Is Another Flash Crash Inevitable

¹²Is Another Flash Crash Inevitable By Year’s End And Will It Be Triggered On Purpose?
Or, Why Broker Entered Stop Losses Are Now Another Sure Way To Lose Out On Stocks.

By Christopher Steiner, Alpha Dogs | 3 September 2010

If we don't see a 'Flash Crash II', we'll certainly see events that mimic the quote volume spike of May 6–one of the factors that many people, inclduing Nanex CEO Eric Scott Hunsader, believe played a role in the short-lived collapse. In fact, adds Hunsader, somebody could be intentionally slowing down some aspects of the market— using excessive quote blasts— to skim profits from clueless competitors. This won't stop, Hunsader says, until the SEC or the exchanges step in and do something about the copious quote volume wars currently taking place.

Nanex, which sends its clients compressed real-time quote feeds and market data, has drawn attention lately for some of the underlying quote patterns it uncovered within the chaos of trading on May 6. But it wasn't odd patterns that drove the market to madness; it was pure volume. On May 6, somebody, Hunsader says, opened up a fire hose worth of quotes straight at the NYSE.

When the NYSE gets deluged, Hunsader says, their quote feed to the Consolidated Quote System (CQS), which is used as part of the National Best Bid and Offer (NBBO) system, gets delayed. The NBBO system requires trades to be executed at the best bid or offer possible, whatever exchange that might be on. NYSE's feed to its subscribers of Open Book, a premium, real-time quote stream, does not slow down, however.

When I visited Hunsader at Nanex's Winnetka, Ill. offices on Friday, he plotted out quotes for various securities on Open Book versus quotes from CQS, showing a clear delay in the latter system on May 6— a delay that could be measured in multiples of seconds— an eternity in the trading world. What's more troubling is that the system has shown big delays more than once since then. It seems that whenever the NYSE receives more than 20,000 quotes per second, its CQS feed, which determines where many equity orders get routed, falls behind. Here at Nanex's site, you can see, plotted, the CQS delay for the NYSE quote of GE on July 21 vs. that of Open Book.

The average total quote volume on the NYSE is 10,000 per second. Things went way past that during the Flash Crash. At one point on May 6, somebody launched 5,000 quotes at the NYSE for the single ticker of Public Storage inside of one second. None of those quotes led to a trade— and that traffic by itself took the NYSE to 25% of its stable CQS capacity, according to Hunsader. Introduce a few more salvos like that, and the NYSE's computers groan to keep up and you get wildly swinging trades for heavyweights such as Apple, Accenture, Exelon and Procter & Gamble.

High frequency traders' algorithms saw the quote disparity between CQS, Open Book and the rest of the market. They knew something was wrong. What other choice did they have but to pull out? (Just as Kansas City's Tradebot and New Jersey's Tradeworx did.) Those moves robbed the markets of liquidity right when they needed it. But how could the algo traders stay in when they knew the data feeds were badly broken?

The better question: who was flooding the system with quotes and why?

Hunsader's answer: it's clear that one trader or perhaps more discovered that by blasting the NYSE, they could introduce added latency in the CQS feed. Knowing that most players were looking at a delayed NYSE feed, anybody in the know could make easy arbitrage plays between the NYSE and other exchanges. "I think somebody figured this out a year ago and they've been playing with it ever since," he explains.

The explanation holds merit. It turns out that several mini flash crashes occurred before May 6. In almost all cases, the incidents started and ended inside of a second, which is why market observers didn't notice. During that second, though, a lot happens. In one repetitive incident, the price of a blue chip stock drops, say, 20 cents after several big sell orders comes through.

Those using the CQS feed, however, don't see the drop at the NYSE when they should, so they may be posting new bids at or near the old price, which an enlightened trader's algorithm— with access to an Open Book feed— seizes on. Inside of that second, 250,000 shares could trade hands, netting whoever capitalized $50,000 or so. On April 28, for instance, the share prices of Wal-Mart and Procter dipped 50 cents for less than a second in what Hunsader calls a mini Flash Crash. If algorithms had been programmed knowing the dip was coming, as Hunsader expects, profits are fat and easy.

And what if this happened across 70 stocks on one day, which Hunsader says happened? Now it's real money— perhaps more than $50 million in profits, virtually risk-free. Hunsader, in addition to enjoying his role as detective, is trying to lighten the workload for his firm, which has to carry all of this added noise— the excessive bids— on his network to his clients. Uncompressed, his data feed now amounts to 500 GB a day.

Before long, he reckons, many people will have to upgrade their data lines to handle the exploding quote volume on the financial markets. Consider that in 2004, our equity markets saw one quote for every 10 shares traded. Now the ratio is one to one, a ten-fold increase. CQS in 2004 carried 25 million quotes per day. Now it's at 800 million a day and growing. Just in 2010, in fact, volumes have doubled.

"There is no extra trade volume here," Hunsader says, "it's just noise."