Barclays Thinks Its Dark Pool Is Fine

If you stick to saying true things, it probably isn't fraud, even if those things are a bit silly. I mean, maybe. Don't try that at home, or at your dark pool.
Eric Schneiderman obviously has his own thoughts on the matter.

A while back I used a Barclays presentation to teach a lesson in the grammar of financial presentations. That presentation featured this chart of "Sample liquidity landscape by category," with a bunch of bubbles showing the various sorts of traders that traded on Barclays's LX dark pool:


Except it didn't show that. It just showed a bunch of bubbles. In particular, it omitted the bubble for Tradebot, a high-frequency trading firm that traded on the LX dark pool and would have earned itself a big bubble. Connoisseurs would not have been surprised by this, because the chart's title started with the word "sample," and, as I said at the time, "'Sample,' in financial presentations, means 'not true.'"

QuickTake Dark Pools

That presentation came from New York Attorney General Eric Schneiderman's fraud lawsuit against Barclays, accusing Barclays of misleading its customers about the prevalence of aggressive high-frequency trading in the LX dark pool. Now Barclays has filed a motion to dismiss that lawsuit, and if you are interested in the grammar of financial presentations you will not find a better textbook than this. Here's how Barclays's lawyers put the point about "sample": 1

Overlooking the explicitly stated purpose of the sample chart, the Complaint alleges that it constituted an "analysis purporting to represent the 'liquidity landscape' of Barclays' dark pool." This is plainly wrong. Nothing in the text suggests that this Sample Liquidity Landscape Chart shows the full set of LX participants as of any particular date, or that it "purport[s] to represent the 'liquidity landscape' of Barclays' dark pool." ... There is no suggestion on the chart that it was intended to be a comprehensive depiction of the order flow in LX. Indeed, the title of the chart prominently states that it is merely a "Sample," which means "serving as an illustration or example." Webster's Third New International Dictionary of the English Language, Unabridged 2008 (definition of "sample").

Like I said: It means "not true." And sophisticated institutional investors -- the sort of investors who own dictionaries -- would have known that.

But there is much here beyond the "sample" thing. For instance, Barclays also advertised that it could "continuously police the trading activity in LX" and "quickly identify aggressive behavior and take corrective action with clients who exhibit opportunistic behavior in the pool." Schneiderman's complaint says, well, OK, but you never did that:

Barclays has never prohibited a single firm from participating in its dark pool, no matter how toxic or predatory its activity was determined to be.

Indeed, Barclays has known about the high levels of toxic activity occurring in its dark pool -- including latency arbitrage -- and has been aware of which firms are responsible, yet Barclays has refused to stop it. 2

Sounds bad, right? Not at all!

The Complaint alleges that "Barclays does not actually police or punish bad trading behavior" because Barclays did not prohibit any firm from participating in LX. This allegation wrongly presumes that policing of the pool requires expelling traders from the pool. In fact, to "police" means "to supervise the operation, execution, or administration of" something. See Webster's Third New International Dictionary of the English Language, Unabridged 2008 (definition of "police").

Barclays's dictionary gets quite a workout in this motion. 3 But actually there's an important substantive point here. Barclays didn't say, "There will be no aggressive high-frequency traders in our dark pool." It said something else:

In line with its representations, Barclays segmented traders based on trader-specific metrics and then grouped those traders into six categories. Investors who were willing to interact with any other trader (including "aggressive" traders) could do so, and those who only wanted to interact with non-"aggressive" flow could do so as well, thereby prohibiting "aggressive" traders from accessing their trading flow. There is no allegation that Barclays stated that it had "prohibited" anyone from trading on LX, and the fact that Barclays did not entirely expel a firm from trading in its pool does not render statements that it would police the pool false.

