2010 Flash Crash

Investors that panicked that day, as the selling intensified, lost their shirts. While the panic was contagious, it originated in the most unlikely place, far from Wall Street. 

Waddell & Reed is a modestly sized asset manager based in a leafy suburb of Kansas City. Known for its Ivy brand of mutual funds. [...] Waddell & Reed decided to hedge its position in the stock market by selling futures on the S&P 500 index. [...] Waddell & Reed wanted to sell 75,000 futures contracts, together worth approximately $4.1 billion, before the market closed. This would be a massive undertaking. In the past, trades of this size had taken several hours to complete. So Barclays launched three different versions of the algorithm that ran in parallel. [...] The selling wouldn’t slow or stop no matter how low the price fell or how quickly it dropped. [...] In each minute, each of the three versions would attempt to sell 3 percent of the total volume traded during the previous minute, for a total of 9 percent of the previous minute’s volume in each subsequent minute. [Each time] using the previous minute’s volume as their yardstick. [...] A lot of volume would suggest to the algorithm that the market was very liquid, and it could release more sell orders without having the market drop too much under its own weight. Relatively light volume would tell the algorithm the market couldn’t handle much additional volume, and the algorithm would taper the number of contracts it attempted to sell. [It is said that] a market’s capacity to absorb selling disappears just when it’s needed most. [Here] as real liquidity dwindled, the Barclays algorithms were telling themselves that liquidity had nearly quadrupled.

Source: Scott Nations. A History of the United States in Five Crashes: Stock Market Meltdowns That Defined a Nation, HarperCollins, Jun 13, 2017. [x]

Waddell & Reed, a moderately large mutual fund founded in Kansas City [...] ordered his staff to hedge their exposure to the stock market by selling seventy-five thousand e-minis worth $4.1 billion. [...] Who puts in a $4.1bn order without a limit price? [...] Waddell & Reed’s traders opted for a variant that would sell e-minis at a rate of 9 percent of the total trading volume over the previous minute. The idea was that, as trading volumes rose and the market demonstrated greater capacity to absorb the order, selling would increase; and when the marketplace slowed down the algo would ease up. [...] They switched the program on at 1:32 p.m. [...] At 1:41 p.m., with nothing to prop it up, the e-mini started falling like a runaway elevator. If Waddell & Reed’s traders had been more responsive, they might have shut the program down. Instead, their volume-sensitive algo actually sped up since—in spite of the exodus—trading activity soared. [...] Waddell & Reed had sold $1.9 billion worth of e-minis. Over the next six minutes, as the market bounced back, it unloaded the remaining $2.2 billion. The entire trade, the biggest e-mini transaction of the year.

Source: Liam Vaughan. Flash Crash: A Trading Savant, a Global Manhunt, and the Most Mysterious Market Crash in History. Doubleday, 2020. [B084]

The E-Mini contract is valued at 50 times the S&P 500 Index. [...] One of the largest E-Mini sellers [the day of the Flash Crash] decided to hedge a $7 billion position in U.S. equities by selling short 75,000 E-Mini futures. It used an off-the-shelf algorithm to feed shares into the market at a pace that would keep them at 9% of the E-Mini market’s overall volume. The hedge was massive, worth a total of $4.1 billion. Waddell & Reed had made similar trades earlier in the year with no outsized impact on the market. But this was no ordinary day. If high-frequency traders had somehow detected the huge order in the market, using their sophisticated algorithms, they might have tried to front-run the order, selling ahead of Waddell & Reed and figuratively adding tons of snow to the avalanche. What Waddell & Reed could not possibly know that day was that buy-side liquidity in the E-Mini had fallen by 55%, from $6 billion to $2.65 billion, by the early afternoon. Buyers had begun moving to the sidelines. The volume appeared to be larger than it was because HFT firms using 15,000 accounts were trading E-Mini contracts back and forth, several thousand times a second, to generate rebates. In effect, [a] game of Ping-Pong. 

