Supercomputer Stock-Trading Robots: What's Overblown and What Needs to Be Fixed

The Knight Capital (NYS: KCG) trading debacle actually itself resolved fairly well. There were no serious market disruptions, no significant harm to investors, no bailouts, and no fallout to the real economy. Knight apparently screwed up, and Knight (and its shareholders) alone paid the price.

Still, there's no guarantee the next mistake to befall the relatively young high frequency trading industry will clean up so easily. After all, the fact that it hasn't harmed investors yet in its brief five-year history isn't a good enough reason to ignore the challenges HFT poses. And the rapid near-failure of a major trading operation -- less than two years after the 2010 Flash Crash -- has once again raised questions about what HFT means for long-term investors and for market integrity.

Of course, some of these concerns are important, while some are overblown.

What's overblown
A commonly worn theme in the debate over high frequency trading is the notion that HFT-related mistakes supposedly have shattered investor confidence by proving long-term investing dead. For instance, in his Wall Street Journal column, Jason Zweig quoted one investor panicked by the May 2010 Flash Crash and Knight Capital debacles:

But that's precisely the opposite of what these events showed. It was HFTs, short-term speculative traders, and investors with itchy trigger fingers who got burned. Meanwhile, long-term investors were unaffected. Someone who bought shares of the Dow (INDEX: ^DJI) the day before the 2010 Flash Crash and who didn't bother to check his or her portfolio for a week wouldn't even know the flash crash had occurred based on a one-week return of 0.3%. An investor who held such a position until today would have gained 20%.

For short-term traders, it's another story. Long-term investors may not compete directly with HFTs, but day trading for most individual traders is an even bigger sucker's bet than ever before. Human traders simply can't hope to compete on short-term strategies that rely on speed against supercomputer algorithms written by the top mathematical minds.

Now, even as HFT reinforces the importance of investing for the long term, it can have an impact on how we ought to execute trades.

Stop-loss orders, which automatically sell shares after a predetermined loss, are less appealing than ever in a world where shares fluctuate more.

Limit orders help to protect you against so-called "micro flash crashes," which can occur when high frequency algorithms momentarily refuse to buy or sell a stock because they're freaked out over strange data, resulting in, say, a market sell order filling at $0.01.

The bottom line is that the threat HFTs pose to long-term investors is vastly overstated. Still, exchanges, regulators, and policymakers need to make sure that HFT doesn't interfere with orderly functioning stock markets.

The SEC has already implemented several safeguards in the wake of the 2010 Flash Crash. For example, exchanges are now required to implement "circuit breakers" that pause trading in stocks when they experience rapid price changes. Unfortunately for Knight, its trading errors began in the first 15 minutes of trading, when circuit breakers didn't apply in order to allow markets more flexibility in setting opening prices.

But perhaps the most important things the SEC has done so far is to acquire real-time trading data from HFT specialist Tradeworx, as well as to begin assembling a database that should allow it to get a full picture of markets and follow the life cycles of specific trades.

Potential dangers
Until the SEC has had time to collect and analyze its new data, there won't be enough information to draw hard conclusions. Nevertheless, here are some of the criticisms of HFT that ought to be considered:

Predatory algorithms: Critics charge that high frequency traders might be utilizing unfair or unethical strategies that harm investors and other traders. For instance, some algorithms might be using "order anticipation" strategies to detect a large order and then quickly buy up all the available shares a penny below a buy order or a penny above a sell order. In this way, they could capture the price differential between buyers and sellers of stocks without contributing any efficiency or liquidity to markets. Alternatively, using so-called "momentum ignition" strategies, HFTs could try to quickly push the price of a stock up or down by quickly issuing and cancelling buy or sell orders to fool other algorithms.

Of course, plenty of legitimate strategies are available to high frequency traders, including a variety of arbitrages -- buying and selling similar securities to profit off of tiny price differentials. Opportunities include the same stock trading at different prices on different exchanges, indexes trading out of sync with their component stocks, and stocks that could be more in-line with their corresponding options.

Many of the high frequency traders we spoke to commented that it would be logistically difficult and legally stupid to manipulate markets, since the SEC can request to see algorithms to ensure that strategies are above-board. Better data should help to confirm or refute the scope of various predatory strategies.

Quote proliferation: HFT has given rise to booming quote volume without necessarily a corresponding increase in trading volume. Critics like data collection firm Nanex point out that massive quote proliferation increases data costs by consuming bandwidth. Some also question whether heavy quoting is being used for more nefarious purposes, say, to manipulate stock prices, confuse competing algorithms, or even gum up exchanges to effectively create a fog of war that might yield an informational edge.

