StockFundoo Methodology

Powered by Fundamental Deep Value Investing and Technical Analysis
Detailed Stock Analysis updated on a Daily Basis
Daily Nifty View and Market Trend Analysis

Thursday, November 1, 2012

Why High Frequency Trading (HFT) is not so profitable anymore!

High Frequency Trading, or popularly known as HFT, is a set of trading algorithms which aim to profit by lightning fast trade executions in electronic trading markets. More than 30% of currency trading in world markets and about 90% of stock trading in exchanges like NASDAQ is carried out by computers using HFT algorithms. These HFT algorithms buy and sell shares rapidly, typically in milliseconds or less, and their total share holding period is fraction of a second. Unlike the typical investor, which buys and holds the stock for a few months or so, HFT algorithms average stock-holding period is just about 11 seconds! And how do they make money in 11 seconds, let’s have a look:

Just to share a few examples, some hedge funds have created HFT algorithms, which will benefit from very small difference in prices of a stock, commodity or a bond between various exchanges in global markets. For example, if Nifty is trading at 5600 at NSE, India and 5601 at Singapore Exchange (SGX), these algorithms or algos can quickly buy at NSE, where it is cheap by 1 point and keep selling at SGX, where it is dear by 1 point, to pocket that small 1 point difference between the exchanges for the same asset class. The small 1 point trade profit in few milliseconds would quickly turn out to be a large amount by the end of the day. This  process repeated over a thousand times by the computer software for all the asset classes which trade globally would result in a significant profit in the range of millions of dollars by the end of the year.

There are hedge funds which have designed HFT algorithms to sense large “whale-sized” orders which reach a given exchange. The large whale sized orders will be typically fed by mutual funds dealers or buy-only insurance funds which manage billions of dollars of assets in stock markets through conventional route of buy-and-hold. The HFT algorithms would then trade in front of the large order to benefit from price fluctuation that the large whale order will bring in the price.

The way this works is that these HFT algorithms would keep their bids and offers active in the exchanges to sense the orders coming and would actually keep on cancelling these bids and offers electronically at microsecond speeds or even faster. As per estimates, over 90% of orders entered by HFT algorithms are cancelled and not meant to get executed. This way, they don’t actually do any trade, but keep fishing for other orders appearing in the market.

Once they sense a large order for a stock e.g. 1 million Buy Order for Reliance Industries or 2 million Sell Order for SBI, just to take a example, they would place their orders quickly in front of these orders using special Order types and benefit hugely from change in the price that these large Buy, Sell orders would bring. Sometime back, there was a statement from Indian Exchanges officials that these algorithms are already responsible for 0.7% to 0.8% impact cost to prices that are available to institutional investors. This used to be a hugely profitable strategy and bread and butter stuff for HFT algorithms for a number of years.

Yet another set of HFT algorithms would benefit from maker-taker model in electronic markets. In several electronic markets, with a view to boost liquidity, traders looking to buy or sell will be the “takers” and will have to shell out a small amount e.g. a few paisa or few cents to get the order fill. And the traders on other side, typically an electronic HFT program would keep fishing for these takers, by fast moving bid-offer quotes and act as maker of the liquidity. These HFT algorithms will be called the “makers” and would then get paid fraction of these cents by “takers” for providing the liquidity to the takers. Replicating this process, day-in and day-out would bring in millions of dollars annual revenue for these maker-taker HFT algorithms.

These are just a few commonly known examples of HFT algorithms. There are newer trading algorithms developed on a daily basis, which use techniques such as Artificial Intelligence or Machine learning, algorithms able to work in dark pools, predatory fast algorithms feeding on other algorithms’ mistakes and slow speed and so on. Trading Algorithm Designers and Programmers, typically known as Quants are Math, Statistics and Physics Ph.D’s who would sweat day in and day out to devise new trading and programming logic to mint money out of electronic markets.

To share the impact of these HFT trading shops, total profit made by HFT based Hedge funds was $4.9 Billion in 2009, when these algorithm strategies actually peaked due to mayhem and volatility in stock markets. Surprisingly, the profit made by HFT funds is in gradual decline from 2009 onwards. In 2010, these funds made a profit of about $2 Billion and this have come down to $1.25 B in 2011. This has caused turmoil in high speed trading funds and many HFT Hedge funds are now closing shop, laying off traders and simply turning off their computer strategies which are unable to mint money anymore. Why is HFT industry in dire straits now?