Owner and chairman of Tradebot Systems, [David] Cummings wrote, “Wow! Who puts in a $4.1 billion order without a limit price? The trader at Waddell & Reed showed historic incompetence.” [However] the 75,000 contracts represented 1.3% of the total E-Mini volume of 5.7 million contracts on May 6 and less than 9% of the volume during the time period in which the orders were executed. [...] Additionally, the aggregate size of this participant’s orders was not known to other market participants. In other words, none of the E-Mini traders could be panicked by a large order because there was no way of knowing there was a large order. [...] Only 35,000 of the trades had entered the E-Mini market during the 20 minutes of decline, half of the order.

Source: Jim McTague. Crapshoot Investing: How Tech-savvy Traders and Clueless Regulators Turned the Stock Market Into a Casino. United Kingdom, FT Press, 2011. [B181]

Whenever something weird happens in the market, people rightfully look at High-Frequency Trading (HFT).

[During] the Flash Crash of 2010, [...] a mutual fund (reportedly, Waddell & Reed Financial, Inc.) submitted a very large order to sell 75,000 e-mini S&P 500 futures (about $4.1 billion worth). Such a large order rarely hits the market; in fact, it had only happened twice over the previous 12 months. [...] HFTs initially provided liquidity. They were net buyers as the market was dropping, but, at 2:41 p.m., HFTs turned around and became net sellers, perhaps to reduce their inventory risk. [The CFTC and SEC report said afterwards that this chain of events:] “generated a “hot-potato” volume effect as the same positions were rapidly passed back and forth. Between 2:45:13 and 2:45:27, HFTs traded over 27,000 contracts, which accounted for about 49 percent of the total trading volume, while buying only about 200 additional contracts net.” [...] Liquidity dried up. [...] Some stocks experienced highly unusual trades as their limit order books were wiped out, and market orders started hitting “placeholder bids” at extreme prices, including a trade at $0.01 for Accenture. [...] The role of HFTs in the Flash Crash was not so much what they did but what they didn’t do, namely provide unlimited liquidity.

Source: Lasse Heje Pedersen. Efficiently inefficient: how smart money invests and market prices are determined. Princeton University Press, 2015. [B114]

On May 6, 2010, the US stock market was hit by what has come to be known as the “Flash Crash” [after] an anxious mutual fund manager tried to hedge its portfolio by selling $4.1 billion in futures contracts. […] This seemingly prudent transaction somehow triggered the computers. The computers bought many of the futures contracts the fund was selling, then sold them seconds later because they don’t like to hold positions for very long. Futures prices started falling and the computers decided to buy and sell more heavily. The computers were provoked into a trading frenzy as they bought and sold futures contracts among themselves, like a hot potato being tossed from hand to hand.

Nobody knows exactly what unleashed the computers. Remember, even the people behind the computers don’t understand why their computers trade. In one 15-second interval, the computers traded 27,000 contracts among themselves, half the total trading volume, and ended up with a net purchase of only 200 contracts at the end of this 15-second madness. […] Some computers paid more than $100,000 a share for Apple, Hewlett-Packard, and Sotheby’s. Others sold Accenture and other major stocks for less than a penny a share. The computers had no common sense. They blindly bought and sold because that’s what their algorithms told them to do. The madness ended when a built-in safeguard in the futures market suspended all trading for five seconds. Incredibly, this five-second stabilization of prices was enough to persuade the computers to stop their frenzied trading. Fifteen minutes later, markets were back to normal and the temporary six hundred-point drop in the Dow was just a nightmarish memory.

Source: Gary Smith. Standard Deviations: Flawed Assumptions, Tortured Data, and Other Ways to Lie with Statistics. United States, Harry N. Abrams, 2015. [B038]

On May 6, the researchers found, there were a total of around fifteen thousand participants active in the e-mini. [...] Of these, they classified sixteen as HFT based on their attributes of trading huge volumes without ever accumulating a large position, and always ending the day close to flat. [...] This group of largely Chicago-based entities was responsible for 29 percent of the trading volume. [...] The most profitable of these firms mostly removed liquidity by aggressively hitting the resting offers of other participants just as the market was about to move, something academics refer to as “sniping.”

Source: Liam Vaughan. Flash Crash: A Trading Savant, a Global Manhunt, and the Most Mysterious Market Crash in History. Doubleday, 2020. [B084]

Postscript: According to research by Kirilenko et al. (2010), HFTs did not lose money during the Flash Crash. They may in fact have made more profits that day than on the previous days. 

