Credit Suisse On Flash Trading And Dark Pools, And General Regulatory Status Update

Tyler Durden's picture

Dan Mathisson of Credit Suisse held a conference call with clients yesterday, providing an update on his views of Flash Trading and Dark Pools.

The debate over Flash orders and Dark Pools has been heating up. Politicians have been railing against high-speed trading, the SEC has hinted at new impending rules, and some exchanges have voluntarily withdrawn their flash order type. Meanwhile, thrown into the regulatory mix are possible changes to Reg ATS and Reg NMS, new short- selling disclosure rules, and a new uptick or circuit breaker rule.

While Dan did have some harsh words for the media world, stating that a lot of misinformation has been spread, in an ironic twist he agreed tacitly with all the allegations that Zero Hedge and others have presented. But let's back up: Dan did correctly indicate that anyone can have access to the BATS datafeed for a nominal fee. However, the question arises how many people have the capacity to actually utilize this information, and especially in a way that conforms with existing trading infrastructure. A firm like RenTec, Getco and Goldman which has spent billions on trading facilitation hardware will of course be the first in line to see what happens in the 25 ms period when trading intentions are being solicited, while the vast majority of the trading world has no way to capitalize on this absent substantial CapEx investments to the tune of tens of millions of dollars.

What Dan, however, did highlight as the gist of the problem was the following: "flash trading gives out one's trading intention to the whole world, and the latter can trade ahead of you" - when you are dealing with hundreds of thousands of blocks hitting advance notice concurrently, it would be foolish to assume that this does not provide a major informational advantage. Indeed, Dan himself highlighted the tradeoffs: positive - collecting a rebate; negative - someone can step ahead of "your" order. Does not seem like that much of an exciting trade off. And Zero Hedge completely agrees with Dan's statement, that this becomes a major problem when this is done by brokers getting ahead of their own clients: "Clients get traded ahead of, while the broker saves money (or collects it via rebates)." Alas, this is exactly what is happening en masse.

Another point that Dan highlights is that Flash trading implicitly violates the spirit of Reg NMS, as a Flash order for the most part creates a locked market where an incoming bid is for brief period of time at the same level as the offer, indeed one of the issues NMS has riled against.

Bottom line: Flash is done, and according to Credit Suisse's DC Liaison and Head of Public Policy Mary Whalen the SEC will ban it shortly, especially as a function of the abnormally short time frame provided for comments (brings back flashback tears of joy to when the SLP was jammed down the throats of the investing public).

Some other, more relevant data points:

Credit Suisse notes that in addition to Flash, Actionable IOIs are next on the chopping block. While different from the IOIA page Zero Hedge discloses every now and then when JPM's ETF desk goes on full buying tilt in 10k blocks of SPYs, these are advance indications of trading interest (analogous from Flash) however which unlike Flash even, are completely hidden from the public view. Actionable IOIs are also over per CS. Amusingly, a client asked on the call how Actionable IOIs are any different from a call a broker places to his top 5 hedge fund clients while the smaller clients have no awareness of the best bid/offer presented by said broker. The response had to do with a semantic play in that IOIs are "messages intended to trigger a negotiation", which is not applicable to the broker example. Of course, this is a joke, and is a very critical angle of attack going forward as it very obviously disadvantages one set of clients (i.e. those who do not have a $10mm plus softdollar account with an i.e., Goldman Sachs) over those who, like PIMCO and Blackstone, spend billions a year in commissions to the biggest Broker Dealers. If one wishes a truly democratic market, this is a key area that also has to be revamped as it creates the most uneven, two-tiered market structure possible - granted not so much in equities but in other OTS instruments such as options and CDS. This will be a topic that Zero Hedge will definitely delve into much more in depth in the coming weeks.

On the topic of Dark Pools there was less conclusive findings, but the consensus was that the SEC will demand much more increased disclosure, potentially including real-time printing and a designator highlighting on which ATS a given trade was done. The pros: can see where a given trade was done, cons: the whole world will be aware of what trade occurred where (post facto), in essence limiting some of the major advantages of dark liquidity. As CS points out - "the loss of anonymity will hurt institutions, and the SEC may mandate regular reporting of volume as well."

Additionally, CS notes that the Reg ATS Fair Access Rules will likely have to be redone, as ATS now can decide which participants to kick off for any and no reason. In an effort to make markets more democratic this particular rule will end up being overhauled as well.

Of course anything that benefits the common traders to the detriment of the big institutions will likely not be received too well by Wall Street's lobby powers, so expect the dark pool fight to be quite intense, although at this point it seems that Schapiro is set on providing substantial color here as well. Even Credit Suisse points out that it does "not believe it is the SEC's view to disband dark pools." With a little prodding the regulator's view may change.

