"Do It Yourself" Latency Arbitrage: How HFTs Can Manipulate The NBBO At Whim Courtesy Of NYSE Empty Quote Gluts

Tyler Durden's picture

Another day, another stunner from the statistical wizards at Nanex. As readers will recall, in our latest piece we discussed the implications of the temporal arbing of the NBBO between the Consolidated Quote System and proprietary pricing tapes, like NYSE's OpenBook, which indicated a major discrepancy in the pricing data in widely held stocks like GE. In summary, at its peak, at 14:45:55 on May 6, the latency between the CQS and OpenBook pricing hit a high of 24 seconds, making a mockery of the NBBO as all those who had premium access to OpenBook were all too aware that 99% of the investing public were seeing pricing data almost half a minute stale, and could trade accordingly on secondary "dark" venues. At the time we were disgusted with the implications this phenomenon had on the NBBO, as this was nothing less than a full-blown NBBO arbitrage opportunity for the haves vs the have nots. Yet today Nanex takes this observation, and our collective blood pressure, to a whole new level, by not only confirming that there is in fact a trigger threshold in terms of quote saturation which immediately causes a latency arbitrage between the CQS and OpenBook, but closes the circle on the ongoing constant presentation of mysterious "crop circle" quote stuffing data. In essence, what Nanex' data implies is that HFTs can create latency arbitrage on demand between the NYSE pricing data dissemination to the CQS, but not to NYSE's own proprietary product, OpenBook, by pushing the consolidated NYSE quote rate beyond a magic number of 20,000/second. This immediately begs the question: just how much of the NYT's as defined "conspiracy theory" for an "on demand" Flash Crash is theory and how much is fact, if the cause and effect of the May 6 events have been inverted, and the NYSE's Liquidity Replenishment Points failed only as a result of HFT quote bombardment.

If confirmed, this is a stunner, as there is no reason why the NYSE should delay data hitting one data stream, the CQS, but not its own premium product, OpenBook - this would mean there is a gating factor that is essentially imposed artificially for the plebes (and the bulk of investors) who use the commodity CQS tape. It would also open up questions as to how often this form of borderline illegal arbitrage occurs on a daily basis on the NYSE: if all it takes is for the HFT participants to bombard the exchange with a few thousand extra quotes (as in "stuffing" the exchange) that have no intention of being traded on, but merely serve to choke the NYSE-CQS pricing translation system, then nobody can have any faith in anything coming out of the biggest public exchange. We can not wait to get the NYSE's, and most certainly the SEC's, explanation for this surprising, and even more market credibility destroying, phenomenon.

From Nanex:

We wanted to see the extent of the delay between NYSE quotes from CQS and
OpenBook on a more recent trading day. So we synchronized quotes from CQS and
OpenBook for GE between 1pm and 4pm Eastern time and plotted 30 minutes worth
of timestamp differences along with the quote price which are shown in Chart 1
below. We were surprised to see the frequency and magnitude of the delay. We
thought high quote activity in a stock would cause a delay in that stock's
quote, but could not find any correlation between the quote activity in GE and
the delay.

Then we decided to focus on the one minute that had the highest delay, which is
highlighted with a yellow circle in Chart 1. This one minute sample of the
delay is plotted in Chart 2.

Chart 1:

Instead of looking at the quote rate for just GE, we decided to plot the
quote rate for all stocks that NYSE sends to CQS. In other words, we plotted
the sum total number of quotes where the listed exchange and reporting exchange
are NYSE. This is plotted in Chart 3, which has the same time interval as Chart

You can see that there is a very strong correlation between the quote rate in
Chart 3, and the delay in Chart 2. Whenever the quote rate in Chart 3 exceeds
20,000/second, a corresponding delay is seen in Chart 2. The higher or longer
the quote rate exceeds 20,000/second, the greater the delay.

We then looked at all 3 hours and noticed the same relationship between total
NYSE to CQS quote rate and a delay in GE.

Then something very disturbing dawned on us. If the average or base quote rate
is around 10,000/second, then it only takes an additional 10,000 quotes/second
to reach the magic 20,000 quotes/seconds where a corresponding delay is seen in
CQS. This 10,000 quotes/second can be in any stock or combination of stocks
that NYSE sends quotes to CQS for.