Identifying a single root cause for a complicated event like the Flash Crash isn't easy. Some have pointed the finger at a very convenient scapegoat. 

IDT was one of a burgeoning number of arcades or “prop shops” sprouting up in Britain and the United States. The business model was straightforward and, for a while at least, highly lucrative. IDT would take on a bunch of wannabe traders and teach them the skills they needed to succeed in the markets. Those who thrived were backed with steadily larger sums, while those who failed were cut. Any profits the recruits made after paying a monthly desk fee of around $1,700 were split, with newbies retaining 50 percent and the most successful as much as 90  percent. IDT also creamed off a small sum on each trade, or “round-trip,” its traders placed, which quickly added up. The arrangement meant it didn’t matter if everyone in the stable was making money as long as they were all buying and selling and there were at least some big winners. As the owner of a rival arcade put it: “During the gold rush, it was usually the ones selling the spades who got rich.” 

Nav [real name: Navinder Sarao] was what’s commonly referred to as a “scalper”. [...] Scalpers analyze the ladder for clues as to whether prices will rise or fall. [A] ladder [is a display that] shows trades occurring and orders entering and leaving the market in real time. [...] At the end of almost every session, [Nav] made sure he had no outstanding positions—that he was “flat,” in the idiom of the trader. [...] He barely withdrew anything from his trading account to live on, preferring to let it accumulate like a high score. [...] Early in his career, Nav made a decision not to talk to his friends and family about his finances because he was worried they would treat him differently. [...] Nav [was] betting two hundred lots a night, a $15 million position that yielded six-figure profits. [Now he was trading] about 1% of the SP 500’s total daily volume. [...] At that size every tick, or 0.25-point move, of the S&P 500 was worth $6,250. [Eventually he earned] himself around $15 million. [Nav once made] $435,185 [in a few hours], more than the value of his parents’ house. [...] If Nav had worked at a bank or a hedge fund, there would have been a compliance officer to step in when he strayed into dodgy territory, but as an independent day trader he was on his own.

Chicago-based software vendor Trading Technologies [...] offered a product called Autotrader that allowed customers with no background in programming to create their own algos in Excel. [...] The “cancel if close” Function [was so good that] the U.S. authorities would later describe as a “layering algorithm”. When activated, it would allow Nav to place a number of large sell orders a designated number of ticks above the best offer. As the market moved higher and lower, Nav’s orders would move in lockstep, always maintaining the specified distance from the current price to minimize the chances of being hit. [...] The problem with spoofing as a strategy was that it was somewhat akin to picking up nickels in front of a steamroller. To alter the order book sufficiently to have an impact, you had to place big orders. But if a massive hedge fund or bank happened to come along at the wrong moment with a mandate to hoover up a billion dollars of e-minis and lifted all your resting offers, you’d be left watching in horror as the market jumped ten levels, costing you millions. [...] The solution he came up with was called “back of the book,” and it was designed to take advantage of the CME’s First In, First Out (FIFO) queuing system. [...] Nav’s brainwave was to have his algo add a single lot every time a fresh order arrived behind him, thereby constantly sending him to the back of the line and out of harm’s way.

Nav activated the “cancel if close” feature and placed four sell orders totaling 2,100 contracts one tick apart, starting three levels above the best offer of 1,163.25. They had a combined value of $120 million. Over the next six minutes, as the e-mini price fluctuated, these orders were automatically canceled and replaced 604 times to ensure they remained in lockstep and therefore unconsummated. [...] On this occasion he left the orders sitting there for more than two hours, [...] bringing the total value of his spoof offers to $200 million. [...] In that final two-hour spell alone, he’d bought and sold 62,077 e-mini contracts with a combined value of $3.4 billion. If, at any moment, the market had rallied, his entire account could have been wiped out. Instead, the e-mini fell 361 points, and he made a profit of $879,018. [...] One minute after Nav shut off the Autotrader program, at 1:41 p.m. CT, the e-mini started to plummet with a velocity and intensity it never had before. 