The most relevant take home message from the call, however, was CS' disclosure that there will likely be a major change to Reg NMS itself, specifically that displayed liquidity will be protected over non-displayed. This means that any NBBO quotes (either bid or offer) would be protected, and any sweep below or above the NBBO would need to take all the offered liquidity. In an example: a bid offer is at 34x35 (the NBBO): post amendments a trade would not be able to occur at 36 absent sweeping all the offered liquidity at 35. Zero Hedge agrees with Credit Suisse that this will be a "huge change to the trading landscape" and together with regulatory changes to dark pool implementation, this will be the most relevant development to keep track of. Of course, the fact that Reg NMS allows this to happen currently is very troubling indeed, as it explicitly renders the whole concept of the NBBO if not useless, then quite close to it, to the benefit of advantaged and privileged players who can completely bypass the NBBO if and when this suits their interests.

Lastly, CS thinks there will be no major changes to HFT, collocation, etc. Zero Hedge would beg to differ on this one.

Comment viewing options

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

Alas, Goldman et al will just find another financial market loophole to exploit before you can say "MARY SCHAPIRO IS A --"

dirkd's picture

Nyse went down 1 minute before open. You could not put in orders or cancel existing orders. I notice while watching price tickers and open book, that most stocks are trading

Risk as it pertains to ... ? Backstop behind the SLPz? So, conceivably, a "past ball" (not seen coming, i.e. flashed in a dark pool) thrown in the dirt could get by the catcher and roll how far (without a backstop)? There is no risk when Bernanke is the umpire behind the plate.
His message was being spread and gaining even more support...therefore he needed to be censored.

Until we have guys like Black back as regulators nothing will change. We just

good articles; my newest bookmarked finance site ..http://www.. hat tip: finance news & finance opinions

Miles Kendig's picture

Tyler - ZH is on a roll today.  I appreciate the excellent catches and analysis.

These are interesting times.  The information flow and dynamics between differing nexus are moving along nicely.

Ducky's picture

So how do you get rid of collocation? Let's say you are @ the cme and you can no longer pay them to collocate. Doesn't that just shift demand for the office space closest to their server. It seems like one floor above the servers is now valuable collocation space leased by somone else. A bidding war starts for their space.

When that lease is up the rent rises and, in the case of the cme, the rent goes to the cme anyhow.

Now if you are collocating @ their alternative site out in the Chicago burbs then you start a bidding war for the nearest land.

How much can things really be slowed down here?

channel_zero's picture

It seems like one floor above the servers is now valuable collocation space leased by somone else.

It doesn't work like that. 

You are talking about taking a building that was designed for limited datacenter functionality and turning it into mostly datacenter.  The power/cooling investment would be monstrous if it was capable of being retrofitted to that degree.  The investment would be monstrous too.

After saying all of that, if the building infrastructure documentation was publicly available, it would give you meaningful insight into the whole issue of front-running.

Ducky's picture

if the exchange was able to do this one floor below then a tenant could do it one floor above

Anonymous's picture

This topic is where ZH's journalism/advocacy shines best. But is ZH a one trick pony? Most other postings are lifted from MSM and given an inverse CNBC spin with a twist of conspiracy to rile up the populist crowd it has attracted.

Ducky's picture

so in other words they are suppling a demand in addition to their journalism/advocacy.

people like coming here to vent. there is something wrong with that?

is your real name Dennis?

Anonymous's picture

11) Evangelism. Spreading belief in the conspiracy theory is of the utmost importance. The conspiracist believes himself to be the sworn enemy of an immensely powerful malevolent enemy, which must be defeated. Telling others what he knows will make him a less appealing target for the enemy. But in spreading the word, he also becomes the hero in a grand cause, a paladin in gleaming armor against the dragon. Converting others to his beliefs will not only lessen his cognitive dissonance (he is, after all, often told that he is crazy), but will also convince others of his heroic stature. In the eyes of converts, he will go from zero to hero in one easy step.

"Conspiracy Theories as Examples of Pure Dogma"

Cheeky Bastard's picture

First. Before you go citing something from a q-rated site you better know what is dogma, what is knowledge and what are the methods to distinguish the difference. All knowledge is a derived probability of falsifiability, and dogma is a belief structure that the metaphysical origin or ontological one, will not become fluid and changed in time or in space. Knowledge is just a more probable belief in a metaphysical static of a being, while dogma is a lower form of epistemological expression of that same probability. So when you try to say you KNOW something, the odds of the universe are that you most certainly do not, and when you try to picture something as dogmatic you are probably false in that claim also. And if you don't understand what i have just explained to you, than you have no right to use those concept in barbaric linguistic attempt of expression to picture yourself or your source as a valuable point from which an argument can be drawn. And if you are not referring to mathematical axioms you present no knowledge just a belief which has not yet been disproved.