Chart 2

Chart 3



Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
Missing_Link's picture

Tyler hits yet another gold mine.  Keep up the great work!

Racer's picture

I hope this brings about a complete failure of the market as the HFTs find no suckers to sell to because humans see no reason to participate in this rigged casino.

The market originally was a market for companies to raise money, pay dividends from their earnings and give something in return for the capital raising.

Now it is just a bunch of machines trying to rig the system and rob blind anyone who dares trying to look at it from fundamental or economic point of view.

Starve the beast and don't participate IMO is the best option, let them eat each other alive till only the tails are left!

mikla's picture


When you think about it, they've created an absolutely brilliant system.  It's the equivalent of a "rounding-the-penny" funneling of fractions into an account-of-choice, which is triggered on-demand.

This means we can shunt capital into any person's or company's account at whim.  And, we can all pretend it is a "victimless crime" (e.g., we're merely manipulating the market, front-running/quote-stuffing, and rounding-the-penny to our own accounts).  It's like having infinite power to control the "winds and the tides" of capital flow and capital accumulation.

Best of all, this type of system leaves no "smoking gun" like the cattle futures trades that burned Hillary.  Rather, the system works because it's hidden in the opaqueness of statistical analysis -- try getting a jury to understand *that*.

Of course, the ultimate problem is market distortion (things are priced incorrectly) and investor confidence (people don't trust the system anymore).  We're long past that, as retail investor outflows continue (nothing in this market is sane, it is dramatically overpriced).

It's probably game theory that brought it down:  If they exercised more discipline, they could have profited forever with no one"catching on".  However, like the failure of OPEC, the incentive is for some player to "cheat" a little more, skim a little more, and so the other players skim-and-cheat more as well.  At some point, the cheating reaches "too high" a level, and the suckers catch on.  That's what's happened (as shown by retail investor capital outflows).

Reminds me of the political system, where everyone hates their official, and nobody thinks government is doing a good job.  Strangely, people still tolerate the system, because it's all they have.  At least we are permitted to leave the capital markets (at some token level).

SWRichmond's picture

Excellent sum, mikla.  It also explains 100% profit trading quarters pretty well, doesn't it?

Assetman's picture


You've summed it up very well, Mikla.

I find it interesting that this Administration has somehow managed to alienate: (a) small businesses; (b) retail investors; and (c) non-government/non-unionized employees-- at the expense of a much broader population base. 

As for HFT, if you're a retail investor, Tyler has done more than enough to demonstrate that it's a losers game for the retail investor.  Moreover, it's becoming apparent that it's becoming a losers game for hedge fund managers (without HFT operations) as well.

What we simply have here is grand theft... with a complicit SEC looking the other way, and exchanges providing finanical rebates for "liquidity".  Folks, it isn't good liquidity when HFTs are circumventing NBBO's by flooding quotes.  They are destroying liquidity by forcing out market participants and weakening the market structure.

The next flash crash should be a doozy, though.

mikla's picture

Folks, it isn't good liquidity when HFTs are circumventing NBBO's by flooding quotes.  They are destroying liquidity by forcing out market participants and weakening the market structure.


It isn't liquidity at all.  It's volume.  HFTs provide volume without providing liquidity.

As you point out, it actually removes lequidity (as everybody figures out that the only way to win is to not play.)

Assetman's picture

How about a nice game of Thermoneuclear War?

Or, better yet... strip poker? :)

Assetman's picture

Well crap... REGULAR poker, then.


Mercury's picture

This is why market structure needs to be totally overhauled.  At this point the market is nothing but a video game played by computers and in this particular case the computer algos and their programmers aren't even trying to anticipate the market or outwit the competition with a better mousetrap but have instead figured out how to quite simply overload the processing power of the system and create a smokescreen between themselves and (most) all other market participants.

The point where additional value was being added in terms of liquidity, price discovery and lower transaction costs by further computerizing and automating the marketplace was passed a long time ago.  The need to co-locate servers just to keep the playing field level was kind of a bad sign.

