ITG Hid a Secret Trading Desk in Its Dark Pool
Here is a conversation that happens with some frequency in the financial industry:
Person A: Hey I have come up with a cool new way for us to make money.
Person B: What is it?
A: We should ____________.
B: Hmm, that would make us a lot of money. The problem is that it is super illegal.
A: Are you sure?
B: Yeah, it says it right here, see?
A: Could we do it a little?
A: What if we just --
It's worth giving this dynamic a moment of sympathetic understanding. The financial industry is built on top of many overlapping complex systems of regulation, so it is not always easy to know what's legal and what isn't. But the financial industry is also built on the dream of making money for nothing, of arbitrage, of perpetual motion machines, of converting intellect into cash without the annoying interventions of hard work and risk. People keep inventing magical sure-fire ways to make money, because the financial industry attracts people who are inventive in that peculiar way. Magical sure-fire ways to make money tend to be, you know, bad, for the people providing the sure-fire money. Regulators keep making them illegal. But the regulations are complicated, so people keep reinventing the same magical ways to make money, and then get shot down when their compliance officers point them to the illegality.
Usually! The Securities and Exchange Commission's action today against ITG is the craziest SEC action since yesterday. ITG is a broker-dealer that runs a dark pool called POSIT, and that provides other brokerage services (trading algorithms, smart order routers) through a subsidiary called AlterNet. "Despite telling the public that it was an 'agency-only' broker whose interests don’t conflict with its customers," says the SEC, "ITG operated an undisclosed proprietary trading desk known as 'Project Omega' for more than a year." In 2009 and 2010, according to the SEC order, "ITG explored initiatives to increase diversification and revenues," and someone came up with a very clever initiative. ITG gets all these customer orders to buy stock, see. So why not -- just hear me out here -- why not look at the customers' orders, buy the stock ahead of them, and then sell it immediately to the customers at a higher price? That's a pretty good risk-free profit. So that's what ITG did:
Step 2: Omega buys XYZ stock for $10.00 per share in a displayed market.
Step 3: ITG algorithm routes ITG sell-side customer order to buy XYZ stock to POSIT.
Step 4: Omega sells XYZ stock to ITG sell-side customer in POSIT for $10.02 per share, resulting in trading revenues of $0.02 per share for ITG.
Great, super. Risk-free money. At this point someone should have stepped in to tell the Omega guys that this was super illegal. And someone did! Just, later.
In early to mid-December 2010, ITG’s compliance department and senior management learned – based on the Liquidity Executive’s admissions – that Project Omega was trading based on a live feed of information regarding sell-side customers’ orders that had been sent to ITG’s algorithms. As a result, ITG immediately suspended Project Omega’s trading.
This stuff started in April 2010, so December 2010 is by my count ... kind of late to notice that it was illegal? And the guy in charge of the project ("Liquidity Executive") doesn't seem to have gotten in any trouble at the time. And "On or around December 21, 2010, Project Omega restarted a modified Facilitation Strategy that did not involve access to the Aleri Feed," which ran until July 2011, and which "continued to have improper access to information identifying POSIT subscribers."
So one thing to say about this is: Hahahaha, that's really bad! Like, paranoid-fantasy bad. The deep worry of modern equity market structure is that high-frequency traders, brokers, exchanges and dark pools are conspiring in some combination to front-run unsuspecting customers: The bad guys know, somehow, that the customers are trying to buy a particular stock, and can, somehow, race ahead of those customers to buy the stock and re-sell it to them at a higher price. And that's exactly what happened here! So, terrible. ITG will pay the SEC $20.3 million, a record dark-pool fine. "'The conduct here was egregious,' Andrew Ceresney, director of the SEC’s enforcement division, said during a conference call Wednesday," and it is hard to argue with that.
The next thing to say about it is: As guaranteed money-making schemes go, this one was pretty meh. This makes the name particularly awkward. If you're building a "Project Omega" it had better be at least an interplanetary death ray. But "ITG’s proprietary trading gross revenues resulting from Project Omega totaled approximately $2,081,304," says the SEC, and a person familiar with the matter tells me that net profits were about $124,000. That's partly because it was kept small -- "its total open positions could not exceed $500,000 at any time" -- but also because the scheme is not quite as sure-fire as it sounds.
