming out of exchanges' co-location facility has to be approved by the exchange. This can impact the way cross-exchange or smart order routing (SOR) strategies are executed in India vis-à-vis developed markets.

Recent events

As tends to happen with increased technological innovation, liquidity has increased, bid-ask spreads have narrowed and information leakage has been mostly plugged thanks to DMA, which has allowed end users to directly execute orders. Even price discovery has improved as algorithms make sure that any mispricing is quickly corrected.

Of course, there are many who believe that they have been 'short changed' by automated traders. This is particularly true of traders who do not have the expertise or resources to invest in the infrastructure required for technology-based trading. Lack of awareness and poor understanding about what automated trading involves and achieves has further shaped this belief.

While the guidelines so far have successfully prevented any significant incidents, more measures may be needed to address the genuine concerns by those market participants to ensure smooth and sustainable growth of Indian financial markets along with the growing Indian economy. Making sure that abundant liquidity is present on the exchanges and avoiding any collusion or malpractice in the industry should be some of the top priorities. The regulator has come up with a discussion paper in August 2016 to address these concerns and to provide fair, transparent and non-discriminatory access to the financial markets.

Automated trading volume versus overall trading volume in commodities on MCX

Figure 02: Automated trading volume versus overall trading volume in commodities on MCX

Discussion paper and way ahead

SEBI's discussion paper on algorithmic trading in India includes proposals such as minimum resting time for orders, frequent batch auctions, randomised speed bumps, order sequence randomisation, restrictions on order-to-trade ratios, segregating co-location and non-co-location orders and a review of market data feeds.

Some of the most significant proposals in the discussion paper are:

Minimum resting time

A minimum resting time of 500 milliseconds has been proposed for every order during which it cannot be amended or cancelled (also known as a 'Minimum Quote Life' or MQL). The idea is to address the concern of the opponents of automated trading that liquidity provided by algorithms is only illusory in nature and vanishes if a trader intends to trade against it. A research and evidence-based discussion on this can help to assuage such concerns.

Increasingly connected global markets have resulted in a considerable rise in the number of news events that affect prices. Algorithms absorb such information and modify the prices accordingly. From the algorithmic market makers' (MMs) perspective, a key concern is ensuring that they are not caught blind-sided when news events happen during the resting order period. Market makers have similar concerns about frequent batch auctions or randomised speed bumps.

Randomisation of orders

The paper has also suggested randomisation of orders received during a given period (e.g. 1-2 seconds). The idea is, that randomisation may help in creating a uniform playing field for both faster automated and slower non-automated traders. While randomisation for a very small period may halt the race towards lower latencies (the flip side being less investment in technology and innovation), such a randomisation might result in inefficient markets as mispricings may take longer to be corrected.

Another side effect could be some market participants sending more orders to increase the probability of being filled on a given trade. This could be controlled by another measure on order-to-trade ratio proposed in the paper. Market makers send a lot of messages to keep their quotes updated to avoid getting filled on stale prices. But modifying orders outside a small trading range should be discouraged, as this impacts the exchange infrastructure negatively for all genuine participants.

Separate queues for co-location and non-co-location orders

Separate orderbook queues for co-location and non-co-location orders have also been proposed where one order is picked from the orders emanating from exchange's co-location and another one from non-co-location in a round-robin fashion. It would provide more equitable access to the order book priority for the non-co-location orders, which may be helpful to the people who do not have their servers in co-location facilities but in other data centers.

In practice this would be difficult to do and would expose the system to further 'gaming' as people jockey for positions in both queues to optimise their overall result. And given that co-located market participants would still be getting market data first, the measure might not achieve its objective.

SEBI has been doing a great job at maintaining a balance of including technology in the trading process while ensuring smooth growth of the marketplace and fair access to as many market participants as possible. The points raised in the discussion paper steer the debate in the right direction and will undoubtedly help in addressing the concerns of various market participants.