If this were an old-fashioned, mechanical pinball machine the *tilt* light would be on right now which everyone could see and understand (transparency - see?).  If we had a simpler and more transparent market structure primarily designed to connect willing buyers with willing sellers, this kind of crap would simply go away. More complex technology, complex regulation or (always politically tainted) prosecutions will not fix this this problem.

Racer's picture

The sooner the companies/people who set up these multiple quotes that have no intention of being filled are brought to justice the better.

The country is broke, how about some nice fat fines on those that do this? After all it is illegal, why not DO something about it?

sheep92's picture

Yes, NYSE technology sucks.  Anyone trading with them or using their quotes is foolish.  This is known to anyone trading algorithmically.  Anyone sitting in front of a screen would have no idea anyway so who cares.  As of two years ago (last time I had direct knowledge) the main reason NYSE can't keep up is because of the specialist system where specialists are given the first look at an order and they have some number of milliseconds (used to be about 180 but I'm guessing its less now) to decide if they want to act on it.  This slows down the order flow.  If the specialists like they can even manually halt the order flow.  Anyone trading with them needs to have their heads examined. NYSE is a joke.

sid farkas's picture

interesting, this is why all this HFT bashing makes no sense to me, in the FX world all of the exchanges monitor the number of orders/price updates you make and automatically boot you if you are sending too many.

rogersails's picture

     Nice to see this come to light. In the internet world, it's called a "Denial of Service" attack.

      This was the topic of a comment I posted 2 weeks ago re: a prior quote stuffing article. My comment appeared for 5 minutes and then disappeared. Interesting. I'm not posting any more good stuff here.



Tyler Durden's picture

Your comments have never been moderated

Azannoth's picture

You mean this http://www.zerohedge.com/article/its-not-market-its-hft-crop-circle-crim...

the system says you commented on it, but i can't find you're comment and and topic is 'Updated' I guess it's a bug in the comment system


Tyler, how do I file a bug report ?

rogersails's picture

Yes, Azannoth, that was the article. The post went online, and after a few minutes, it disappeared. Maybe I'm just a cynic (hey, I'm a ZH reader), but since it disappeared, I've been waiting for this new article to appear.



Waterfallsparkles's picture

You probably hit the back button.  I have done that.  Yet, if you go to the top and hit home and check again, your comment is there.

Maybe you should go back to the original article and check to see if your comment is there.  I bet it is.

MichaelG's picture

It was a long thread, and the interjection of replies has pushed your comment onto page 2.  I've posted the link to it above.  I don't think the black helis were involved this time.

Edit: OK, that link's not even working for me, now!  But srsly - go to the article, scroll down to the end, click the link for page 2, and CTRL-F to find your name.

on Sat, 07/31/2010 - 18:21

Can they stuff a pipe so full it slows down, and use their own pipes to race orders to their execution point - in effect, frontrunning ?


rogersails's picture

Michael G

     You are correct - I see it on page 2.  I stand corrected.Tyler, my appology.

The point of my initial post is also validated by the current article. Quote stuffing can be used as a denial (or delay) of service tactic to arbitrage or front run other players.


Azannoth's picture

This beats all not only the HFT can set prices they can also control what(or rather when) other market participants see, it's like playing a shooter online against a GameMaster, u ony win if he alows you to

nonclaim's picture

Come on Nanex/ZH... So, different products that provide different data sets somehow behave differently and this is a stunner?

Thanks for the laugh, but the ignorance in the matters is the real stunner. But, to be honest, this is getting beyond ignorance and becoming a method... which is a bigger problem.

Tyler Durden's picture

The "different products" have the same input source. What is the difference: OpenBook is used by <1% of the trading community and has no latency. The CQS is seen by everyone else, and is irrelevant when it comes to prices during market crashes.

What was your point again?

sheep92's picture

Open book is used by whoever wants it.  Costs about $75 a month.  Used to be $5000.  Has plenty of latency relative to BATS, ARCA, and INET.  Anyone trying to trade HFT without looking at all the major books deserves to lose money.

