This is the paper that led to a Fed Policy Change. One of Einstein's great contributions to mankind was the theory of relativity, which is based on the fact that there is a real limit on the speed of light. Information doesn't travel instantly, it is limited by the speed of light, which in a perfect setting is 186 miles (300km) per millisecond. This has been proven in countless scientific experiments over nearly a century of time. Light, or anything else, has never been found to go faster than 186 miles per millisecond. It is simply impossible to transmit information faster.
On August 1, 2012, starting at market open (9:30 EDT), our monitoring software alarms went off on hundreds of symbols. Looking closer at the data, we found and reported (at 9:38am) that many NYSE stocks had extremely high trade rates: several sustaining over 100 trades per second.
Links and commentary on an explosive paper that changed everyone's view of HFT. On November 14, 2012, Adam D. Clark-Joseph published Exploratory Trading, which analyzes CFTC audit level trading data in the eMini S&P 500 futures market. This is a special, "regulators-only" data-set that contains all orders and trades, and each order and trade has a trader identifier. What this paper exposes is astounding.
In June 2010, while analyzing the Flash Crash, we noticed that many stocks had extremely high rates of canceled orders (1000+ per stock, per second). We then looked at data back to 2004 and found hundreds of thousands of examples: the first beginning in July 2007, which not coincidentally is when High Frequency Trading (HFT) began exploiting the flaws of Reg. NMS.
Below are portions of a comment letter submitted by R.T. Leuchtkafer to the SEC on April 16, 2010, just 3 weeks before the flash crash. The second paragraph in the excerpt below, unknowingly describes exactly how the flash crash was started. The letter goes on to alert the SEC on the dangers of High Frequency Trading (HFT), phantom liquidity and other concerns.
Mentioned on The Daily Show by Jon Stewart. On October 1, 2012, we detected a new form of quote stuffing that tries to hide under the radar. It involves hundreds of stocks and millions of bogus quotes during the trading day. The algo that generates them is very careful to spread out the stocks targeted and limits each test to a blast of exactly 200 quotes in 25 milliseconds or less.
We have found many more examples of HFT algos running out of control since our recent discovery of algo tests occurring during active regular trading hours. We now know that there are thousands of similar events every trading day: we missed many of these because our filter parameters were set too high. In the example below, you can see how much it affects prices investors receive. HFT proponents would like you to believe the spread of CARB is 5 cents or less (the quiet period between 13:25 and 13:40) And they are right! The spread is narrow.. when there isn't any trading activity. But once there is a hint of trading activity, the BBO oscillates to such a degree that it has no meaning. It might as well be infinite.
The March 23, 2012 IPO of BATS was brief. Apple (AAPL) stock began trading at 11:14:18.436 with an initial price of $15.25. Within 900 milliseconds from opening the stock price had fallen to $0.2848. Within 1.5 seconds the price bottomed at $0.0002. 567 trades were executed before the stock was halted (532 are shown below).
During the market close in SPDR S&amp;P 500 ETF Trust (SPY) (and dozens of related ETFs) on February 29, 2012, a delay of about 1.25 seconds developed in quotes from Nasdaq relative to quotes from other exchanges. Execution trade prices from Nasdaq appeared synchronous with trade prices from other exchanges, which shouldn't be possible, since Nasdaq's quote was behind. We also noticed an anomaly in the trade condition.
On October 9, 2012 between 10:02:20 and 11:01:15, there were at least 449 bad prints in 253 stocks (view in google docs, or download it) which caused price spikes on charts. All bad trade prints were from NqTRF (dark pools) and ended in 0.xx40 or 0.xx60. At the end of the day, 437 bad trades were corrected with new prices, and 12 trades were outright canceled. Almost all of the corrected trades have new prices that differ from the original price by $0.444 which is bizarre.
On July 31, 2012, starting about 15:59:57 EDT (3 seconds before the close of equity trading) our monitoring software alerted us to an unusually large trade of over 60,000 eMini (ES September 2012) contracts. At the same time, we also received a series of alert showing large and unusual trades in ETFs.
On October 25, 2013, the Federal Reserve announced it would add an Internet Kill Switch (among other measures) to prevent future leaks of FOMC news announcements. We proved with market data that the September 18, 2013 Fed (FOMC no-taper) news release must have been secreted out of the Treasury Department and placed on computers in New York and Chicago trading centers sometime before the official 2 PM release from Washington, D.C.
We have found significant evidence that the stability of NBBO spreads has decreased since the implementation of Reg NMS in March 2007. Combing through over 0.7 trillion quote and trade records, we sorted stocks into bins by the changes in NBBO spread every second of the trading day. We then plotted the results. You can clearly see the decline in NBBO stability since 2007 in general, and during very active trading. An unstable NBBO is a indication of a drop in liquidity, or that visible liquidity might be an illusion.
From the 1990's through the end of 2006, electronic trading increased transparency which attracted liquidity, leading to to narrower spreads and greater market stability. The graphic in this recent New York Times article shows the steady reduction in trading costs up until about 2006. Since that time, trading costs have not changed significantly on the execution side but on the analysis side (not factored into the New York Times article), both the number of quotes and the cost of receiving and processing them has exploded.1 Spreads have also become considerably less stable.
On April 24th, 2012 at 15:51:44, the number of quotes for a single second in one stock set a new record: 47,138. The stock, PSS World Medical (symbol PSSI), is not active: only 1,992 trades (317,127 shares) traded the entire day. This event represents a clear and present danger to market stability, and needs to be investigated immediately. Due to the way stocks are processed, many other stocks were affected by this event.
It's not high frequency trading (HFT) that concerns us. It's high frequency quoting, and it should concern everyone. The two images below tell the story. The chart on the left shows the growth of high frequency quoting. The chart on the right shows the (lack of) growth of high frequency trading.
The SEC conclusion that the cause of the flash crash was sale of 75,000 emini contracts, is based on faulty logic and a lack of understanding how the algorithm that executed those contracts works. Our detailed study of the actual trade execution data for these 75,000 contracts immediately uncovered a major disconnect between the SEC report and what actually happened. Over the next year and a half, further analysis and discussions with people who knew exactly how the algorithm worked led to probing questions put towards those associated with the SEC report. On April 9, 2012, we finally found a discrepancy that not only nullifies the SEC conclusion, but questions the basic honesty of the research behind the report.
We have found significant evidence that overall NBBO spreads have not decreased since the implementation of Reg NMS in March 2007. Combing through over 0.7 trillion quote and trade records, we sorted stocks into bins by NBBO spread every second of the trading day. We then plotted the results and were surprised to find no evidence of tighter spreads. What is worse, we found that the NBBO spread has become less stable since 2007.