Barclays doesn't advertise LX by saying, "There are no HFTs here," or even, "There are no aggressive HFTs here." It advertises something more like: We can categorize orders and traders, and then you can choose whether or not to interact with the sorts of order and traders you don't like. If you are afraid of aggressive HFTs, you can avoid them. If you just want to trade as quickly as possible, you can trade with whoever's there. 4

Now, Schneiderman would say: But you actually can't do that, because Barclays classified lots of objectively (by its own metrics) "aggressive" HFTs as not aggressive, so opting out of interacting with "aggressive" traders wouldn't do you any good. Again, there is an answer:

The basic premise of this argument is erroneous: Barclays did not represent that it was categorizing traders based solely on certain stated "objective profiling criteria." It said the opposite: the marketing materials explicitly told clients that "Analysis of Liquidity Profiling factors is ongoing" and Liquidity Profiling would consider factors other than those mentioned, "for example, . . . parent-level metrics." A summary that Barclays made available to clients also stated: "Barclays reserves the right as operators of LX to override the profile of any participant."

Barclays advertised that it was "able to monitor the quality of flow in the pool on an ongoing basis by using a set of quantitative metrics to grade and classify each participant." But it didn't advertise that it did that. It just said it was able to do that. In fact it explicitly said that it wouldn't rely solely on those quantitative metrics: It would rely on the quantitative metrics, plus whatever else it wanted to think about. So if for instance it classified one of its own desks as passive, even though it was (according to Schneiderman 5 ) "objectively" aggressive, that was fine.

There is a lot of other delightfully technical parsing in this motion, but I guess at this point it has to go downstairs. 6

Stepping back from the parsing, though, Barclays makes a couple of really good points:

  1. Everything we said was true.
  2. Even if it wasn't true, there's no indication that it was material to anyone's decision, or that it misled anyone.
  3. Even if it misled people, there's no indication that they were harmed.

And I actually find myself broadly in sympathy with all of those points? Like, I have made fun of the "sample" thing incessantly, because it is funny, but let's be fair: It's 100 percent right. Barclays really didn't say, "This is a chart of the trading in our dark pool." It said something more like, "This is a chart showing that we can produce a chart of the trading in our dark pool." That was literally true. That chart is darkly amusing, but it really isn't evidence of fraud. 7

The rest of Barclays's responses here strike me as similarly persuasive. Technical and nit-picky, but persuasive. If you stick to saying true things, it probably isn't fraud, 8 even if those things are a bit silly.

The argument that no one was harmed is in some ways less important to the case, but more interesting to the world. As I said when Schneiderman brought his complaint, it might be "long on evidence of false advertising," but it's "shorter on evidence that the 'predatory trading' was actually predatory." Does that matter for the case? Well, maybe; you do need to show damages to recover damages. 9

But the broader question is, does it matter for Barclays? If its customers -- which, it can't stop reminding you in this motion, were very sophisticated -- got better execution with Barclays than they did from other brokers, then they'd keep coming back. And if they didn't, they probably wouldn't. That, more than the shape and size and color of some bubbles on a two-page handout, is probably what drives investors to or from Barclays. As Barclays says:

LX's customers are highly sophisticated traders and asset managers responsible for investing millions or billions of dollars of assets, who execute trades across multiple markets and ATSs, are capable of closely monitoring the quality of execution they receive based on extensive data, and can select from multiples platforms on which to execute their trades based on detailed execution data, not on the glossy marketing brochures or quotes from magazine articles the NYAG cites.

And the fact that LX was, before Schneiderman's complaint, the second-largest dark pool, suggests that the actual execution was pretty good, regardless of what the handouts did or didn't say. Part of that, by the way, was probably because of the presence of high frequency traders, who allowed LX to trade a lot of shares and provide good execution.

Now that's just my intuition, and the fact that volumes on LX are now a fraction of what they were before the suit might be read the other way: Now that Schneiderman has exposed what's going on, everyone seems to have left. But that's not conclusive: Customers may be fleeing because they don't like lawsuits, not because they don't like what this lawsuit has revealed. 10

There's a third possibility too. Perhaps Barclays's execution was fine, and every other broker's execution would have been similarly fine, and no customer was harmed or benefited much either way. 11 But perhaps Barclays was one of the top dark pools not because it provided better execution than other dark pools, but because Barclays was better at creating the impression that it protected customers from evil high-frequency traders. The way you do this is not by lying about how many high frequency traders are in your dark pool -- and Barclays was always honest about that 12 -- but rather by shading and insinuation. You talk about categories and aggressiveness and sophisticated monitoring tools. Your customers don't quite know what they dislike about high-frequency trading, but they know it's something. And your job is to nod sympathetically and convey to them that you feel the same way.