Did one of the five biggest S&P traders in the world really operate from a semidetached house in suburbia? [...] In the twelve days, [NAV’s] layering algorithm canceled or modified orders 182,000 times, corresponding to $35 trillion in notional trades—double the size of America’s gross domestic product. On eight of those days, not a single one of those orders was hit. The size of the orders was also immense: an average of 504 contracts, where the average across the market was seven. On the day of the crash, the layering algo accounted for close to a third of all canceled trades in the e-mini.

The value of a single e-mini contract—the minimum one can wager— is calculated by taking the current value of the S&P 500 and multiplying it by $50. In mid-2007, when the S&P 500 was trading at around 1,500, a single contract, or “lot,” was worth $75,000. [...] Even in extremely volatile conditions, the S&P moves around by only a few percent in a day, so rather than requiring their customers to put down the total size of their position, they ask for a smaller sum, known as “margin,” which is calculated to cover any potential losses. [...] Trading volumes in the [S&P 500 “e-mini”] contract far exceed the amount of buying and selling that goes on in the underlying stocks. 

At exactly 1:45 and 28 seconds, the e-mini ladder froze. The CME’s “stop-logic” function had kicked in after the rate of the fall had breached a set level and, for five long seconds, no trading took place. [...] The Dow had fallen more in a five-minute period than at any other time in its 114-year history. When trading resumed, the e-mini started to climb as rapidly and miraculously as it had tanked. [...] Fueling the exodus was a technical issue at the biggest stock market of all, the New York Stock Exchange, which happened to be upgrading its IT system that day.

Accenture sold for a solitary cent. At the other end of the spectrum, Apple and auctioneer Sotheby’s both transacted at $100,000 a share. [...] The mystery of why some shares changed hands for less than a cent and others for $100,000 came down to an arcane rule that obliged market makers in a given stock to provide quotes at all times, regardless of conditions. The stipulation was supposed to help prevent events like the Flash Crash by guaranteeing there was always somebody willing to trade, but, as ever on Wall Street, some dealers had found a way around it by leaving extremely high or low quotes in the order book. Before May 6, 2010, it had seemed inconceivable that these so-called “stub quotes” would ever be hit. 

In response to the clamor, the government did what it often does and formed a committee. [...] “A temporary $1 trillion drop in market value is an unacceptable consequence of a software glitch,” they wrote. [...] Even seven months on, the SEC didn’t have the data it would have required to understand what happened to equity markets over a single half-hour period at a granular level. If the umpires couldn’t watch the game, how could they hope to referee it?

FINRA had struck a deal with the exchanges to cancel any trades that occurred more than 60 percent away from their price before the crash started. Around twenty thousand mostly equities trades were scrapped, but not all transactions met the threshold, resulting in some big losers. [But] FINRA’s erroneous trade agreement didn’t cover derivatives.

[Nav] was the first alleged market manipulator ever to be extradited, and some of the counts against him carried sentences of up to twenty years in prison.

Source: Liam Vaughan. Flash Crash: A Trading Savant, a Global Manhunt, and the Most Mysterious Market Crash in History. Doubleday, 2020.  [B084]

Postscript: Nav was found guilty and sentenced to 1 year house arrest (i.e. home detention) in a Chicago court in 2020. At trial, it was revealed that Nav suffers from severe Asperger. Nav's lawyers also claim that their client lost all the money he made as a trader (about $70 million).

[It was] alleged that Sarao had attempted to manipulate the price of E- Mini S&P 500 futures contracts on the Chicago Mercantile Exchange, a side effect of which was the Flash Crash. On November 9, 2016, Mr. Sarao pled guilty to one count of wire fraud and one count of “spoofing” (a form of price manipulation). [...] Since then, we’ve experienced flash crashes in U.S. Treasury securities (October 14, 2014), foreign currencies (March 18, 2015), and exchange-traded funds (August 24, 2015). [...] And the two- and- a- half hour Bloomberg terminal outage (April 17, 2015) that postponed a multi- billion- dollar government debt issue.

