Added_110203

Quote Rate Pirates are Attacking the Option Market (Continued)

A “Splash Crash” is coming

CNBC’s Jeff Cox warns that the next May 6^th type market meltdown will extend beyond stocks and is already being dubbed the “Splash Crash”

http://www.cnbc.com/id/41408510

Last summer a significant amount of time was spent researching and reporting to the PHLX Option Exchange, the NQEX Option Exchange, the OCC, and the SEC information on the high number of quotes being generated on the same option symbol in a short period of time in the options market. The findings of this research was compiled in a Comment to the SEC’s Concept Release on Equity Market Structure

http://www.sec.gov/comments/s7-02-10/s70210-283.pdf

And further detailed in a webpage

http://sites.google.com/site/quoteratepirates/

The following report is a continuation/follow-up of the above study.

On January 19, 2011, OPRA reduced its data stream projections for July, 2011, to 17.3 Billion messages per day from 17.9 Billion, though it did increase its required capacity to 5,257,000 messages per second from 4,069,000.

http://www.opradata.com/specs/upd_traffic_proj_11_12.pdf

It is unclear why the expected daily requirements would be down if the required capacity is up by 29%. But as demonstrated in the previous report the numbers are way too small if the high quote rate issue is not resolved. The previous report demonstrated that 819 Billion quotes per day could occur if all exchanges allowed all symbols to quote at 100 quotes per second. At the time of that report there were only 8 exchanges, now there are 9. Even if just the PHLX allows all symbols to go to 100 quotes per second the number would be over 100 Billion per day!

The Link below is to a spreadsheet that shows a more comprehensive study of the history of high quote rates in the option market. The data source for the study was Nanex’s OPRA “Best Bid and Offer” for the option market. (Note: to do the same research on all exchanges would require more time and resources than could be justified.) The quote counts per second are filtered at greater than 100 for this study. (Note: though the previous study indicated all future studies would be on 20 quotes/second or more, it was determined that the amount of time and computer resources would be impractical until the bigger numbers were stopped.) Here are some interesting observations:

• Most important, the problem of high quote rates in the option market still exists.
• It does appear that the frequency of occurrences has reduced but days like 01/10/2011 at 150,164 attacks proves that the mechanics for high days are still in place.
• The Pirates have been sharpening their tools as can be demonstrated on 01/21/2010 when there were two attacks of over 3,000 quotes in one second on a symbol: 3,541 on MCD 2011Jan22 C 75.00 at 13:19:11 and 3,483 on TCK 2011Jan22 P 60.00 at 10:41:00. This is almost double any number seen previously.
• Attempts were made to connect low volume days to holidays and snow days with the idea that the quote stuffing Pirates did not get to work on those days. Jan 11^th and 12^th were big snow days in Philadelphia. Jan 12^th did have an extremely low count of 8,400 but Jan 11^th had a count of 21,979 which is only relatively low. Evidence to support such a theory is inconclusive at this time.
• The peak period appears to have occurred in June and July 2010, with occurrences frequently reaching above 200,000 attacks and as high as 381,837 on 06/10/2010.
• On 07/20/2010 there were 122,543,177 quotes that can be attributed to quote stuffing out of a total of 1.2 Billion BBO quotes. In other words, 10% of all quotes in the BBO market for the day were attributable to the over 100 quotes/second counts!
• The PHLX is still the dominate problem exchange. The estimate still is that the PHLX is used in over 90% of the attacks. Further, as previously demonstrated, it can be assumed that the attackers are exempt from the PHLX’s Cancellation Fee policies.
• The previous report was unclear as to when this type of activity started, but it did indicate that the level of sophistication was increasing as the year 2010 went on. This report demonstrates that the development of the attacks was on going for several years back.

Link to OptionQuoteCounts spreadsheet (Note: to follow the text, it might be easier to open another browser window and paste this link into that window):

https://spreadsheets.google.com/ccc?key=0Atf2d5P3QgrMdEFuUEhkZzFDYmh5VF91aXBvS1R6eXc&hl=en&authkey=CKC5-ckJ

Explanation of Headers:

FileName: OptionQuoteCounts_20110110_GH_100_PHLX

• _ 20110110 shows the date in the format YYYYMMDD
• _ GH means that the data is OPRA’s Best Bid Offer (GR tapes used in the more comprehensive study of 20100818 showed all changes by all exchanges)
• _ 100 is the filter size for number of quote changes per second before reporting (previously it was indicated that 20 was a more reasonable filter, but the number of issues at 20 is still so high that it is impractical from a time perspective to create such reports)
• _PHLX exchange means the study was further filtered to only show if the PHLX exchange was a Best Bid or Best Ask in the quote change. Most reports end in BBO as the individual exchange is not being studied.

TotCount: Is the number of occurrences in the day where an option symbol had more than 100 quote changes in a second. One might note that there are 23,400 seconds in a day so numbers like 150,000 means that on average 6 symbols per second are being attacked through out the whole day.

TotSumCount: For each occurrence that is identified there is a number of quote changes that is greater than 100. This is the sum of all of those values.

TotAvgCount: This is the average calculated as TotCount / TotSumCount.

NumOpt: Often the same symbol is attacked for several seconds; this is the number of different symbols attacked during the day.

NumUnder: Often, several option symbols of a particular underlying are attacked. This is the number of different underlyings attacked during the day.

HiHi: The highest quote count for one symbol during the day.

HiHiTime: The time it occurred.

TotOptQuotes: The total number of option quotes for the day.

SumCount/OptQuote: The percentage of option quotes that are involved in the attacks during the day.

As can be seen from the above table, Jan 10, 2010 was a relatively high count day with 150,164 Total Quote Counts. In the table below a separate report was run for each of the 9 exchanges. As can be seen the PHLX was involved in 147,753 of the attacks which is approximately 98% of the total. The PHLX was involved in the Highest problem symbol for the day with 1282 quote changes in one second at 10:04:27 on the symbol AMR 20120121 10 call. There was, on average, an attack on 6 option symbols per second for the entire trading day. 1.71% of all PHLX option quotes for the day that made the BBO stream were involved in some kind of a ridiculous Quote Stuffing scheme. If the PHLX in fact was charging the $2.10 cancellation fee on the parties involved in these attacks it would have collected $40,198,918.20.

(Note: Click on image to see full view)

In summary, we are looking at only the tip of an iceberg. The 08/18/2010 detailed study was on a day that had 2.9 Billion quotes on all exchanges which is well within the above OPRA projections. This was considered a typical day for the study period. Only 1.41% of the quotes were related to quote stuffing at 100 quotes/second (Note: 2.61% of the BBO quotes). What happens if this number is left unchecked? It still seems that the bulk of the activity is at the PHLX, though it is obviously happening at other exchanges as well. The activity is still isolated enough that the individuals involved can be identified and should be stopped before the problem becomes more destructive.