Anonymous's picture

Read the article Cheeky before you comment.

Let me just point out that just as it is appropriate for ZH to point out journalistic "bias" in CNBC's reporting, it is just as appropriate to point out "bias" on ZH.

Over and out.

Cheeky Bastard's picture

i have read the article, and that's why i said what i have said. he is being quasi-philosophical and mentions the " creation of knowledge " and many other false terms. First of all knowledge is not created but derived. it can be applicable or pure, but not created because the ontological nature of knowledge is consistent and static. He also uses meta-structural examples ( catholic church ) where any other super-structure would do just as same. Also, the article tries to academically define and battle the concept of conspiracy theory throught various fields of science, and yet provides no conclusio but an opinion. I am sick and tired of everyone who thinks that being a philosopher is the same as being speculative. It is not. And if you think it is, try to fly in an airplane which was engeneerd by someone who thinks that being an engineer is the same as knowing how to use a hammer. I have studies and mastered epistemology and logic ( and philosophy in general ) and i have a pretty good BS detector when it comes to recognizing quasi-intellectualism and quasi-philosophy.

dnarby's picture

True, but being long on polysyllabic bibble-babble and short on facts is a poor way to evangelize.

So is hypocrisy.

Anonymous's picture

I am trying to understand this statement. Perhaps if you could explain what you mean by "metaphysical static of a being" it would help. Do you mean metaphysical status? And what is a "lower form of epistemological expression of that same probability"? The form of expression of a probability should not matter as long as the probability, a number, is expressed. If knowledge is a belief, and dogma is a belief, then they are both the same type of epistemological expression, namely, they are beliefs.

Cheeky Bastard's picture

metaphysical static of a being ------> it is un-changable in essence ( being in philosophy is everything that is ( rock, soul, unicorn, atom)

also we only interpret noumen and phenomena ( we do not KNOW IT, we believe in it with a certain probability that they will change/not change in some time in the future ( see Carnap, Wittgenstein) )

" lower form of epistemological expression of that same probability " ( numbers are yet not defined, but we can agree that they can represent quantity, but that is only one of their ontological characteristics ) this would mean that something can be epistemologicly strong at the present time and considered as knowledge ( ie. Newtons mechanic ) but become dogmatic when the antithesis is derived as probabilistically more closer to knowledge ( ie. quantum mechanics, or the most recent model; String Theory ).

Also, number can be just one of the expression of probability, but its not THE only expression of it; one is language ( see. Logic), also it can be intuition( see Schopenhauer), it can be method( see Feyerabend),context etc.

Anonymous's picture

OK, that clarifies some of what you said. But the "lower form..." phrase is still not clear to me. Don't you just mean that a dogma has a weaker epistemological standing than knowledge on a scale determined by your epistemological theory?

Where we really get into trouble is the thing you mention - intuition (and worldview). As you know, most philosophical debate ends in conflicting intuitions. So we can expect political/historical debate to be even more dependent on unexpressed intuitions and worldviews.

Cheeky Bastard's picture

yes, i said lower form because there are different forms of knowledge

a) semantic

b) logical

c) systemic

d) empirical


the problem is that only systemic axioms are non-temporal and are evaluated as truths without any need for analytical clarification ( ie. Euclidian geometry).

semantic, logical ( minus the axioms of logic. see above ) and empirical knowledge are all forms of knowledge that can variate as the time goes by. but to make a long explanations short; empirical knowledge is mostly dialectic ( ie. you know the chair is in front of you in a time point A, and you know that the chair is not in front of you in a time period B ) also.


semantic knowledge is the knowledge about the meaning of words and is closely correlated with logical knowledge ( but logical knowledge does not need semantics to be discrete, while semantics need logical to be transferable ) and it is NECESSARY fluid, because if it was not so the evolution and progression of languages would be impossible.

and finally the logical knowledge is a form of knowledge which depends on defining the relation and co-relation between atomic formulas and language expressions.


Now, the point I'm trying to present here is that logical knowledge and systemic knowledge are axiomatic system with well defined static truths, and form a basis of our understanding, while semantic and empiric knowledge are fluid forms of knowledge and are dialectical. That is best seen when observing the evolution of natural sciences. While the particle of empiric knowledge is always defined with certain logical and systemic knowledge it fluctuates when exposed to semantic and empirical knowledge when it is a prevalent thesis/synthesis.

