Today’s world of electronic trading and computerized trade-matching has allowed a proliferation of programmed high-frequency algorithmic traders to enter the arena under the guise of liquidity providers. There is good reason for this trend. Computers have tremendous power, not just in terms of speed of execution, but also the added advantage of processing multiple calculations on the fly that humans simply cannot accomplish with the same speed.
We, as mere mortals, can be pushed back on our heels when going up against these programmed machines, but not all is lost. Many complain about algorithmic, or algo, trading, which some contend has changed how traditional trading patterns form and unwind in the markets. The high-frequency algo trading systems have managed to change the landscape and, for many strategies, confuse the overall picture. (Algorithmic trading encompasses a lot of different types of traders and is not always high-frequency trading [HFT], but HFT nearly always is algo trading).
The average high-frequency algo trading approach has been programmed to recognize pretty much all known trading strategies. Trading systems range from Fibonacci to Gann, to seasonals, to moving averages. More complicated systems trade approaches such as pure auction volume, volume-weighted average price (VWAP), time-weighted average price (TWAP), price destination, implied shortfall and end-of-day. The list is limited only by the imagination.
Many of these systems include a version of pattern recognition, as well. There even are news-driven algo systems that instantly scan government reports for some form of better- or worse-than-consensus results. More complicated systems have correlations, probability analysis and cross-market basis studies, where multiple layers of orders — some long, some short — will re-create synthetically either a basket or even a currency. For example, a long Dax, short S&P, long U.S. bonds, short German bunds equates to a long euro/dollar position.
It was no accident that Fidelity bought Wealth Labs — a build, design and backtester of systems — allegedly because the company wanted to know what others were doing, so that it could take advantage of them. In other words, perhaps Fidelity wanted to assume the opposite side of the systems on the basis that the vast majority of systems are designed to be trend-followers.
Understanding algos
The role of the high-frequency algo trading system, however, is not to trade the trend, but to enter and exit the market as frequently as possible for miniscule individual profits in a range environment. It does this by identifying what have become known as “child orders.” These are large orders that have been chopped down into small, minor orders, partly to hide the order’s footprint and partly to avoid disturbing the market’s existing liquidity and driving price away from intended execution levels.
The controllers of the high-frequency algo trading system can manipulate the degree of aggression in the system. They can be aggressive by hitting bids or lifting offers, or be passive by using resting limit orders, or even neutral by running a long/short book.
They can create iceberg, or submarine, orders where only a part of the order is visible. Once the portion of the order that is shown has filled, it immediately re-loads with additional orders. All this occurs in the blink of an eye, and the system, almost imperceptibly like the speed of a paddle wheel, can accelerate or slow or shift from buying to selling instantly. These algos more likely are attempting to spot other traders’ icebergs and exploit them.
Algos frequently are turned off en masse. This is most noticeable in the forex market when a government report is awaited. In the few seconds just before release, the machines may widen their bid/offer spreads dramatically. Some even seem to switch themselves to a latent state, creating a massive degree of either gap or volatility in the immediate few seconds post release. For example, most experts believe the May 6 flash crash was more likely the result of high-frequency algos leaving the market than pushing it.
Firing all cylinders
Many systems are programmed with 17 to 20 risk parameters to take into account the number of concurrent systems, some that will send buy-orders and some that will send sell-orders. The objective is to fire all cylinders where there is an element of trades that match each other, others that create a net-short in some markets and others that create a net-long . The algo wants to find that level where the trades create a slight movement that triggers competitors’ orders to participate in the price discovery process.
These days, high-frequency algo trading strategies also have been designed to analyze competitors’ orders to try to take advantage of either size or speed. This creates an additional element of algos feeding off each other.
The long-term effect will be a fundamental change in market structure. Unless regulatory authorities step in and somehow prevent orders designed to create the impression of activity, eventually the machines will have nothing to feed off. However, authorities do not know where to look or even what questions to ask. The elusive needle in the haystack is buried so deep within millions of lines of code that no audit trail could find it. Additionally, all notions of churning, which in regulatory terms refers to the excessive buying and selling of securities by a broker for the purpose of generating commissions, are being ignored, as is order front-running, something machines have finessed. All this is perceived as liquidity, but in reality it is a mirage. When the machine is overwhelmed, such as on May 6, 2010 when the Dow dropped 1,000 points in less than an hour, then chaos occurs.
