Embedded Files
Quote Rate Pirates are Attacking the Option Market
A bigger May 6th is coming
This is a research project in progress, areas of this are unfinished, partially researched and may be speculatively assumed. But the findings are such that it is important to make them available to those that might be able to help stop what is obviously the potential demise of the option market…
For the option market, like any market to be fair, there has to be equal access to the known data. Governmental financial regulations and the SEC have already mandated this. It is this study’s purpose to demonstrate that a fair option market is in jeopardy under the current course of action by at least two option exchanges.
Some Definitions:
High-Frequency Trading (HFT) is an algorithmic based, computer generated, trading style that has added competition and liquidity to the equity markets. HFT is legal and generally beneficial to the market, insuring tighter bid/ask spreads and plenty of market volume. But, there are other computer generated activities that are counter productive to the markets and not legal. One of these is called “Quote Stuffing” or sometimes referred to as chatter or flicker. Quote Stuffing is the systematic entry, removal and replacement of quotes (bids to buy, offers to sell, and size of bid or offer) into the market at a very rapid pace.
Allowed to go unchecked, the excessive quote activity caused by Quote Stuffing could potentially cause what computer geeks would know as a denial-of-service attack (DoS) on the computer resources of the market’s quote distribution system (data-feed). Stuffing too many bogus quotes into the data stream of legitimate quotes can cause the data-feed to fall behind. The gap in time between when the quote is generated and the time that it is actually posted by the data-feed to the public is often referred to as latency. A quote is said to be “Stale” if it is not the current and most accurate quote.
Nanex has demonstrated in its study of the equities market “Flash Crash” of May 6th that there was a huge data-feed latency potentially caused by “Quote Stuffing.”
http://www.nanex.net/20100506/FlashCrashAnalysis_Intro.html and
http://www.nanex.net/FlashCrash/FlashCrashAnalysis.html
In a CFTC Technology Advisory Committee meeting studying the causes of the May 6th Flash Crash, an individual referred to the Quote Stuffers as “Quote Rate Pirates” dubbed after the old “SOES Bandits” slang used for level II traders of the late 1990’s - thus the title of this paper.
Quote Stuffing in the Options Market:
David Koran, a programmer at Nanex, and I have created a program that is able to monitor the number of quotes that each option exchange is posting on each option symbol every second of the day. With this program we are able to filter for those option symbols that have an excessive number of quotes per second (referred to as “quote count” or “quote rate” throughout this paper) at the exchange level.
To give the reader a scope of this project, there are roughly 350,000 option symbols coming from over 3,300 stocks. There are 8 option exchanges that potentially trade these symbols. Each option symbol has a bid and ask from each exchange as well as other trade information.
Now suppose a Quote Rate Pirate on just one of these exchanges decided to start placing and changing (“stuffing”) quotes on each of the option symbols at a rate of 100 per second. That would result in 35 Million messages per second. Currently a message from OPRA (the central clearing house for option quotes) uses 66 bytes. That would be 2,310 megabytes of data per second from just one option exchange.
There are 10 bits (start bit + 8 data bits + stop bit) in a byte. 2,310 megabytes(MBps) of data is 23,100 megabits(Mbps). A typical DSL line is 3Mbps; a typical cable connection is 12Mbps… The fastest network line is an OC-768 which is 1.33Gb or 1333Mbps. It would take 18 of these OC-768 lines to every entity that wanted to trade real-time options to handle the potential data from just this one exchange should the scenario be allowed to happen. The point is that trying to transmit this much data is not possible (or at least impractical) and will only lead to a disaster.
Another way to look at this is there are (60 x 60 x 6.5) 23,400 seconds in a trading day. If we had 100 updates per second on all options we would have (23,400 x 350,000 x 100) 819,000,000,000 (819 Billion) quotes per day. That is 150 times the May 6th peak of the sum of all three data-feeds as shown in the Nanex chart...
http://www.nanex.net/FlashCrash/QuoteRates.html
Note: This Nanex chart showing Quotes per Day will be referred to several times throughout this research project. The chart is designed to show general dates when peaks in quote rates occurred. The scale for OPRA data (the red line) is on the left where the scale for the other two data sources is on the right. This two scaled approach is confusing for our research project. It is important for the reader to realize that if the data for all three were put on the same scale, the OPRA data peak on May 6th would be over 5 times the height of the blue CQS data source.