What was your point again?

Tyler Durden's picture

The point is bolded and highlighted. Feel free to reread the article as many times as needed.

Also, free free to peruse the following DC Circuit decision that the SEC needs to do more to justify its decision to collect fees for giving access to NYSE Arcabook (and look who is representing the NYSE - the one, the only Eugene Scalia - gotta love supreme court junior capture). Oh wait, so the regulators love a two-tiered market? But yes, what is $75 to someone who knows the market is broken.

And let me paraphrase your sarcastic conclusion: "Anyone trying to trade HFT without looking at all the major books deserves to lose money."


sheep92's picture

Your only point seems to be that NYSE has technology problems and that they can be overwhelmed by volume.  Gee, thats news.  You also seem to believe that anyone that 'trades' deserves to lose money.  That's not very insightful.  If you are an investor then none of this makes a whit of difference as who really cares whether you get filled 18 milliseconds earlier or later.  In fact, if you are an investor you should love flash crashes cause then in you have a GTC order in at a low price it might get hit.

If you are a trader, then yes, you ought to know what you are doing. 

Tyler Durden's picture

"if you are an investor you should love flash crashes cause then in you have a GTC order in at a low price it might get hit."

Judging by the 15 weeks of outflows, investors are certainly loving the flash crash...and eagerly awaiting the next one (alas, without GTCs, which will get cancelled past whatever limit the PTBs determine is too far away from the NBBO for an execution price).

As for traders, with stock volumes at 1 year lows, they certainly don't appear to know what they are doing.

sheep92's picture

If you find yourself consistently alongside the public when it comes to investing then you ought to rethink your methods. Over time the public consistently underperforms any benchmark you might come up with.

I don't get your assertion that trader's knowledge is correlated with stock volume.  Does that mean traders were brilliant during the 87 crash when volumes hit a then record?  Or was the first day after 9/11 another time of collective wisdom?




Tyler Durden's picture

Oddly enough, if you find yourself in the most HFT concentrated stocks you have lost money on a 6 month, 1 and 3 year basis. (http://www.zerohedge.com/article/complete-q2-hedge-fund-holding-update-which-we-discover-181-hfs-hold-apple-stock). Maybe HFs should rethink their methods. As for your volume question, you answered it yourself perfectly.

sheep92's picture

You have an interesting style. When you are pointed out to be factually incorrect you wander off to another topic.  All this blather about HFT stealing billions from unsuspecting daytraders and 'ruining' the markets is just that, blather.  Are you secretly shilling for some specialist firm hoping to get spreads back to an eighth?


Tyler Durden's picture

No, not at all. Are you, on the other hand, not so secretly shilling for an HFT lobby group hoping get order cancellation rates up to 99.99%? Or has the recent heat by regulators gotten you a little on edge this morning? As to your false observation "If you try to place an order that will cross the NBBO it gets rejected" look up Rule 611 A exemptions - Qualified Contingent Trades: "Can cross stock outside the NBBO if part of a larger "properly hedge" options trade." And guess what, since the rule is retroactively compliant, the options and the stocks can have crossed anywhere in our disjointed market place, in both time and space, in essence making NBBO arbs more than feasible. "According to the SIA Exemption Request, a contingent trade “is a multi-component trade involving orders for a security and a related derivative, or, in the alternative, orders for related securities, that are executed at or near the same time."

But of course you knew that.



sheep92's picture

No, I trade for myself and a few friends.  I do no HFT cause I am not a broker-dealer.  If I get a rebate it about covers my trading and clearing costs so I don't 'make markets'.  If you think that there are free arbs in the market then why aren't you doing it and getting rich instead of whining about it.

Tyler Durden's picture

Because since this is a blog, and not a hedge fund, we do what we are supposed to do, which is to inform the people. You know, like what a non-corrupt media should be doing. But if you believe our IRR would be better trading in a broken market, we are more than happy to accept venture capital.

sheep92's picture

Then by all means inform them.  But don't make sh%t up or spin it so that it seems important.  All hft goes home flat for the night.  So no matter what some phd math/comp sci person has discovered about gaming the market its effect on an investor is near zero.  If you want to be a day trader then I'd suggest that no one owes you anything. 