Strategies
For retail traders, taking advantage of the algo trading landscape is a challenge. However, there is some hope. The retail traders’ objective is to align themselves with the high-frequency algo traders rather than fight them. In other words, determine when the algos are buying or selling and trade directionally with it.
This is a difficult proposition. The high-frequency algo trading systems are not designed to be directional, but like the old floor local working the bid/offer spread to its maximum advantage while nevertheless moving price up or down like a game of Chutes and Ladders. The VWAP is one of the more simplistic tools for retail traders to follow that taps into some of the order-flow peculiarities of the algo-trading phenomenon.
The VWAP, available as an indicator on nearly all major charting programs, is calculated by dividing the sum of the dollars traded (price multiplied by share volume) for every transaction by the total shares traded for the day.
The VWAP formula is below:
Where:
Pj = the price of trade j
Qj = the quantity of trade j
j = each trade taking place over the time period
“Volume alert” (below) charts the March euro FX futures on Jan. 27, 2011. The chart identifies the trading days of the major financial centers.
The center green line is the VWAP with one and two standard deviations (STD) marked. The blue and red arrows are suggested entry points, buying and selling to take advantage of the price discovery. The 100-period simple moving average (pink line) can be used to aid with trend direction.
When the price bar closes through the VWAP, then a move to the second STD should be anticipated. Use the VWAP as a median line similar in nature to a pivot that once penetrated and confirmed by a close indicates that further advancement is likely. On subsequent bars, if there is no further advance, then a pullback to the first STD can be expected — again, using the close as a confirmation with a stop and reverse strategy either on a failure below the VWAP or at the top/bottom of the range and the second STD.
Unlike algo traders, other day-traders tend to figure largely into the category of momentum strategists. They look for breakouts and like to buy high in the hopes of selling higher. They will tend to get trapped by the whipsaw of the algo and end upside-down on their positions.
Long-term traders have a time horizon that instead looks at the market in a holistic way and targets the opportunity for a larger advance. Many algo systems are programmed to detect long-term trader behavior and establish the degree of power of the buying pressure. The primary question is whether the long-term trader will be passive and try to buy low and create a basing action, or be aggressive and buy as price rises? This is when child orders come into play. By monitoring child orders, algo traders can sniff out long-term trading action and effectively front-run orders, believing that if there was a buyer at x, then they will pay x-plus to fill additional positions.
TWAP is the domain of the institutional trader. Essentially it is akin to a dollar-averaging process. Institutions may wish to achieve a particular percentage or market share where time becomes an added influence rather than purely volume. For example, they may buy a line of stock whereby they remain constant at, say, 10% of the volume transacted. An alternative would be to transact a certain quantity split into x units every minute over a 15-minute time period.
This creates the impression that they will not upset the natural order flow and balance of the market. The reality is that an institution, despite being price conscious, needs to reflect the average of all their buys vs. sells rather than the actual price. This is complicated additionally in that the objective of some funds is to invest and monitor performance on a relative basis. This might be performance measured against their peers or some benchmark, whereas other funds are more about actual performance.
Unlike these large players, retail traders are not blessed with inordinate amounts of money and simply seek fair value calculation identified by volume. However, pure volume probably is irrelevant, as it is the pattern created that becomes key. After all, how many times have traders intoned how volume is weak yet price continues to grind higher? Tools such as VWAP can correct this by providing a relative view of the markets, and traders thereby can align themselves with high-frequency algo traders rather than fight the machines.
Alex Benjamin Sassoon is a 35-year veteran of the financial markets and has worked in every aspect of the business, from market maker to portfolio fund manager and broker to individual trader. He also runs TradingClinic.com, a financial portal focused on trading education. E-mail him at alex@tradingclinic.com.
