Algos Gone Wild: Stock Trading Glitches Strike Again

Algos Gone Wild: Stock Trading Glitches Strike Again


Everyone knew that yesterday would be a busy, even frenetic, day for financial markets around the world. The Federal Reserve was scheduled to issue its latest verdict on interest rate policies, with economists holding their collective breath in hopes of a third round of quantitative easing. There was a cluster of fresh economic data due out. And today is the much-anticipated meeting of the European Central Bank.

Still, while lots of traders were anticipating volatility, nobody expected anything like the kind of gyrations seen in as many as 148 different stocks, with blue-chip names like Verizon (VZ) and Dole Food (DOLE) among those hit with massive trading volumes. Dramatic gaps emerged between “bid” and “ask” prices, while some stocks briefly plunged 10 percent or more.

For most of the market, the mayhem was all over in less than an hour – except, that is, for Knight Capital (KCG), the trading firm in whose direction fingers quickly began to point as being responsible for the erratic trading. By the ended of the day, Knight’s own stock had nosedived nearly 33 percent, as some authorities investigating the extreme volatility appeared to be attributing it to technology glitches with one of its algorithmic trading programs.

No kidding. Experienced traders had figured that out within a few seconds of the wonky chart patterns showing up on the screens in front of them. What is harder for them to understand is why – two years after the “Flash Crash” of 2010 – this kind of market turmoil still rears its head. Of course, few of the trading events like this make it into the headlines the way the “Flash Crash” did, so regulators aren’t under tremendous pressure from outside Wall Street to come to grips with the impact of algorithmic programs on the very nature of trading in the stock market. But within the market? Well, what started as irritation and impatience is morphing into anger and even outrage.

RELATED: SEC Cracks Down to Prevent Another 'Flash Crash'

This time around – insofar as it’s possible to piece together what happened – it seems as if the computers ran amok, sticking Knight not only with damage to its reputation on Wall Street but with a loss that some have estimated could be as high as $300 million. That could be wildly inaccurate, but it does seem – given that trading is a zero-sum game – as if this time human traders profited from the mayhem. And, as defenders of high-frequency trading were quick to point out, the problem was nowhere near as acute as the “Flash Crash.” That’s certainly true: Even if Knight’s losses are as large as $300 million, that’s a drop in the bucket when set beside the $862 billion that was (temporarily) wiped off the value of the U.S. stock market in 2010. High-frequency trading systems and the algorithms they use, these advocates argue, add liquidity to the market, which is a Good Thing.

Well, not really. Not it results in a major crisis of the kind we saw two years ago and a slew of smaller trading anomalies, day after day, week after week, month after month on top of that. Less than two weeks ago for instance, traders reported seeing a bizarre “sawtooth” pattern of trading in a handful of large-cap stocks, including Coca-Cola (KO) and Apple (AAPL). Their prices swung higher and lower with an uncanny degree of synchronicity, zooming higher every hour on the half-hour, and lower once more thirty minutes later. More algorithms, traders muttered gloomily to one another.

It’s not that they mind the computerized trading, the human traders insist. It’s just that the two systems, human and cyber, don’t have to play by the same rules. “If a human being tried to do what these systems routinely do – test the market to see how people will respond by submitting orders and then pulling them back in milliseconds – then the SEC would be all over them” for violating trading rules, says one veteran trader and portfolio manager.

Like many of his peers, this trader doesn’t want to ban algorithms and computers from trading. He just wants to reduce the probability that improper or bizarre trades end up distorting the market. This particular trader suggests that all that might be needed is to ensure that any algo trade is a “real” trade, by requiring the system to keep it posted long enough for others to react, respond and – if they wish – transact. “Market participants will always behave like the players in any game, and push right to the limit of where the rules are,” the trader says. “That’s where the computers are right now – right on that line.”

Does it matter as long as these short-lived hiccups don’t produce another “Flash Crash”? Absolutely. Each time trading in even a single stock is obviously distorted in ways that have nothing to do with a real transaction (such as a big investor either taking a position or unloading shares), news (earnings releases, an analyst’s report, a lawsuit) or a broad market trend (as when economic data convinces investors that it’s time to lighten up their stock market holdings), it undermines confidence in the integrity of the price discovery function that is at the heart of the market. And that means that it undermines confidence in the market itself.

Just imagine, for a second, that the algo in question had gone temporarily insane not in the U.S. stock market, but in the European sovereign debt market. Assume, just for a nanosecond, that yields on Spanish or Italian debt had soared to 10 percent, 12 percent or 15 percent. Imagine how global investors would have reacted to that; how they would have responded by assuming that something fundamental had changed and that the entire European economy was about to collapse.

The potential for trouble is almost unimaginable – “almost” because some traders already fear what might happen in just such an eventuality. Simply because a hiccup in trading appears technical and short-lived, doesn’t mean it’s something that couldn’t have serious consequences for our financial markets – and it’s certainly not something regulators and policymakers can confine to the back burner for much longer.