Source: Andrew W. Lo. Adaptive markets: Financial evolution at the speed of thought. Princeton University Press, 2017. [B116]

Most have described the series of events on that day as a snowball or a hot-potato. What came first, the S&P 500 ETF, or one of its more volatile constituents that day? It's probably the former, but once the vicious cycle starts going, it's hard to untangle the two incidents.

At 2:40, a wave of sell orders for Procter & Gamble, the consumer products giant, hit the floor of the NYSE. The market for P&G suddenly became unbalanced. There were far more sell orders than buy orders. A computer system to slow trading at the NYSE kicked into gear, routing orders to designated market makers on the floor. That slowdown created another problem. Because traders couldn't sell on the NYSE, they routed their orders to other markets, such as Nasdaq. Waves of orders sloshed away from the NYSE to other corners of a rapidly fragmenting market. In a matter of moments, P&G shares collapsed, losing 35 percent of their value. Across Wall Street, hundreds of trades started to occur at unheard-of price levels. Accenture, a global consulting company that normally traded for about $50 a share, swapped hands for a penny a share at 2:47.53. Boston Beer, the maker of Sam Adams brews, also hit a penny. Cigarette giant Philip Morris tumbled from $49 to $17. […] In total, roughly one trillion in assets vanished from the market like a conjurer's trick. At the other end of the scale, Apple, which normally traded for about $250, sold for nearly $100,000 a share. The explanation for this insanity went back to Nasdaq's integration of algo trading. Years before, high-speed market makers operating on Nasdaq were told that they always needed to stay in the market. But there was a loophole: They didn't need to post bids or offers close to the price of a stock. Instead, they could put up wildly wide quotes, such as an offer to buy for a penny or sell for $99,999. The trick allowed the firms to stay in the market without actually trading.

Just before 2:43 Eastern time, a massive sell order for several thousand E-mini contracts hit the tape, eating right through the order book. At the exact same moment in New York, a massive wave of ETFs mirroring the S&P 500, the Nasdaq index, and the Dow industrials were sold. […] Citadel was one of the most prolific traders of NYSE and Nasdaq stocks in the world. It also executed many of the trade orders from retail brokers such as E*Trade and TD Ameritrade […] Since internalizers matched buy and sell orders from clients "internally," rather than route them to an exchange, it was flow the exchanges rarely saw. But at about 2:45, Citadel asked all clients to route their orders away from its computers due to a technical glitch. […] Suddenly, a wave of retail order exhaust flushed away from Citadel and into already overwhelmed exchanges such as Nasdaq and the NYSE. [Then] a trigger snapped and the exchange's so-called Stop Logic Functionality halted trading in the contract for one, two, three, four, five seconds. Those few seconds gave the high-speed Bots the time they needed to catch their breath. […]The feedback loop was broken. The machines regrouped. Instead of selling, they started buying. The market recovered-and then it surged.

Source: Scott Patterson. Dark pools: The rise of the machine traders and the rigging of the US stock market. Crown Business, 2013. [B082]

Remember, market orders can be dangerous! 

[A] large imbalance between buyers and sellers sent prices rushing downward and triggered CME’s stop logic functionality, pausing E-Mini trading for 5 seconds. [...] Many of the algorithms had instructions to liquidate all positions and exit the market in the event of unanticipated trading patterns. That’s what many of the machines began to do, draining liquidity from the market. Sell volume was so unusually high that, absent the traders, the exchanges could not handle them all.

“The data coming back from the exchanges was garbage,” said one trader. In some instances, the information about trade volume and prices was 20 seconds old. Rather than risk making erroneous trades based on the faulty market data, these traders simply shut down their computers. Part of the problem was that the NYSE was upgrading quote systems for its listed stocks, and the work was only half done. [...] The time differential on May 6 sometimes gave HFT firms a 35-second informational advantage. [...] The tape delay had given them a head start to the exit.

There was also some highly circumstantial evidence that some high-frequency traders deliberately had flooded the market with trades to slow down the consolidated tape. These traders were among those who had the special feeds from the exchanges to their collocated servers that provided them with an advance peek at actual pricing data.

