Here’s a little trivia to begin with. Do you know much time does it take for the human eye to blink? Don’t blink at this question! The answer is: 1/10th of a second. And given that 1,000 milliseconds make for a second, a blink takes 100 milliseconds. Most humans blink about fifteen times a minute, or every 4 seconds. Now, if someone were to ask you what’s faster than an eye blink, what would your answer be? Blinked again! It’s called high-frequency trading (HFT). Driven by sophisticated computers with complex mathematical models, HFT is a subset of algorithmic trading that involves executing thousands of buy-sell orders in milliseconds to exploit decimal differences in the price of any asset across financial markets. Firms engaged in trading are doing their transactions in microseconds (one-millionth of a second) and nanoseconds (one billionth of a second). Though not all algorithmic trading is high-frequency, all high-frequency traders do use algorithms.
In his book, Flash Boys, author Michael Lewis succinctly describes what HFT is all about. “The US stock market now trades inside black boxes, in heavily guarded buildings in New Jersey and Chicago. What goes on inside those black boxes is hard to say — a ticker tape that runs across the bottom of cable TV screens capturing only the tiniest fraction of what occurs in the stock markets. The public reports of what happens inside the black boxes are fuzzy and unreliable — even an expert cannot say what exactly happens inside them, or when it happens, or why.”
The black boxes mentioned are servers placed by traders closer to the US exchange’s trading platform. The book’s plot reveals how certain traders get their servers placed strategically close to the main exchanges to receive trading orders a split second ahead of the rest of the market and use complex algorithms to make profits.
In India, HFT trading made its debut in 2008 when the market regulator allowed the National Stock Exchange, which controls 90% of the country’s stock market turnover, to offer direct market access to institutional clients. This facility enabled them to directly tap into the exchange’s trading system by using their broker’s infrastructure. In doing so, algo traders, for the first time, gained access to the country’s stock market without the brokerage punching in orders. Two years later, to further bring down the latency — the time taken for order matching and trade confirmation — co-location services that allowed market participants to rent servers situated within the exchanges premises, were introduced. The NSE’s National Exchange for Automated Trading has a latency of single digit millisecond for all orders.
Today, almost 60% of incoming orders at the NSE are from co-located servers. More importantly, automated computer-driven trading or algo trading accounts for over 40% of executed orders in India, compared with an average of 32% across Asian markets in 2015, mentions Aite Group, the Boston-based market research firm. While HFT players are not biased towards any particular asset class, in India they largely operate in the derivatives segment owing to higher liquidity.
While the popularity of algo has caught on, so have the accompanying issues. In the US, the introduction of National Best Bid and Offer (NBBO) mechanism has ended up institutionalising inequality between high-frequency traders and rest of the market. NBBO allows orders to be routed to different exchanges, if the price desired by an investor is not available at the exchange first sought. As HFTs can see the best prices before the rest of the market, they can buy shares at those prices and sell it at higher rates to investors accessing the NBBO.
Closer home, while there is no NBBO mechanism, the Securities and Exchange Board of India is planning to regulate algo trading to create a “level playing field” for those investors who lack access to such advanced and expensive systems. The regulator has floated a discussion paper titled ‘Strengthening of the Regulatory Framework for Algorithmic Trading & Co-location’ and envisages to put certain restrictions to curb the tech-edge availed by a particular class of traders. The discussion paper cites ‘market noise (excessive order entry and cancellation)’ as one of the issues that call for regulatory intervention for HFT in India.
The curbs being planned include “speed bumps” to randomly delay executing orders and forcing exchanges to take orders from co-located servers and other sources alternatively, thus negating the advantage enjoyed by HFT platforms. Besides, the regulator is also considering implementing a minimum resting time between orders (resting time is the period between an order received and its final execution), randomising orders received during the predefined time period of 1-2 seconds and forwarding a revised queue with a new time priority to the order-matching engine.
The global Futures Industry Association (FIA), representing HFT firms such as Citadel, IMC and Optiver among others, has taken umbrage to the development. The association, in a letter to Sebi, has raised concerns over the fact that any kind of regulatory intervention would result in “potentially detrimental impact to market liquidity, increased risk and increased trading costs for investors, which outweigh potential regulatory benefits.” The FIA feels that the introduction of separate queues and order-validation processes for co-located and non co-located orders would unnecessarily introduce a level of complexity to the trade-matching process as well as exchange systems.
Domestic market participants, too, believe any blanket restrictions on algo trades would be a retrograde step. Rajesh Baheti, managing director at the Mumbai-based HFT firm Crosseas Capital, says, “Unless you have statistical data to compare price volatility, unexpected sharp swings, market depth, liquidity in post-algo and pre-algo era, it is not correct to curtail HFT on a generalised fear. It is a premature paper without any scientific evidence.”
Rishi Kohli, managing director at ProAlpha Systematic Capital Advisors, that manages Rs.1,000 crore of assets through quant strategies, feels HFT will lose its meaning if speed bumps are applied on all algo orders. “The challenge is to segregate the indirect front-running trades from pair trades and arbitrage trades where speed bump should be avoided,” he says. Typically, in an algo model, pair trades are generated every second. There can be some kind of tagging like one order number for both the legs of the trade. When there is only one-leg order, a speed bump can be used since there is a chance of indirect front-running to shore up value of an instrument. “However, orders that have multi-legs are likely to be arbitrage, options arbitrage or pair trades where speed bumps can be avoided,” adds Kohli.
