Are HFT Algos Taking Aim At Dominating And Manipulating The Wonderful World Of ETFs Next?

Tyler Durden's picture

While many have speculated that the May 6 flash crash was a combination of High Frequency Trading (primarily), quote stuffing, ETF participation, and overall liquidity reduction, few, and certainly not the SEC, have been able to pinpoint the participation of HFT in disruptive ETF movements. Indeed, HFTs have been isolated in individuals stocks (best seen in the infamous "crop circles" images from last summer here and here) and specific futures contracts (most recently the NG NYMEX contract which experienced a truly bizarre algo driven sine wave pattern before flash crashing with no fundamental input) but rarely in actual ETFs. Perhaps this has been due to the relatively high volume of trades in some of the most popular ETFs such as the SPY, where the impact of one single algo would rapidly get lost in the noise. Well, a few days ago, Nanex once again was the first to catch the NatGas "sine wave" in action in what is possibly the most actively traded product in the stock market: the SPY or Spider ETF. Today, Nanex once again brings something very jarring to popular attention by focusing not on the most trafficked "synthetic CDOs" but on numerous ETFs that have not been front and center in the public's eye, yet which could serve as a great practice springboard to total market manipulation via HFT strategies - strategies that if taken beyond their reasonable limit, could crash the overall market very much how the NatGas algo crashed the price of gas by 8% in seconds. Presenting the RETF algo....whose purpose is currently unknown, but whose presence in the market should be known by everyone who trades stocks.

From Nanex:

On May 2, 2011 we began noticing a strange algo phenomena that would begin
immediately after the official market close. It started with one issue,
affecting only the ask price. It soon progressed into more stocks and soon
began affecting both the bid and ask price. As of 6/10/2011 it is seen in over
two dozen issues. The algo has several unique characteristics:

  1. To date we have only seen the algo run after normal market hours
    (specifically between 16:00:00 EST and 16:15:00 EST).
  2. It runs in short and long ETF issues (i.e. an ETF and it's Inverse ETF).
  3. It averages approx. 2000 quotes per second on each issue it runs on.
  4. Quotes from multiple exchanges follow the price.
  5. It consistently raises the ask price and recently began dropping the bid
    price (many times until the bid hits zero). In the majority of cases it also
    moves the NBBO to the same outrageous price levels.
  6. We have not seen any trades occur while the algo is running (yet). As it
    runs in fairly illiquid ETF's and after hours, this is likely due to the fact
    that there are no trading opportunities at the time.
  7. It has increased it's scope of ETF issues every week.

It would be easy to say this algo is a test of some type -- if it was only seen
on one or two days. However, seeing it run every day for six solid weeks (as of
this writing), increasing in both symbol scope and frequency, raises many
questions. One thing however is painfully clear; all one needs to bring a
stocks quoting price to zero (on the bid side) or into the stratosphere (on the
ask side) is the will to do so (and an algo that will blast thousands of quotes
a second to raise/lower the NBBO in small increments, thus not violating the
recently implemented stub quote
rule
).

Click for a larger image

First Sighting

The first time this algo was seen was on 05/02/2011. Just a few small blips in
one symbol on one of our most basic high frequency quote monitors:

What was unique was that it occurred after market hours (which is not that rare
but enough to make it a curiosity) and when examined we found it involved
multiple exchanges (normally we see run-up run-down algo sequences originating
from a single exchange). This is not to imply the algo itself is running on
multiple exchanges (although it is within the realm of possibility), but that
quotes from other exchanges are following the price movement and quoting at or
near the same (outrageous) level.

In the following charts, trades and quotes are plotted sequentially as they
occur. As such, no data is lost. Exchange's bid and ask prices are colored
according to the legend on the left. When quotes are the NBBO, a red circle is
drawn for the Best Ask, and a Green circle for the Best Bid.

Click on any chart for a high resolution image.

Symbol RXL

Chart #1: The first sequence of 250 quotes shows the ask price being run
up incrementally from $60 to over $400.
Click for a larger image

Chart #2: This chart shows the ask price continues to rise, although due
to the bid price being so low the scale makes it appear to be flat:
Click for a larger image

Chart #3: Removing the bids we can see that the ask price continues to
incrementally rise and affect the NBBO as it does:
Click for a larger image

Chart #4: This rise to over $500 occurs in approx. 6 seconds, with
approx. 18,000 quotes being generated in the sequence:
Click for a larger image

On week two the algo begins to affect both the bid and the ask price and also
begins to run on numerous additional symbols. It does not take the bid price to
zero but to a much lower level before stopping, while the ask price continues
to rise:

In the following charts, trades and quotes are plotted sequentially as they
occur. As such, no data is lost. Exchange's bid and ask prices are colored
according to the legend on the left. When quotes are the NBBO, a red circle is
drawn for the Best Ask, and a Green circle for the Best Bid.