Some large players left the market anywhere from 5 minutes to 15 minutes, whereas others left for just 30 seconds. The bottom line was that stock prices fell like an avalanche because few people wanted them at that moment. Initially, the pickup in velocity was due to “stop loss” orders. These are like ejection seat buttons in a jet fighter. If the jet is falling to earth, the pilot ejects to save his life. Some investors place stop loss orders below the current price of a holding, to lock in a profit. [...] Regulators later theorized that stop loss orders triggered a chain reaction. [...] Traders, it seems, overlooked their “stub” quotes. Under the SEC’s rules, some market makers and exchange specialists were required to constantly post prices on both sides of the market for the securities they specialize in. A lazy man’s method of complying with the annoying regulation was to post one bid price and one asked price so far out of the money that under normal circumstances, they never, ever would be “hit” by an actual trade. A buy order might be listed at a penny and a sell order at $10,000.

Brokers stopped automatic execution of customer orders, also known as internalization, which on most days accounts for nearly 100% of retail trades. [...] The damage to retail investors was substantial. Twenty thousand trades, totaling 5.5 million shares, were executed at a price 60% or more away from pre-Flash Crash price levels, and thus later were deemed invalid. At least half of those were retail orders. And, of course, that says nothing of the countless trades done at discounts of less than 60% but still large. [...] The 60% threshold was the result of an arbitrary decision. More than 20,000 trades had fallen from 20% to 59% and were left to stand.

It was also tough luck for savvy traders who had bought Accenture at a penny, realizing it was a steal, and had quickly sold it when it climbed back above $30-per share. They were now left with a short sale. They had to go back into the market and buy Accenture so they could deliver it to the purchaser, and they ended up paying more for the stock than the amount for which they had sold it!

According to the SEC, some 200 stock equity issues actually traded at 0, including stalwarts like Exelon, Boston Beer Company, and CenterPoint Energy. [...] There were some equally wild rises on the upside, with both Sotheby and Apple shares hitting $99,999 at one point, presumably because they hit stub quotes in the bid side of some trader’s book. [...] Those penny stub quotes were legitimate because at the time of the trades they had met the definition of a market order, which was “the best available price at the time of execution.”

Following the Flash Crash in May, investors immediately yanked out $25 billion from equity funds. [...] Everyone wanted to finger a culprit because the market had failed dismally, if only for 20 minutes. [...] The staffs of the two agencies, [the SEC and CFTC,] cranked out an 80- page report on the Flash Crash by May 18—a period of 12 days. [Yet] the SEC’s staff for the most part didn’t spend a lot of time in the market trenches and was surprisingly ignorant about the market system’s inner workings. 

Source: Jim McTague. Crapshoot Investing: How Tech-savvy Traders and Clueless Regulators Turned the Stock Market Into a Casino. United Kingdom, FT Press, 2011. [B181]

No wonder it apparently took the SEC 5 months to rebuild the consolidated tape and create an audit trail for May 6, 2010.

At the time of the Flash Crash, there were 11 registered exchanges exclusively for stocks. [...] There were more than 70 Alternative Trading Systems (ATSs) that were registered as broker-dealers and matched orders of buyers and sellers for all sorts of securities. And there were 5 electronic communications networks (ECNs), which were essentially ATSs that elected to display their best buy and sell orders on the consolidated quote so all investors could see their quotes. Thirty of the ATSs were so-called “dark pools” that did not trade in the “lit market” and thus did not display their quotes. Large institutions such as mutual funds and pension funds that want to sell large blocks of a particular stock—say 150,000 shares—would not want to put such a large order into the public market, where it would be displayed on the consolidated tape because the sudden appearance of such a huge supply would cause the price to drop sharply before it could unload all its shares. Dark pools attempted to cross these big orders by finding other big traders or by breaking the big order into smaller ones and dribbling the shares into the market. In the over-the-counter market, broker-dealers “internalized” trades, matching one customer’s order with that of another customer. The buys and sells, which are not publicly disclosed to the market, represented 17.5% of the total shares traded in the National Market System. [There were also] fourteen designated Control Markets were in operation in the commodities marketplace on May 6. Six of them were selling products that were so tightly correlated to the equities market that price direction on the venues moved in virtual lockstep. At one time, the commodity exchanges were associated strictly with agricultural products like pork bellies and hard assets like gold and other precious metals. By 2010, they also were offering an array of equities products such as those based on major stock indices like the S&P 500 and the Dow Jones Industrial Average (DJIA).