 So while something can be understood as a upper form of knowledge( systemic, axiomatic)at one moment in time, in another one can become a lower form of knowledge ( empirical, semantic ) because there is a dialectical antithesis which holds ground better in systemic and axiomatic form than the thesis which tries to lower. 

furthermore; when it comes to intuition; it goes to lower forms of knowledge simply because it is arbitrary in its ontological nature. Habermas wrote excellent pieces on that and the need for re-defining consensus and discourse.


if you want to know more you can read Popper, Wittgenstein, Carnap, Russell, Feyerabend ( excellent thinking regarding dogmatic superstructures, and particularly the scientific methodology)


i hope i was of help to you,and that i didnt confuse you with all this.



Anonymous's picture

No the ideas are quite familiar. Your terminology is sometimes unfamiliar.

Cheeky Bastard's picture

i know, i have granted myself a poetic license ..

Anonymous's picture

It is my working hypothesis that Cheeky is unemployed.

hypothesis. (n.d.). Unabridged (v 1.1). Retrieved August 13, 2009, from website:

If employed, it is my conjecture that Cheeky will soon be unemployed for spending so much time commenting on ZH.

conjecture. (n.d.). Unabridged (v 1.1). Retrieved August 13, 2009, from website:

Cheeky Bastard's picture

i can not fire myself, now can i .... I'm on a vacation, and you'll see me a whole lotta less here in a week

Ducky's picture

evangelism? like cnbc? or like the current p/e ratio of the s&p 500?

as for "Spreading belief in the conspiracy theory is of the utmost importance" I think my Senator from Illinois did a perfectly fine job of that when he said of the financial lobby's ability to get  whatever they want "Frankly they own the place"

dirkd's picture

"Can today actually turn out to be a down day...?"
No. PPT is on the job.

Risk as it pertains to ... ? Backstop behind the SLPz? So, conceivably, a "past ball" (not seen coming, i.e. flashed in a dark pool) thrown in the dirt could get by the catcher and roll how far (without a backstop)? There is no risk when Bernanke is the umpire behind the plate.
His message was being spread and gaining even more support...therefore he needed to be censored.

Until we have guys like Black back as regulators nothing will change. We just

good articles; my newest bookmarked finance site ..http://www.. hat tip: finance news & finance opinions

Anonymous's picture

what are you talking about? Reg NMS protects top of book already (NBBO) unless you use ISO orders in which case you are required to route to all the NBBO orders

Anonymous's picture

I know, this description sounds exactly like..... HOW REGNMS ALREADY WORKS.

Heh.. but if you try to read between the lines, perhaps he was saying that non-displayed liquidity which may currently take precedence in front of any displayed liquidity that is routed to might no longer get precedence.

Consider -- bid/ask of 10.00 x 10.25, and on the venue where the 10.25s are, there is a hidden offer of 10.24 sitting in front of it. Currently, if some other venue sends an ISO order to them to lift the 10.25s as part of their obligation under RegNMS, the 10.24s will get the fill. Perhaps he's saying that the displayed liquidity would truly be protected, and get that fill instead. (note the loss of price improvement for the buyer)

ig0r's picture

thats not really relevant. majority of trading is done in penny wide markets and non-displayed liquidity on the NBBO is not protected as far as I know (I dont trade equities right now though)

Anonymous's picture

This part of the post made absolutely no sense to me.

Anonymous's picture

It's good that they are making changes to Reg NMS. They should also respect time priority. if I am the first to bid at a level, I should be the first one filled if it trades there. Nobody should be able to pass me in the queue by bidding on a different ECN. The new market should also respect price priority. If 2 venues have the same price, but charge different fees for taking liquidity, then they don't really have the same price.

Anonymous's picture


You wrote:

> while the vast majority of the trading world has no way to capitalize on this absent substantial CapEx investments to the tune of tens of millions of dollars

The problem is not with CapEx - it is the lack of ability to program a computer efficiently.

A quad-core 2.6Ghz Dell server with 4Gb of RAM is more than enough horse power to track ever Level-2 quote as well as all flashed quotes, and still have enough left over to built a trading strategy that could execute a trade in under 25ms.

Anyone with the technical skill set and ~$100K can get up and running.


ig0r's picture

No way, not if you want to make money and trade more than a few symbols (which you would need to do to make consistent money). And under 25ms is 1995-slow

Anonymous's picture

25ms is very slow - but 30ms is the flash period, and 25ms is the time that Tyler mentioned.