OPRA is projecting maximums of 4 to 5 Million quote messages per second for the year 2011-2012:
http://www.opradata.com/specs/Upd_Traffic_Proj_2011_2012.pdf
The above calculation of 35 Million messages per second is 7 to 8 times OPRA’s projections. OPRA is projecting peaks of 15.5 to 20.7 Billion messages per day. The above calculation of 819 Billion messages per day is 40+ times the OPRA projections.
Again remember the above calculations were based on the potential of just one of the eight (and soon to be 9) exchanges. Hopefully, this is enough information to make the reader realize at 100 quote updates per second a disaster is inevitable.
Quote Rate in Perspective:
To help put 100 quote changes per second into perspective, think about the following examples. A humming bird flaps its wings 18 to 78 times per second. The average computer monitor has a refresh rate of 60 Hz which means it reprints the screen 60 times per second (at that rate the eye can not even detect a flicker). A high end TV probably has a 100 Hz rate. Click on this link and you can hear what 100 Hz sounds like
http://www.fearofphysics.com/Sound/100hz.mp3
Data can only travel at the speed of light (186,282 miles per second), and the distance from New York to San Francisco is approximately 2,582 miles. A piece of information can get to San Francisco in no less than .013861 seconds (or approximately 14 ms). This assumes a direct straight line connection, which is not reasonable. The actual time is probably more like 25ms. Using the assumed 100 quotes per second would imply broadcasting one quote every 10ms. The quote is not even valid by the time it reaches San Francisco. In fact the research has shown that when activated the chatter from quote stuffing is averaging over 200 quotes per second with peaks above 2,000 at the PHLX. That would mean that the quote and the following 27 would be “stale” by the time they reach the West Coast.
The point is that 100 quotes per second is a ridiculous number of times to update an option symbol per second. The price of an option is based on the price of an underlying stock or index. Because options are leveraged instruments and are typically traded in nickels and dimes, the rate of change in the price of a stock very seldom, if ever, warrants the updating of prices anywhere near 100 quotes per second.
What does Quote Stuffing in the option market look like?
This 10 second clip of the activity of part of the PHLX’s index XEV’s option chain captures quote stuffing behavior. The resolution of time on the capture program is not fast enough to capture each change, but it updates enough to display the problem. Note: this is actual speed, not a fast forward…
http://www.nanex.net/Dave/iXEVOption.htm
The clip starts at 14:35:29 on June 22nd. Notice at the start the 111.5 put ask price is in the process of counting up from 0.21 to 5.00. Once it reaches 5.00 it resets to 0.21. Then the 114 put ask price starts counting up from 0.22 to 5.00 (it actually increments in .01 steps, but the capture program is not fast enough to display all of the updates). Then the clip ends with the 121.50 put ask price starting a similar count up.
There is no trading purpose to activity like this. But, it will be shown that it goes on continuously in the option market throughout the day, especially at the PHLX and NQEX exchanges.
What is Legal and What is not?
Financial regulations dating back to 1975 require that there be a fee based public data source (Consolidated Data Stream) for both the stock and option markets. For years as can be seen in this 2005 SEC Commission ruling
http://www.sec.gov/rules/final/34-51808.pdf
it has been debated and accepted that exchanges can offer additional data sources provided “the integrity and reliability of the consolidated data stream must not be compromised…”
In the equities market the mandated consolidated data stream for all NYSE listed stocks is called the Consolidated Quote System (CQS). In the option market it is OPRA.
Under the SEC Commission Regulation NMS ruling most if not all exchanges also have other data-feeds (“Direct Connection Feeds”) that entities can connect to for additional fees. Again, these are allowed provided they do not compromise the integrity and reliability of the CQS and OPRA feeds.
A New York located trader connected to a data-feed may have as much as a 50ms advantage over a trader in San Francisco connected to the same data-feed using the same trading parameters. (Based on the assumption that it takes 25ms for the data to reach SF, the same amount of time for both traders to process the data, and 25ms for the trader’s orders to get back to New York.) This is a perfectly legal advantage and traders that take advantage of this type of lag are typically “Latency Arbitrage” type traders. This is the justification for many of the co-located computers in and around New York.