Experts warn that certain proposals, in fact, may be seen as regressive by the international investing community. Anshuman Jaswal, senior analyst at financial research firm Celent, says, “In a financial market, you can’t regulate everything. Stopping tick-by-tick data (TBT) would actually put Indian markets behind the curve as TBT is a standard now in most markets.”
Not surprisingly, Manoj Kumar, Sebi’s head of market regulation, at the India Fix conference held on September 1 in Mumbai, mentioned that the regulator “will not do anything in haste and will keep on consulting everybody.” But the damage has been done.
While regulatory curbs are threatening to spoil algo’s party, taxes are already giving heartburn to traders. The share of algorithmic and co-located trading is trending downwards. (See: Speed Bump) The securities transaction tax (STT) on selling options was raised to 0.05% from 0.017% in the FY17 budget. As a result, trading through algos and co-location, which together accounted for 47% of the total derivative volumes, last July, has slipped to 37%.
Samssara Capital’s managing director, Manish Jalan, who also worked as an algo trader with Merrill Lynch and Credit Suisse in Tokyo, cites the rise in transaction costs as one of the reasons for the drop in volume. “Trading strategies focused on selling equity options are now facing issues,” he says.
HFT strategies are typically asset-class agnostic and look to continuously cash in on the difference of basis points between prices. A single transaction for an HFT may be minutely profitable, but the trader needs to enter into several transactions in a day to stay profitable. For example, in the currency market, the tick size is 0.25 paisa. So, on a lot size of 1,000, a trader can make Rs.2.5. One has to do a lot of trade in a day, but this also increases the transaction costs.
The concern for a HFT trader is that an increase in algo trading will turn non-remunerative. “You have a transaction cost of 1 basis point, which is say Rs.1,000/crore. As an HFT, you are essentially trying to make money on a couple of basis points’ trade. If you have to give away one basis point to the government in terms of STT and other costs like stamp duty, etc, you are looking at 1.5 basis points per trade. Then there are other costs in terms of setting up co-location servers and having dedicated lines to exchanges,” elaborates Jalan.
Rajib Ranjan Borah, who heads the HFT firm iRageCapital, that focuses on market making strategies, says that a market-maker’s prime objective is to capture the bid-ask spreads and they are already operating in an environment of high transaction cost. The proposed regulatory slowdown only makes life more difficult for liquidity providing market-markers. “Retail investors have actually benefited immensely from algo-based high frequency market- makers. Because of the resultant increase in liquidity, both impact cost and volatility in the market has reduced. Furthermore, derivatives which are used to hedge exposures have become more liquid with the advent of algo-based high frequency market-makers. Slowing down algo-based HFT might take us back to the age of manual market-makers, and the key beneficiaries will be traditional manual market-makers and liquidity taking arbitrageurs,” he adds.
Playing it safe
In 2010, a flash crash saw the Dow Jones lose 9% of its value in a matter of few seconds. A report by the Securities and Exchange Commission and the Commodity Futures Trading Commission attributed the fall to a combination of a large algo-based sell order (by a mutual fund) executed in a matter of minutes and rapid buying and then re-selling by HFT firms. The rise in volume triggered further selling by the mutual fund as its algo was coded to respond to volume movements and not price changes. But instead of a knee-jerk reaction by putting curbs on HFT trading, the US chose to walk a different path. “The US has got an exchange -— IEX (Investors Exchange) — dedicated to equalise markets in terms of speed. HFT is no more an arms race,” says Jaswal.
While Indian markets have not yet witnessed any flash crashes triggered from algo-based high-frequency trading, the regulator’s desire to rein in high-speed movement of shares in a matter of few seconds doesn’t seem surprising. “Trading is happening at high speeds and such high volume can affect the pricing of pretty much the entire market. The moment it goes wrong even at one broker or one participant, it is very difficult for the regulator to step in. By the time the exchange realises something is wrong and regulator course corrects, a lot of damage would have already been done,” points out Jaswal.
In fact, in October 2010, the NSE had witnessed a flash crash of 920-points on the Nifty caused by a dealer’s ‘fat finger’ error. Though coded robots are yet to lay a wrong finger in India, the broking fraternity doesn’t want to take a chance.
Referring to the Sebi paper, Alok Churiwala, vice-chairman of BSE Brokers Forum says, “There are two sets of people: one, who are using this facility; and a large majority who are not. Some don’t even know what this is. They believe that there is neither a level playing field, nor any concrete economic value-add to a lay investor.”
It is not surprising that when the forum assembled on the 8th floor of PJ Towers, the iconic Dalal Street building on September 7 to discuss Sebi’s paper, a heated discussion ensued with some members questioning the logic of regulating high-frequency trading in India when the world over it remains largely unregulated. After a two-hour debate, the divided house came to a consensus that some regulatory oversight is better instead of none.
Though Sebi has not decided on a time frame for implementing the new rules, the very fact that domestic traders are blinking first is an indication that algo trading has hit a speed bump in India.