Look back at the SEC's step-by-step description that I quoted above. A stock is quoted at $10.00 bid, $10.02 offered. Omega buys at the bid ($10.00) and sells to an ITG customer at the offer ($10.02). That's a free two cents. But to get the free two cents, you have to be able to buy at the bid. If the best bid in the market is $10.00, and the best offer is $10.02, you can buy immediately at $10.02, or sell at $10.00, but there's no guarantee you can buy at $10.00. You submit your bid to buy at $10.00 and wait alongside (really, behind) other bidders for a seller to cross the spread and sell to you at $10.00. And then there's not even a guarantee that you can sell at $10.02: You know your customer is looking to cross the spread and buy at $10.02, which helps, but other people are looking to sell there, and you need to get to the customer first.
These are not insurmountable problems. ITG's feed, which Omega used, included information about "parent" orders. A customer might ask to buy 10,000 shares at the volume-weighted average price over a period of time. The customer would enter that order into an ITG algorithm, and the algorithm would figure out how many shares to buy at a time and in what locations, splitting the parent order into smaller orders and sending them to different venues. If you know about the parent order, and if you also know how ITG's routing algorithms work (and the Omega guys did!), then you can be pretty sure that the customer will be sending lots of orders to POSIT to buy at the offer. So you can just put out a small bid -- remember, Omega was limited to $500,000 of open positions at a time -- somewhere and, if it executes, you have a very good chance of running and selling it to the customer on POSIT. Omega had a distinct advantage, but it was as it were a statistical advantage. It enhanced this advantage by taking "steps to ensure that the sell-side subscribers with which it was trading in POSIT were configured to trade 'aggressively' in POSIT": Omega would only trade with customers who it knew were likely to cross the spread and buy at the offer in POSIT, since those were the customers off whom it could make money. But it still couldn't just perfectly clip two cents a share off those customers. The stars needed to align a bit for this trade to work.
The third thing to say about this is how very close it is to things that are totally legal. (Despite being, itself, totally illegal.) Again, read that SEC step-by-step description. ITG is accused of buying at the bid and selling at the offer. That is exactly what market makers do. Customers go to market makers to buy and sell stock. The market makers charge the customers for liquidity in the form of a bid-ask spread: The customer can buy at $10.02, or sell at $10.00, and the spread is the market maker's profit. That's how it works.
What's bad here is that customers weren't going to ITG as a market maker. Its market-making business was secret. "During the entire time it was in operation," says the SEC, "Project Omega’s existence and trading activities were kept confidential and were not disclosed to ITG customers or POSIT subscribers." And: "Even within ITG, Project Omega was only to be discussed on a 'need-to-know' basis, and even the customer-facing side of ITG was not informed of Omega’s existence." Instead, customers were going to ITG to help them find the best price for stocks, on a pure agency basis. If there was stock available somewhere for $10.00, customers wanted ITG's routers and algorithms to find it for them. What ITG did instead was, at least some of the time, find that stock for Project Omega, and then sell it to the customers for $10.02. That is bad because it is not what the customers signed up for, and because it was not disclosed -- but not because it is intrinsically bad.
That is often the way that people get in trouble over market-structure stuff. There are lots of ways to make money off of clients, and many of them are allowed, even the ones that look kind of weird. You just have to give the customers a fair shot at figuring out how you're making money off of them.
I mentioned this morning that most dark-pool misbehavior seems to be about natural information asymmetries where people who make a living off trading-venue complexity know more about the trading venue's complexity than regular customers do. But there are exceptions, with the ITG case being the honkingest of them.
Or sell, etc., ceteris paribus.
It's just a feed, don't worry about it:
For the period of approximately April to December 2010, Omega’s Facilitation Strategy, which was designed by the Liquidity Executive, involved trading based on a live feed of information (the “Aleri Feed”) relating to open orders routed by sell-side subscribers to ITG’s trading algorithms for handling. The Omega team accessed the feed by connecting to a software utility called “Aleri” that was used by ITG’s sales and support teams. The feed contained various categories of real-time information regarding “parent” orders routed through virtually all of ITG’s algorithms, including: (a) client identifier, (b) symbol, (c) side, (d) quantity of shares, (e) filled shares, (d) target price, (e) the ITG algorithm in which the order was located, and (f) time parameters.