The bubble chart doesn't say, "We don't allow high-frequency traders." It says something more like, "We hear your worries about high-frequency traders, and they worry us too, and here are some soothing bubbles."

Now that's not fraud. You might find it a little annoying, because the natural next step would be to say, "and we have solved all your worries about high-frequency trading," and then to solve them. But that requires knowing what they are: identifying actual harm that aggressive high-frequency traders cause to nice investors, and then fixing that harm. Another dark pool, IEX, was started with that specific goal in mind, with specific clearly explained measures to address certain clearly identified problems, and has gotten a lot of traction. 13

But in general if the problems were easy to identify and fix, then -- well, for instance, then Schneiderman's complaint would have identified the harm. Right? If Barclays was deceiving people into trading with predatory high-frequency traders, then there'd be some evidence of that predation. LX would have provided bad execution. People would have lost money. Schneiderman doesn't bother to argue that investors were harmed because, I assume, he doesn't know if they were.

On this theory, Barclays's customers were attracted to LX by presentations that were not false, really, but that took advantage of general paranoia about evil high-frequency trading without actually making any sort of explicit promise, or effort, to prevent it. 14 And now they've been scared away from LX by a lawsuit that takes advantage of that same unfocused paranoia -- and that also makes no effort to figure out what the problem is, or how to fix it. Or even if there's a problem at all.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
  1. Here and everywhere I quote from the motions, citations are omitted. Except the dictionary citations of course. The boldface is my emphasis; the italics are in the original.

  2. It goes on:

    For example, on January 16, 2014, senior leaders in the Equities Electronic Trading division were provided an analysis identifying over a dozen major high frequency trading firms engaged in significant trading activity in Barclays' dark pool. That analysis discussed those firms' history of sending "toxic" order flow. One high frequency trading firm was described in the analysis as "historically . . . very toxic." Another firm was described as having "[trading activity that] is very toxic, and the client is up-front about this." Another firm was described as having "[k]nown latency arbitrage flow" in the dark pool.

    Barclays has not denied any of those firms (or others) access to its dark pool, despite its representations that it will identify "aggressive behavior, [and] take corrective action" to "refuse a client access" to the dark pool if such aggressive or toxic high frequency trading strategies are discovered.

  3. Tangentially related, but here's Chris Arnade:

    You remember that kid in elementary school, the one who would argue during a game of tag: "You said you have to tag the person. Well you only touched my clothes. That isn't a person."

    Remember that kid? That kid is Wall Street.
  4. Something similar is going on in Barclays's ATS disclosure, which says that the dark pool allows "conditional orders" -- in which an HFT firm can send in a nonbinding order to see if there's interest at that level -- but also allows customers to opt out of interacting with those orders.

  5. By the way Barclays objects to that classification, and it's right:

    There is also nothing misleading in the fact that Barclays rated its own desks as "passive" traders. The NYAG's only attack on this rating is a conclusory statement -- with no citation -- that Barclays rated one of its desks as passive even though that desk allegedly "engaged in high-speed, high-order volume trading akin to high frequency trading" and so, based on unidentified "objective profiling criteria," should have been rated aggressive.

    There's no evidence in the complaint, beyond an anonymous griper, that the desk should have been rated "aggressive" by Barclays's objective metrics, whatever they were.

  6. One favorite:

    The NYAG alleges that Barclays' statement that the trading activity in LX was 6% aggressive was false, because Barclays "categorized approximately 25% percent [sic] of the orders taking liquidity in its dark pool as aggressive" during discussions with an unnamed "prominent high frequency trading firm." These allegations, however, draw a comparison between two entirely different calculations. In its marketing materials, Barclays discloses that 6% of all "trading activity in LX" is "aggressive." The 6% number is calculated as a percentage of all orders, i.e., orders taking liquidity and orders providing liquidity. In contrast, the 25% "aggressive" figure is, as acknowledged in the Complaint, calculated as a percentage of "orders taking liquidity in [the] dark pool." There is no other allegation in the Complaint that these two numbers are inaccurate, apart from the fact that they were based on different metrics.