Each exchange had its own fail-safe system. [...] The NYSE later would report that prices at its exchange [...] did not reach the ridiculous levels reported by competitors who had no braking mechanisms at all for chaotic conditions. [Nevertheless] panicked investors merely sidestepped that exchange and placed their orders on one of the other 70-plus trading venues that comprised the national market system. [...] There was nothing untoward in the brokers avoiding the trading halt on the NYSE and taking their customer orders elsewhere. Market rules allow brokers to “trade through” exchanges where they cannot get a response to an order to buy or to sell, even if the exchanges are displaying the best bid and asked prices.

The problem was that the imposition of market-wide circuit breakers was little more than window dressing to get the Congress off of the regulators’ backs. Academic studies going back to at least the 1980s had cast doubt on the efficacy of the devices. [...] In practice, circuit breakers caused even more panic because sellers in search of liquidity were denied it.

Source: Jim McTague. Crapshoot Investing: How Tech-savvy Traders and Clueless Regulators Turned the Stock Market Into a Casino. United Kingdom, FT Press, 2011. [B181]

See Also: The August 2007 quant liquidation/meltdown.

The morning of May 6, 2010, SEC Chair was delivering a speech in which she said: “In the wake of the financial crisis, it has become a cliché that regulators cannot keep up with innovators in a market as dynamic as finance. I don’t think that’s true.” Five hours later, the Flash Crash happened and exposed the SEC's ignorance. The SEC had no choice to later come clean to the investing public:

The SEC explained that most of the delay [in investigating the Flash Crash] was because it did not have a consolidated data trail. [...] SEC Chair Schapiro was quoted as saying: “The idea that the regulator of the largest capital markets in the world cannot easily reconstruct trading when there has been a problem, or when there is a suspicion of manipulation or misconduct, is not acceptable to me.”  

A $4 billion order, one that had worked before without difficulty, and barely half of which actually traded on the way down, helped trigger $700 billion in losses in the Flash Crash. [...] Selling snowballed even more when “internalizers”— firms that normally buy and sell with their own customers instead of sending customer orders to the exchanges—“began routing most, if not all, of these orders directly to the public exchanges where they competed with other orders for immediately available, but dwindling, liquidity. [...] The rules allowed internalizers to decide when to trade against their customers or just throw their orders overboard to the exchanges—so they did. 

HFT market makers would shut down and run for the hills at the first sign of stress. Under duress, HFTs would not be liquidity providers. They would be demanders and consumers of liquidity the likes of which the public has never seen. [...]  The story of the Flash Crash is that the market failed that day. [...] Limit order books can quickly empty and prices can crash. 

A “liquidity vacuum" [pushed] Phillip Morris (MO) from $48 down to $17 and right back up to $46. [...] When Accenture traded at $0.01 during the Flash Crash, the bid that was hit was a stub quote. According to the SEC Flash Crash report, “Executions against stub quotes represented a significant proportion of broken trades on May 6.” Eliminating stub quotes was a no-brainer in our opinion. What is needed, however, are more stringent market maker obligations. Currently, market makers need to quote within only 8% of the NBBO for most securities. There are even some securities where they are allowed to quote within 30% of the NBBO.

Source: Sal Arnuk and Joseph Saluzzi. Broken markets: How high frequency trading and predatory practices on Wall Street are destroying investor confidence and your portfolio. FT Press, 2012. [B083]

In fact on May 6 only market-wide circuit breakers were operational. […] Those did not trigger on May 6, 2010. […] On 10 June, 2010 the SEC decided to remedy the problem [of absence of a stock-specific circuit breaker] by introducing a mechanism that places a five-minute hold on stocks with a movement of more than 10% during the previous five minutes. […] For less traded stocks and ETFs priced at or above USD1, a price move of 30% would trigger a pause, and 50% price move will trigger a pause for securities trading below USD1.