There are several simple reasons why HFT trading cannot make money ad-infinitum. The simplest reason being lot of High Frequency algorithms are dog-eat-dog kind of strategies, where one strategy feeds on mistakes or slow speed of another HFT strategy. When strategies get smarter and faster, and people behind these programs correct their mistakes, there is very little money left on the table for the winner to pick. Similar to human evolution theory, winners remain to fight another day and loser strategies pack their bag and leave.

Another reason why HFT trading isn’t making lot of money now is that most of the strategies which were feeding on spread between asset classes globally, now act so fast and with such great volumes on these spreads that they have in-turn caused the spreads to shrink to almost nothing now. E.g. If multiple high speed programs are buying Nifty at an exchange where it is 1 point cheaper and selling at an exchange, where it is 1 point dearer, very soon this 1 point difference will evaporate. This is good news for investors and buy-only funds as they can buy and sell at much thinner spreads and save on their trading costs. But, this is a death knell for algorithms predating on spreads and very soon they turn unprofitable or barely profitable.

Yet another set of customers whom HFT algorithms used to predate on was the large Buy Side customers – the whale orders coming from buy-only mutual fund and Insurance companies looking to invest in markets for longer duration. Now, after getting skinned for a long time and seeing the front running HFT algorithm making money off them, most of these large investors have developed or bought buy side HFT algorithms. The buy side HFT algorithms break the single large buy or sell orders in thousands of small orders, fed into the market using a randomized process to fool the HFT predators. This has caused an algorithm-race between HFT algorithms trying to fish the whale orders, and smart-buy algorithms which break a large order into randomly placed small order throughout the day. This race has eaten into profits made by front running HFT algorithms and pushed many large buyers into dark pools where HFT algorithms find it difficult to fish for whale sized orders. This is another reason for declining profit for HFT industry.

In 2008 and 2009, when markets were volatile, hugely bearish and prices used to crash during the day causing big gaps in asset valuations across cash segment, future prices and option prices, several algorithms were hugely profitable. In today’s calmer times of 2010 and 2011, where volatility has died down, several of these HFT algorithms have found it difficult to predate and make money with the ease seen in 2008-09 timeframe.

With declining profits due to the above reasons, and rising cost of hardware – now super hardware is required to process penta-bytes of data per day -  and increasing data feed costs from exchanges, several of these cottage HFT funds have realized that it’s no longer worth the effort. Large Hedge funds now need to innovate very frequently, weed out loss making strategies and design new strategies regularly to keep running their shop profitably. HFT is here to stay for now, but Innovate, improvise or fade out is the new mantra for this erstwhile tremendously profitable HFT industry for the next few years. 


  1. I began making this point years ago, when algos and HFT were smaller, and simpler. The few then were taking out barrows of money, like shooting fish in a barrel. As there was no attempt to leash them (Greenspan even considered them one of his "tools" for awhile), they flourished.
    You particularly have, as in most articles on the subject, ignored the most egregious rat pack in the group. The ones that I call "The Cyborgs." (see my archived charts)The minor Cyborgs are the ones who keep all of the major indices in lockstep. The major Cyborgs are the one(s) who set the intraday pivots, and regularly crash the markets, generally for half-hour intervals. They are, in my opinion, at least as dangerous to the market as "The Fiscal Cliff." Legal pickpockets, now "TOO BIG TO JAIL."
    It was clear years ago that with a feckless SEC that just got fecklesser and fecklesser, "... our only hope may be that they will become their own worst enemies." And, competition did begin to have an obvious effect over the years. In those early years, it was a matter of cannibals dividing the pie into ever-smaller pieces. Eventually, they had to begin eating each other.
    Knight Capital is a recent untold story of how the availability of uncontrolled HFT tools have corrupted the market. They got caught, but after how long?

  2. Good information about High Frequency Trading it is very helpful for peoples.
    historical option prices


Related Posts Plugin for WordPress, Blogger...