Liquidity Executive seems to have been Hitesh Mittal, who left in 2011 and was, until this past weekend, the head trader at AQR Capital Management. The SEC says that "After resuming trading in late 2010, Project Omega continued to engage in live trading until on or around July 11, 2011, when ITG terminated the Liquidity Executive as an employee and discontinued Project Omega’s operations."
Which sort of undermines my description of this as, like, guy invents risk-free way to make money and then implements it for a while before compliance catches up to him. I mean, it was that. But his intention was probably less about setting up a way to front-run customer orders, and more about setting up an actual proprietary market-making business. It's just that, instead of hiring market-making people, it used employees from the customer algorithm side:
None of the Omega team members had experience with proprietary trading. Instead, the Omega team consisted almost entirely of ITG employees with significant experience in ITG’s algorithmic trading group designing, building and/or writing computer code for ITG’s trading algorithms. Based on that experience, the Omega team had detailed knowledge regarding how ITG’s algorithms operated.
The SEC order doesn't break out whether the project was more profitable before or after it stopped using the customer order information, though, again, it was shut down in July 2011.
I have complained in the past that modern discussion of high-frequency trading uses the term "front-running" too broadly:
Here is Wikipedia: "Front running is the illegal practice of a stockbroker executing orders on a security for its own account while taking advantage of advance knowledge of pending orders from its customers." But then came "Flash Boys," which used the term to describe high-frequency traders "seeing an investor trying to do something in one place and racing ahead of him to the next," eliminating the broker/customer relationship (and the illegality) from the definition.
But this case preserves the broker/customer relationship (and the illegality): ITG was acting as a broker, and raced ahead of the customer. I don't know that it's technically "front-running," but it's not good.
That's on a total of about 1.3 billion shares traded, according to the SEC, meaning that ITG grossed about 0.16 cents, and netted about 0.01 cents, per share traded. Here is an estimate that Virtu makes a net profit of about 0.27 cents per share traded, presumably without access to customer order information.
That is for what was called the "Facilitation Strategy." There was also a "Heatmap Strategy" (see paragraphs 52-59 of the SEC order), which was roughly similar but used ITG router information to do similar trades in external dark pools.
Or: If you are a retail investor and you want to buy stock, you will end up in a situation very similar to that of ITG's customers here. If the stock is $10.00 bid, $10.02 offered, your retail broker will route your order to a "wholesaler." The wholesaler will offer "price improvement," meaning that it will sell you the stock at say $10.019, a fraction of a penny cheaper than what you'd get buying at the offer on an exchange. And the wholesaler will, in a perfect world, simultaneously buy the stock for itself for $10.00, risklessly making most of the spread for itself.
Consider also the UBS dark pool case, where UBS got in trouble for allowing sub-penny pricing in its dark pool. That would have been fine for its wholesaling business, but is apparently not allowed for high-frequency traders' orders in the dark pool.
One implication is that, if this worked frequently, then that would point to a problem in ITG's customer algorithms. Like, if you're using ITG to route your orders, and it's regularly finding you stock at $10.02 when there's stock available at $10.00, that's bad regardless of who is buying the $10.00 stock. If ITG was buying the $10.00 stock and then reselling it to the customer at $10.02, sure, that's illegally bad. But if Unrelated Firm X was buying the $10.00 stock, and Unrelated Firm Y was selling ITG's customer stock at $10.02, that's legal but sort of bad work by the router.
Obviously this happens sometimes to everyone: Markets are fast and fragmented, and nobody can always buy all the shares they want at the best possible price. But presumably ITG's algorithms often did work for customers, meaning that presumably the customers sometimes got those $10.00 shares before Omega could. Which, again, would help explain why Omega wasn't a guaranteed money-printing machine.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author on this story:
Matt Levine at firstname.lastname@example.org
To contact the editor on this story:
Zara Kessler at email@example.com