Click on any chart for a high resolution image.

Symbol IFNA

Chart #1: The first 300 quotes:
Click for a larger image

Chart #2: 1000 quotes:
Click for a larger image

Chart #3: 5000 quotes:
Click for a larger image

Chart #4: 84,000 quotes:
Click for a larger image

We now see the algo affecting both the bid and the ask price and in many cases,
taking the bid price (in incremental steps) to near zero (0.01). When the bid
price hits 0.01 it then returns to the normal price range, while the ask price
continues into the stratosphere. We also see the algo filling our high quote
frequency/algo monitors every day:

In the following charts, trades and quotes are plotted sequentially as they
occur. As such, no data is lost. Exchange's bid and ask prices are colored
according to the legend on the left. When quotes are the NBBO, a red circle is
drawn for the Best Ask, and a Green circle for the Best Bid.

Click on any chart for a high resolution image.

Symbol GASL

Chart #1:. The first 500 quotes of the sequence:
Click for a larger image

Chart #2: 1,000 quotes:
Click for a larger image

Chart #3: 5,000 quotes:
Click for a larger image

Chart #4: After approx 14,000 quotes the bid price hits zero, then
rebounds to a previous price before the sequence began:
Click for a larger image

Chart #4:Symbol GASL - The Ask price continues to rise. 84,000 quotes
shown below in approx. 41 seconds:
Click for a larger image

On 6/10/2011 we see another change in behavior from the algo. As opposed to
running for 2-10 minutes in each issue (thus taking the bid to possibly zero
and the ask as high as it would go in the time span), the algo only ran for 1
full second in each issue. It is unclear (as of this writing) if this will
become the new normal, if it will soon stop altogether or if it's previous
behavior will continue.

In the following charts, trades and quotes are plotted sequentially as they
occur. As such, no data is lost. Exchange's bid and ask prices are colored
according to the legend on the left. When quotes are the NBBO, a red circle is
drawn for the Best Ask, and a Green circle for the Best Bid.

Click on any chart for a high resolution image.

GASX
Click for a larger image
RETL
Click for a larger image 
RETS
Click for a larger image
SBM
Click for a larger image 
SDK
Click for a larger image
SJL
Click for a larger image 
SZK
Click for a larger image
UCC
Click for a larger image 
UGE
Click for a larger image
UPW
Click for a larger image 

 

Additional examples of the algo in action can be seen here.

And before readers decide that this algo does not impact an ETF they care about, here is the list of names in which the RETF has been seen:

As of 06/10/2011 we have seen the algo run in the following issues:

BIB  ProShares Ultra Nasdaq Biotechnology
BIS  ProShares Ultra Short Nasdaq Biotechnology
BGU  Direxion Daily LG-Cap Bull
BGZ  Direxion Daily LG-Cap Bear
GASL  Direxion Daily Nat Gas Bull 2X
GASX  Direxion Daily Nat Gas Bear 2X
IFNA  iShares FTSE EPRA/Nareit
KRS  ProShares UltraShort KBW Regional Banking
KRU  ProShares Ultra KBW Regional Banking
LTL  ProShares Ultra Telecommunication
MWN  Direxion Mid-Cap Bear 3X - Triple-Leveraged
REK  ProShares Short DJ US Real Estate Index ETF
RETL  Direxion Daily Retail Bull 2x Shares
RETS  Direxion Daily Retail Bear 2x Shares
RXD  ProShares UltraShort Health Care Fund
RXL  ProShares Ultra Health Care Fund
SAA  ProShares Ultra Small Cap 600 Fund
SBB  ProShares UltraShort Small Cap 600 Fund
SBM  ProShares Short DJ US Basic Materials Index ETF
SDK  ProShares UltraShort Russell MidCap Growth Fund
SIJ  ProShares UltraShort Industrials Fund
SJF  ProShares UltraShort Russell 1000 Value Fund
SJL  ProShares UltraShort Russell MidCap Value Fund
SKK  ProShares UltraShort Russell 2000 Growth Fund
UCC  ProShares Ultra Consumer Services
UGE  ProShares Ultra Consumer Goods
UKK  ProShares Ultra Russell 2000 Growth
UKW  ProShares Ultra Russell MidCap Growth
ULE  ProShares Ultra Euro
YCL  ProShares Ultra Yen

 

How long before this test attempt to manipulate prices in C-grade ETFs comes to an SPY near you... And what happens when all the other algos, unfamiliar with this electronic trading pattern pull out of the market, and we see yet another historic collapse in a market that is sustained purely on hope and algorithmic levitation?