With a single quad core at 2.6Ghz, you can trade in under 1ms, provided you have a broker that is fast enough, or you have sponsored access.

10ms is fairly easy to acheive with $100K in capital and a machine that costs well under $10K co-located with your broker (not with IB however).

If you think otherwise and have experience, then I suspect you should look into your own software and implementation.

I'm not saying that a $5K machine is going to compete with the likes of GS, because it likely will not... but the notion that you need millions in cap-ex to look at and execute against flash trades is BS. What you need is the ability to write very fast software in a low level language (i.e. assembly, well written C, exceptionally well written from a speed perspective C++).

Millions of dollars as Tyler suggested would get you so much hardware, that you could pretty much assure yourself of NOT being able to come up with a computing farm that could respond in under 30ms,, because the design would be incredibly difuse and would require far too much inter-machine communication.

Well written software working on a single machine with fast memory and nice CPU caches is all that is required. You can't use anything but an in-memory database with the timescales involved, and you can't access disk. Given how fast the data is changing, a good CPU cache is very important, which means that you run into problems as you add CPUs, since cache consistency becomes an issue, so there is not all that much you can do from a hardware perspective. Just get a fast machine with 1 or 2 fast CPUs, lots of fast memory, a well written OS (Solaris, Linux), and compact well written (i.e. fast but ugly) code.

As for the number of symbols traded - I don't buy it. I track over 3K symbols on a single quade core machine and have plenty of headroom.


Anonymous's picture

Tyler you are dead on the liquidity sweeps on NBBO

You can see this on the intraday block trades ALL THE TIME...the time and sales will show 50 cent swings in less then a the following...

13:01:43 SPY 300 95.51
13:01:44 SPY 100 95.52
13:01:45 SPY 5000 96.00
13:01:46 SPY 200 95.53

ZERO RISK...becuase SIGMA X has both sides of the trade...brilliant. Watch especially towards the end of the day when the "Liquidity" is balanced by HAL 9000...

The instiutions are having their pockets cleaned by the SLP's (SIGMA X) via Goldman...

Of course if the SEC decided to get the tapes and look at all the executed orders via SIGMA X they would find all of this out but there is NO WATCH DOG, NO RULES ...just whatever they fucking want to's cool with Mary. Funny how there is a "DEBATE" when they want to take something down and send it through the "normal channels" for discussion. Ahem....where was the fucking debate to get it implemented MARY!!!!

This is just one example...

Oh that's right they DID get the tapes before...remember 911 ...that's how they rolled up many of the terrorist...they got the tapes and trading accounts for those who shorted the airlines and insurance companies prior to 9/ fault...

We just allow our financial terrorists here to do whatever they want...well done Uncle Sam and politicians you are all fucking scum.

peterpeter's picture

You can't look at the trades... they can be reported up to 90 seconds after the fact before they are considered "late".

For something as liquid as SPY, that's what I assume is going on.  Best guess (not going to invest time looking) is that the trades you are referencing are for a price that was in the NBBO within the prior 90 seconds.

This is one of the more obvious issues with trying to put the uptick rule back in place using last rather than bid - but even the SEC in their own call for comments on Reg-SHO pointed this out (hence i am assuming no price test, or at worst a price test built on national best bid - which is still a horrible idea filled with technical issues since each person's view of the national best differs over small enough time quantum).


Anonymous's picture

peter^2, doublecheck your assumption on T&S reporting. While trades can be reported up to 90 seconds after fact, order flow/execution has to be documented as of time of event. Which raises question as to how clocks are sync'd.

peterpeter's picture

For a 50 cent SPY move, it would have to be a very very poorly synced clock.

I still have my money on a trade that happened earlier, where the actual timestamp on the trade is being lost in the likely numerous levels of data processing prior to arriving on ZH blog.

I'm not going to look myself, but if someone were really going to see if there was a problem here, the exchange reporting the trade and the exchange sequence number would come in handy.

I just don't believe that any trade for SPY is happening that far from the NBBO at that moment in time.

KidDynamite's picture

TD - question - perhaps i have inferred incorrectly that you are against dark pools (although I think you are)... could you please clarify this view?  I would think that if you are of the opinion that HFT algos are ripping of other traders, you'd be in favor of dark pools to level the playing field.


Deferred Comp's picture


How ironic that the holy grails of equity market structure and execution have  been shown for what they are :  More examples of our financial innovation gone awry with  information leakage at the speed of light whitewashed with a few mils of rebate rationalized as latency arbitrage.