Source: Charles-Albert Lehalle and Sophie Laruelle. Market microstructure in practice. 2013. [B073]

The book Contrarian Markets concluded that "according to CFTC commissioner Scott O’Malia, three other such crashes that have occurred since the May 2010 flash crash have also been related to algorithmic trading". These events – and others like it – have eroded investor confidence. Let us not forget the lessons from that day as 2010 progressively fades in the rear view mirror!

The Flash Crash was a dramatic illustration of [market fluctuation] risk: Investors who submitted market orders wound up paying $100,000 a share for P&G and selling those same shares for a penny apiece. [...] Before the Flash Crash, 67 percent of U.S. households owned stocks; by the end of 2013, only 52 percent did: The fantastic post-crisis bull market was noteworthy for how many Americans elected not to participate in it.

In March 2012 the BATS exchange had to pull its own initial public offering because of “technical errors”, [a glitch later referred to as the BATS Crash]. The next month, the New York Stock Exchange canceled a bunch of trades by mistake because of a “technical glitch.” In May, Nasdaq bungled the initial public offering of shares in Facebook Inc. because, in essence, some investors who submitted orders to buy those shares changed their minds before the price was agreed upon—and certain Nasdaq computers couldn’t deal with the faster speeds at which other Nasdaq computers allowed the investors to change their minds, [an incident nicknamed the Faceplant]. In August 2012, the computers of the big HFT firm Knight Capital went berserk and made stock market trades that cost Knight $440 million and triggered the company’s fire sale, [in an event dubbed the Knightmare]. In November, the NYSE suffer what was termed a “matching engine outage” and was forced to halt trading in 216 stocks. Three weeks later, a Nasdaq employee clicked the wrong icon on his computer screen and stopped the public offering of shares in a company called WhiteHorse Finance. [...]  In April 2013, the price of Google’s shares fell from $796 to $775 in three-quarters of a second, for instance, and then rebounded to $793 in the next second. In May the U.S. utilities sector experienced a mini–flash crash, with stocks falling by 50 percent or more for a few seconds before bouncing back to their previous prices. These mini–flash crashes in individual stocks that now occurred routinely went largely unnoticed and unremarked upon.

Source: Michael Lewis. Flash boys: a Wall Street revolt. WW Norton & Company, 2014. [B026]

A particularly ironic example of a flash crash happened in the spring of 2012, when a company called BATS Global Markets attempted to go public. [...] The SEC granted BATS an exchange license in 2008. The company planned to list its shares on its own stock exchange and go public on March 23, 2012. What happened instead was a disaster [...] and forced BATS to rescind its IPO. [...] The first BATS trade was at $15.25 a share, but almost immediately market makers on NASDAQ quoted BATS at $10.98 a share, and then at $8.03, and then $4.17, and then ever lower at blazing speed. It took precisely 1.372 seconds for the price to crash all the way from $15.25 to a fraction of a penny a share, a textbook example of a mini flash crash.

Source: Sal Arnuk and Joseph Saluzzi. Broken markets: How high frequency trading and predatory practices on Wall Street are destroying investor confidence and your portfolio. FT Press, 2012. [B083]

In August 2010, it emerged that a single Chicago high-frequency firm, lnfinium Capital Management, had triggered a brief $1-per-gallon pop in oil prices due to a trading glitch that machine-gunned two to three thousand orders per second for an oil contract minutes before the market closed on February 3. Washington Post Company shares tumbled in the blink of an eye in June due to a computer-trading glitch, triggering new circuit breakers that had been installed to halt wild trades. In September, shares of Progress Energy, a North Carolina utility with eleven thousand employees, plunged almost 90 percent in a matter of seconds due to a glitch.

Source: Scott Patterson. Dark pools: The rise of the machine traders and the rigging of the US stock market. Crown Business, 2013. [B082]

[Since May 6, 2010] there was a series of mini-flash crashes in individual stocks. [For instance] on June 16, three orders totaling 766 shares pushed shares of the Washington Post up 99% in one second, from $462.84 to $929.18.

Source: Jim McTague. Crapshoot Investing: How Tech-savvy Traders and Clueless Regulators Turned the Stock Market Into a Casino. United Kingdom, FT Press, 2011. [B181]

Postscript: The markets pared back the losses and ended the day with a "mild" 3.2% decline.