Manoj Narang, CEO of Tradeworx, offered us a more innocent explanation for surging quote volumes: Current SEC regulations disallow locked markets -- that is, when bid and ask prices are the same. When trading moves more quickly than the consolidated data feed, this can cause problems for traders trying to trade at current prices. As a result, those traders end up having to place and cancel quotes until the data feed catches up.

Phantom liquidity: Except for official market makers, HFTs aren't required to post offers at all times. This can give rise to a phenomenon known as "phantom liquidity": large orders can't get filled because algorithms get spooked by their presence. But liquidity is meaningless if it's not there when markets need it. As the SEC observes, "high trading volume is not necessarily a reliable indicator of market liquidity."

Systemic risk: The 2010 Flash Crash brought into focus the potential HFTs could have for increasing large-scale market failures. In its autopsy of the event (link opens PDF), the SEC found that phantom liquidity played a role when multiple HFT algorithms suddenly curtailed trading for their own safety.

In addition, some worry that algorithms interacting with one another can trigger a cascading effect where HFT buying or selling self-perpetuates. And HFT algorithms that trade across multiple markets increase their interconnectedness, so that a hiccup in one market could lead to trouble in others.

Although HFTs obviously take care to avoid writing flawed code, the relatively short shelf life of algorithms (weeks) and the need to write efficient computer code require frequent updates to programs and an incentive to minimize encoding safety mechanisms.

Finally, HFTs' speed and opacity amplify the potential for systemic risks by making it difficult for exchanges and regulators to anticipate problems and offer timely responses.

Of course, Knight's near-collapse didn't create systemic failures, and the flash crash was short-lived and didn't directly affect the real economy. The securities that HFTs trade are listed on exchanges and run through clearinghouses, so the danger of a market freeze-up like the one that befell the opaque mortgage-backed securities and over-the-counter derivatives markets in 2008 is smaller. Still, HFTs are a new phenomenon, so it's important to look ahead to where the dangers might lie.

Reduced investor confidence: The evidence that individual investors have lost confidence in the stock market due to highly publicized HFT fiascos is much shakier than commonly portrayed. Still, anecdotal stories suggest that market failures haven't instilled confidence in the system -- even if long-term investors have been unharmed.

Waste of resources: HFTs pulled down $12.9 billion in profits in 2009 and 2010 and spent billions more on state-of-the-art high-speed servers, fiber-optic wires, and compensation for the top mathematicians and physicists. Like other traders, the HFT arms races consume huge amounts of financial and intellectual resources that might arguably produce more social value if they were employed in other capacities than shaving milliseconds off trading times and keeping bid and ask prices within a penny of stocks' momentary prices.

Other ways to reform trading
Depending on what those findings reveal, additional changes might be necessary, ranging from tweaks and circuit breakers to more serious measures.

If predatory trading practices show up in the data, the SEC ought step up enforcement against price manipulation and consider banning other predatory practices like order anticipation. The European Union issued a draft guideline that would require exchanges to implement procedures for reporting quote stuffing, momentum ignition, and other tactics to the proper authorities.

If we discover that the exploding volume of quotes results mostly from the time lag between exchanges and HFTs, as Narang suggested, infrastructure could be upgraded to remove the lag, or the SEC could tweak the rule that requires exchanges to verify the best price before they allow trades to take place. If quote spamming turns out to be a big problem, exchanges could try to more effectively target offenders.

Those who strongly oppose HFT have suggested placing a speed limit on trading, increasing the minimum (penny) increment for stock prices, or implementing a small tax on financial transactions. Each would likely result in wider spreads between exchanges and between bid-ask prices, and so are unlikely to be implemented unless the current market structure is deemed to be too risky.

For the time being, long-term investors don't need to be overly worried about the changing face of trading. But these are the sorts of things the people who operate stock markets need to think about. We'll be sending a team to the upcoming SEC roundtable on high frequency trading and will report back on emerging developments to Fool.com. This issue is not going away.

To stay up-to-speed on financial reform and investor advocacy, simply shoot a blank email to imoscovitz@fool.com.

« Previous
The Real Story About Investors and Trading Fiascoes

The article Supercomputer Stock-Trading Robots: What's Overblown and What Needs to Be Fixed originally appeared on Fool.com.

Ilan Moscovitzdoesn't own shares of any companies mentioned. The Motley Fool has adisclosure policy. We Fools may not all hold the same opinions, but we all believe thatconsidering a diverse range of insightsmakes us better investors. Try any of our Foolish newsletter servicesfree for 30 days.

Copyright © 1995 - 2012 The Motley Fool, LLC. All rights reserved. The Motley Fool has a disclosure policy.