Tyler Durden's picture

Make sh%t up? You mean like a conclusion that assumes the validity of your own initial assumption despite evidence to the contrary. And while most certainly not all HFT go home flat (yes, you will amazed to learn how many HFT funds file 13Fs), HFT is very pregnant during the day (and during the futures session, which surprisingly they trade too). Which is precisely where the next crash will come. But you must have been perfectly prepared for May 6, right?

sheep92's picture

So in your best blog informative style could you give some idea of the AUM of the hft funds that go home with a net bias and some idea of what that bias is.  That way we can at least bracket their influence.


No, on may 6th I was hiking with my girlfriend.  SInce I run a mostly flat book I'm not too scared of the next crash but thanks for your concern.


Pillage's picture

If I had a few GTC's down at zero.........got filled and cancelled, I'm not loving it!

"HFTs can create latency arbitrage on demand "........but do they offer NFL GAME DAY?........

mitack's picture

"Costs about $75 a month.  Used to be $5000"

So, you are saying they post old info on their web site ?



reads "$5000 for the access + $50/display"...

nonclaim's picture

Same source, different data sets.

One shows the best quote, the other the order book up to a depth. One has an old convoluted flow, the other is a direct stream. Latency is a problem on both systems, more on CQS aged system, less (but still there, frequency update) on OpenBook due to system/hardware capacity.

But you know all that. You *must* know all that.

It is the "let's find a conspiracy" mindset that bothers me. Plus the misuse of technical terms by Nanex that, while doing a good job presenting the side effects of algo trading, partially destroys the value by dis-informing the reader.

You have an intelligent readership. No need to dumb them down.

[This is a constructive comment, BTW. ZH is the best out there and I want it the stay that way.]

Tyler Durden's picture

You said a lot of interesting things. Yet it does nothing to change the fact that there are times when there is a 24 second latency between CQS and OpenBook, which can be effectuated and arbed.

sheep92's picture

You cant arb a consolidated quote feed and a real book.  You can only trade on a real book.  If the book is hopelessly latent then when you try to actually get filled then nothing will happen. There are no such things as intermarket arbs because that would imply that the books are crossed across markets.  If you try to place an order that will cross the NBBO it gets rejected.

defender's picture

If you look at the graphs, the latency is introduced at points that have an implied momentum.  As people try to front run the momentum, the latency makes them put in bids that are biased one direction or the other.  This allows the algos to extend buying or selling points whenever they feel like it.

Also, the latency isn't in the real book, it is in the CQS system.  Everyone thinks that they are getting a bargain because the price they got was better than the tape, when in reality, the price had already moved on and they were left holding the bag.  To put it another way, anyone will sell above the tape price, or buy below the tape price; but if the numbers that they hit are on the wrong side of the real price then they are just being screwed over.

sheep92's picture

let's say you have just two markets A and B.  On one market IBM is bid 100 offered 101.  On another it is 95 bid at 105.   For some reason the consolidated quote feed is stale and shows the market 98 bid at 99.  If you put in an order to buy at 99 then nothing happens in the real world because your order is below the best offer. If you are a seller at 98 on the wide market then your order is rejected because it would cross the books.  If you put in a market order you will be filled at the market price.

No arb.  No nothing.



Assetman's picture


You sound like a smart guy.  What do you do for a living?

sheep92's picture

I trade with a small group of friends. About half the capital is traded algorithmically about half is 'invested' the old fashioned way.


Assetman's picture

Thanks... very interesting... are you making more money these days algorithmically... or the old-fashioned way?

scratch_and_sniff's picture

Sheep, forgive me, but are you trying to tell me that when quote stuffers have intent to create latency for market makers then you see no wrong?

The possibilities for exploiting market makers are endless if you can create latency. If two market makers A & B, quote the same price at one point, then you create a latency for A and wait for the quote from B to rise and then place your bids with A. If the actual B quote doesn’t rise in the span of latency you have created, then you do it again until you get the arb, no matter how small. I cant quite figure out how you can say this is not a problem?