Under the “integrity and reliability” requirements above, a New York trader connected to a NYSE Direct Connection Feed must not have an advantage over another New York trader connected to the CQS feed. The Nanex website demonstrates that there was a Direct Connect Feed advantage on May 6th and that this has been occurring quite frequently for some time. It is hoped that the SEC Commission on the Flash Crash will address this issue.
The Nanex study of May 6th demonstrated that it is possible that Quote Stuffing contributed to the lag between the two NYSE data-feeds. If this is the case the Quote Rate Pirates associated with the Quote Stuffing should be found and prosecuted.
More about OPRA:
The Option Price Reporting Authority (OPRA) data-feed is the CQS of the option market. As such it has the same mandates of integrity and reliability.
OPRA’s responsibility is to receive all the information from the 8 option exchanges and disseminate it to the subscribers. There is some unavoidable lag caused by processing and transmitting the data to a central location. Certainly, there are tolerance levels for this lag that allow OPRA data to meet the integrity and reliability mandates. I do not know if the Option Exchanges are delaying their Direct Connection Feeds to account for this lag, and if not, if this a violation of the “integrity and reliability” mandate.
But, what happens if OPRA is flooded with useless data that it has to process? It would be reasonable to assume that an entity with a Direct Connection Feed to an exchange would certainly be able to profit from any gap caused by the flood of data. This becomes an inappropriate or illegal Latency arbitrage! This arbitrage is particularly illegal, if the trader is associated with the cause of the lag in the first place.
I have discussed this issue with Joe Corrigan, the Executive Director of OPRA, his attitude is that OPRA operates at the direction of the 8 option exchanges. OPRA is paid by the exchanges for the amount of data it handles. OPRA’s solution to more data is to just create a bigger pipe and require the customers using the data to follow suit. OPRA does not care what manipulations are going on at the Exchange level.
The remainder of this paper will demonstrate that Quote Stuffing is occurring in the option market. Hopefully, it will encourage OPRA, the 8 Option Exchanges, and if necessary the SEC to take action to resolve the issue.
The Original Research:
The original research was done using 100 quote changes per second as a filter. It was found that high quote rates actually started before the beginning of this year. By the first trading day in 2010 there were already over 18,000 occurrences of Quote Stuffing a day, by May 6th the there were over 60,000 per day, by June the occurrences were as high as 380,000 per day. Over 90% of the activity was found to be at the PHLX Option Exchange with the remainder believed to be at the NASDAQ (NQEX) Option Exchange (both owned by NASDAQ). The research was only done on the OPRA best bid/ask and not on the activity that did not affect the best prices.
This research resulted in a complaint to the SEC and PHLX in late June. The complaint focused not only on the potential May 6th like attack on the option market as a financial event, but also as a potential terrorist event. The information was kept relatively quiet because the assumption was that the SEC and PHLX would want to find the “Quote Rate Pirates” before they knew they were being investigated.
The PHLX/SEC investigations probably resulted in the $2.10 cancellation fee rule that was implemented by the PHLX on August 2nd. This additional fee has turned out to be an ineffective solution only hurting the retail customer. It is quite obvious now that the “Quote Rate Pirates” are in the inner circle of the PHLX and the NASDAQ. The research shows that if the PHLX were collecting the $2.10 fee for the unnecessary quotes it would be receiving in excess of $100 Million per day just from the filtered quotes. I doubt seriously that this is the case and conclude that the problem is within the entities that the Exchanges exempt.
What is an acceptable Quote Rate for the option market?
In late 2008, Themis Trading proposed to the SEC that a quote should stand for 1 second before being canceled (1 second Quote Life) as a way to stop the problems it was seeing in the equities market.
They continue to be a strong advocate of implementing controls to restore equality to the unfair equity market activities that ultimately resulted in the “Flash Crash” of May 6th.
Nanex concluded its May 6th study with a suggested 50ms “Quote Life” rule for the equities market.
Applying Nanex’s suggestion of 50 ms per quote to the option market would reduce a symbol quote count to no more than 20 counts per second per exchange. This would certainly reduce the risk of a catastrophe significantly. However, even this limit would overwhelm all available network capacity. If each of the 8 option exchanges updated a symbol 20 times per second there would be 160 updates per second per symbol (this paper has already demonstrated that 100 quotes per second for all symbols is not sustainable).