    Now, on the one hand, that seems like a fair point. On the other hand, if 25 percent of orders that (execute and?) take liquidity are aggressive, and zero percent of orders that provide liquidity are aggressive, then, I think, by definition, 12.5 percent of orders that execute are aggressive? Barclays's argument is that the "6% number is calculated as a percentage of all orders," not just executed orders, but the actual chart is unlabeled -- it says "New liquidity profiles" and then has percentages expressed just as percentages, not as percentages of something. So the statement is true on some interpretation, but it's not clear how you're supposed to interpret it.

  7. Especially since, as Barclays points out, there's no time period on the chart. You don't know if that's the "liquidity landscape" for a second or a day or a month or a year, and whether it's sampled as of today or as of six months ago. If you actually wanted to know the landscape, questions like that would matter to you, and you'd ask for actual data. If you just wanted the comfort of a pretty chart, the sample chart would suffice.

  8. Hoo boy is that not legal advice.

  9. Though the stereotype of New York's Martin Act, under which Schneiderman brought this case, is that it basically applies to everything and doesn't require much in the way of showing reliance or harm. But actually Barclays has a pretty good argument that the Martin Act should not apply here, because it is "limited to actions for fraud in the purchase or sale of 'securities,'" and there's no purchase or sale of securities here.

    Now this sounds a little nuts -- after all, you only use Barclays LX to purchase or sell securities -- but hear them out:

    Because the LX trading venue is not a "security," any alleged misrepresentations about LX, and the services it provides, are not misrepresentations made "in the issuance, exchange, purchase, sale, promotion, negotiation, [or] advertisement, . . . of any . . . securities." None of the alleged misrepresentations by Barclays involved "fraud in the making of an investment decision." Accordingly, any misrepresentations alleged in the Complaint are outside the scope of the Martin Act.

    Federal courts have recognized that misrepresentations made by a broker-dealer about the manner in which trades are executed cannot constitute securities fraud. ... Similarly, in Abada, plaintiff alleged that he opened a trading account with defendant based on defendant's misrepresentation "that [defendant] would provide fast, high quality execution of trades." The court rejected the contention that the fraud was "in connection with" a security, noting that the losses claimed by the plaintiff related to the defendant's processing of orders, not "misrepresentations concerning the risk of a particular investment or investment system."

    This seems ... probably right to me? (There are also related arguments about federal preemption of market structure regulation over state law.)

    Now this is interesting for Barclays, but it's even more interesting for Schneiderman's broader "Insider Trading 2.0" crackdown. If the Martin Act just doesn't apply to exchanges and dark pools -- if misrepresentations about who gets what access when to data and execution and speed and so forth can't, as a matter of law, be Martin Act fraud -- then Schneiderman loses a lot of his leverage to influence market structure.

  10. Especially since the customers are mostly institutions who are fiduciaries for investors, and trading with a venue that's being sued for fraud might seem unacceptably risky to them.

  11. I remind you of this post from a former electronic trading product manager, which I interpret to mean, among other things, that running your own dark pool is not so much about improving execution for customers as it is about saving some money on exchange fees. Which doesn't necessarily mean that it makes execution for customers worse! It can just be neutral for them, but a money-saver for you.

  12. "Barclays explicitly told clients that ELPs were a major component of the pool," says the motion. (ELP, "electronic liquidity provider," is Barclays's preferred euphemism for high-frequency trader.)

  13. And if it turned out that IEX's famous box full of wires was not actually full of wires, well, there'd be hell to pay.

  14. Right below the bubble chart Barclays would show this chart:

    [imgviz image_id:igWf7H0fs0uA type:image]

    Now what does that say? I think it says something like: "Instead of talking about how many high frequency traders there are, we're going to talk about something else." Like, there's an arrow between those two rings, but it doesn't represent any change in the world. It represents change in classification. We have 35 percent high-frequency traders (ELPs), yes, but let's not say that. Let's say instead that we have 9 percent aggressive traders. It sounds nicer.

To contact the author on this story:
Matthew S Levine at

To contact the editor on this story:
Zara Kessler at

Before it's here, it's on the Bloomberg Terminal.