While algorithmic trading systems are designed to provide a fast, anonymous and cost-effective means of executing block trades, traders are finding that algos often require careful oversight and a feel for when to use them and, even more importantly, when not. Most professionals know that algo systems, at least in the case of the "first-generation" products widely used today, work best for large-cap stocks with millions of shares in average daily volume. What is not commonly known though, but is a fact that more and more traders and asset managers are discovering, is that even some of the Street's most popular large caps, ones with seemingly high liquidity, are actually not good candidates for algo trading.
When preparing to execute an order and deciding whether to employ an algo system, traders need to do their homework and pay careful attention to five critical factors:
* Average trade size (in shares) of the equity
* Average intraday price volatility (the spread between the daily high and low)
* How many times during the day the stock typically spikes up and down
* Average day-to-day price volatility
* How your effectiveness is measured (commission cost, execution price, time, etc.)
We've found that, while a number of large caps that appear to be ideal algo candidates, they actually trade more like small caps. These include some well-known, very active names. In many cases, average daily volume is in excess of 10 million shares, and yet average trade size is less than 300 shares. These stocks also exhibit relatively high price volatility, both within the trading day and from one day to the next. That's even when there is low overall market activity and no company or industry news. Despite these factors, their big volume attracts high levels of algorithmic trading. Ironically, it actually may be the algos themselves that are making these and other stocks poor candidates for this kind of electronic trading.
We call these equities Whip Stocks because they demonstrate exaggerated, "whip-like," up-and-down trading patterns. If there is disproportionately heavy algorithmic trading in a particular stock, large numbers of small, automated block trades are fed into the market to match, for example, the day's volume weighted average price (VWAP). The latter is one of the most popular first-generation algorithmic trading methods. This is why a stock's average trade size is so important. Increased algo activity creates an abnormally high number of smaller trades, as large blocks are "chopped" into smaller lots, in an attempt to limit market impact. If there's too much algo activity, though, market impact becomes a problem.
The Problems of Algos
Here's what happens. A large block goes up on the tape and acts as a catalyst for the algo systems to begin automatically feeding their orders into the market. The algos chase that volume, and one another, up or down the price ladder, driving the market beyond where it should be. Once this flurry of activity runs its course, algos often send their opposite trades into the market, driving the price too far in the other direction. This cycle can repeat itself multiple times in a given trading day. As a consequence, Whip Stocks have large spreads between their intraday high and low prices, an unusually high number of up and down spikes, and their closing prices may fluctuate considerably from one day to the next.
Compounding the problem are savvy traders and "predatory" algo systems that have learned to spot these patterns and trade ahead of them, thus exaggerating the Whip Stock effect.
Whip It
Here's an example we recently encountered when trading a large cap, Nasdaq-100 equity for a client, with average daily volume of more than 3.5 million shares, priced at around $28 per share. At 9:46 a.m., a large trade of 38,700 shares went up at $28.04. Immediately, the stock started to spike further upward as a series of small trades flowed into the market. In five minutes, 45,000 shares traded in small lots, running the stock up 1.4 percent to $28.43. There was no company or industry news prompting the move, just that initial block and then the wave of small trades chasing the VWAP (and each other). By 10:01 a.m.-just 15 minutes later-the stock traded back down to $28.10. By comparison, during this same time period, the Nasdaq-100 Trust (Nasdaq: QQQQ) was relatively quiet, moving from $41.03 to $41.10 (up 0.17 percent).
As this example shows, prices will eventually settle, but if you get caught at the wrong end of a whip, profitability can suffer. While you think the algo systems have done a good job, matching or beating the VWAP, in reality they have increased your market opportunity cost. This forces you either to buy the stock at a higher price or sell it at a lower price than what you could have done by "manually" navigating the order through all the various electronic trading systems. All this movement will have likely skewed the VWAP itself as well.
This brings us to another important test for whether you should "go algo." How is your trading effectiveness measured? If it is based on obtaining the lowest possible commissions and hitting the VWAP (regardless of how true a benchmark that may be), then, even in Whip Stock situations, algo trading accomplishes your goals. But if you are measured on the basis of profitability, or if you share in profits like most hedge funds, then focusing on reducing your market opportunity cost is paramount.
Another factor to consider is time. How fast do you have to buy or sell the stock? Sometimes, you have to be out of a stock that day.
If speed is of the essence, algos work well, since you can program a finite period to complete your trades. On the other hand, if you have more flexibility, and profitability is a primary factor, it may pay to explore whether more traditional trading methods work better.
A good trader, who knows how to work a stock and find liquidity wherever it exists, may be able to achieve a cost effective execution. You may pay a penny or two more per share in commission, but lower market opportunity cost savings can make up the difference.
Savings
It is important to note that a second generation of algo systems is beginning to be offered by sellside brokers and trading technology vendors.
The key improvements of these systems are their ability to increase the number of trade parameters, access more liquidity in more markets and act in a stealthier manner.
They may reduce the Whip Stock effect and expand the usefulness of algos for mid and smaller caps. The downside, however, is that the more parameters set for any trade, the fewer trades are likely to be executed within the given period. Indeed, as the second-generation algo systems become as common as the first, they may create a new set of market inefficiencies.
Algorithmic trading is not a cure all. It is another tool in the trader's bag. But an algo can't do everything. Any product or service can compete on the basis of price, quality and time, but almost never all three at once. If you're not careful, what you thought was a bargain could end up costing more.
Joseph Saluzzi is co-founder and co-head of Equity Trading at Themis Trading LLC of Chatham, New Jersey.