That being said, most of our future research will be done using 20 counts per second (50 ms).
Here are some current findings:
In the over 100 quotes per second studies filtered by exchanges (not just by OPRA best bid/ask), NQEX is now as active as the PHLX in number of occurrences. The PHLX frequently has quote counts of over 1,000 quotes per second with peaks reaching 2,131 quotes per second. (Remember this is on one option symbol for one second!). Activity is also found at the BATS, BOX, CBOE and PSE exchanges. As suspected there is a great deal of hidden activity away from the best bid/ask (as demonstrated above, potentially slows down OPRA’s ability to disseminate a best bid/ask).
In a study done on the full option data for 08/18/2010 (not just the best bid/ask data), with the filter set to 20 quotes per second (based on the Nanex suggestion of forcing quotes to stand for 50ms) the identified PHLX and NQEX quotes account for 3.92% of ALL of the option quotes for the day. The other 6 exchanges accounted for another 1.03%. This totals 4.95% of the total option quotes for the day.
The 5% may seem like an insignificant number, but it accounts for over 141 Million unnecessary quotes in the day. On the same day the NASDAQ and AMEX equity (not option) quote messages equaled 119 Million quotes. That is where most all ETFs and 4 plus letter stocks are quoted/traded…
The table below is of the activity of 8/18/2010 (no significance to the date, in fact a relatively low data day). The row titled OptionQuoteCounts_20100818a is based on OPRA best bid/ask data with count filter set at 100. The remaining rows with GR in the title are based on the complete option data feed (ie not just the best prices). The rows with _20 are using a quote count filter of 20, the ones with _100 are using 100.
This table shows the first few lines of the detailed study of the GR_PHLX_20 from above. The table is sorted by highest quote count in a second. Individual options are on the left and a summary by stocks on the right.
As can be seen the Quote Rate Pirates are active in the Option Markets. If the PHLX was collecting the $2.10 cancellation fees on this activity (obviously the Pirates are somehow excluded) the fees would be $123,029,856.60 for the day - that rule only hurt the individual trader.
Solution:
So how would such a 50ms rule be implemented? One simple solution would be as follows. All option exchanges now have a cancellation fee policy in place, but they are geared toward the consumer customer (who obviously is not the “Pirate” causing the problem). Redirect those fees and computer resources to a rule that states that all orders must stand for 50ms as in the Nanex scenario or even 1 second as in the Themis suggestion. Drop the computer overhead of tracking cancellations fees on the retail customer’s occasional attempt to buy an option and changing their mind and focus the attention on where the problem exists.
Nanex concluded its May 6th study with the following recommendations. I think the same should be applied to the option market.
- Quote and trade data must be time stamped by the exchanges at the time it is generated. This will ensure delays can be detected by everyone.
- Reasoning: Changing the procedure to time stamp at the time a quote or trade is generated is a near trivial exercise. It probably comes as a surprise to many that time stamping isn't done that way now.
- Quote-stuffing should be banned.
- Reasoning: It is a manipulative device designed to overload the quotation system. Quote and trade dissemination (data feed) is a finite resource, and should be treated as such.
- Add a simple 50 millisecond quote expiration rule: a quote must remain active until it is executed or 50ms elapses. If the quote is part of the NBBO, it may be improved (higher bid or lower offer price) at any time without waiting for the expiration period.
- Reasoning: The exchanges must protect the integrity of the National Best Bid/Offer system. What is the point of having a National Best Bid/Offer, if not everyone in your nation (apologies to Alaska/Hawaii) can reasonably execute a trade against it? 50ms is approximately the time it takes light and electronic communication to travel from New York to California and back. It is impossible to transmit information any faster. This rule would not limit quote/trade rates. So long as trades are executing, quotes can update thousands of times a second. Only a small percentage of quotes today would be affected and the potential for catastrophically high rates would be minimized.
As stated in the beginning this is a research project in progress, areas of this are unfinished, partially researched and often speculatively assumed. But the findings are such that it is important to make them available to those that might be able to help stop what is obviously the potential demise of the option market… Hopefully this work will motivate someone with the power to step forward and stop the activity before it is too late.
Harold R. Lanier
251-650-4436 Fairhope, AL
Page updated
Google Sites
Report abuse