This is where I'd want to hear a speech about high-frequency trading. Photographer: Louis Lanzano/Bloomberg

Why Do High Frequency Traders Never Lose Money?

Matt Levine is a Bloomberg View columnist writing about Wall Street and the financial world.
Read More.
a | A

So I didn't like a lot of things that Eric Schneiderman said in his "Insider Trading 2.0" speech against high-frequency trading on Tuesday, but here is a particularly puzzling bit of it:

Unlike the rest of us who invest in the markets, as I hope you all do, some high-frequency traders appear to trade with virtually no risk. This is something we're very interested in. Last week, a large high-frequency trading shop disclosed that it made money on every trading day over the course of four years. Out of 1,238 trading days, they made a profit on 1,237 of those days. I do not begrudge anyone their right to make money, but if something seems to be too good to be true, it usually is. And we question whether there are some traders that are just so smart that they never, ever lose money without some special advantage.

Puzzling because, one, why does he seem to hope his entire audience of lawyers day-trades, and, two, there's a serious misunderstanding of the trading industry here. Schneiderman is referring to Virtu Financial's much-discussed recent IPO filing, but similar complaints have been leveled against Goldman Sachs and JPMorgan and ... well, Goldman Sachs, a lot. Every firm that is in the trading business seems to make money on trading on a whole lot of days, and lose money on a whole lot fewer days.

QuickTake Trading on Speed

And sure, it does seem unfair that, in the uncertain world of securities trading, some firms would make money (almost) every day. Imagine how suspicious it would be if, for instance, your local supermarket made money on every carton of milk that it sold. That just seems too good to be true, doesn't it? How can they know the price of milk before you do? Shouldn't they be losing money on half of their milk, and making it on the other half, so that things balance out? Doesn't the fact that they always make money suggest that they're ripping you off?

So ... no, right? Here is how Virtu describes its business:

Virtu is a leading technology-enabled market maker and liquidity provider to the global financial markets. We stand ready, at any time, to buy or sell a broad range of securities, and we generate revenue by buying and selling large volumes of securities and other financial instruments and earning small amounts of money based on the difference between what buyers are willing to pay and what sellers are willing to accept, which we refer to as "bid/ask spreads."

That is, Virtu (like Goldman) is selling a product, and that product is liquidity, and it charges for that product. High-frequency trading firms are in the business of acting as middlemen, providing a valuable service by letting buyers and sellers trade as soon as they want to, rather than waiting for fundamental sellers/buyers to come in on the other side of the market.

Now you could reasonably object to that business, but your objections should have the same form as legitimate objections to other businesses. And "that business consistently makes money!" is not a legitimate objection to most businesses. In fact it is the reverse: Businesses that consistently make money are typically thought of as good businesses, not bad ones.

But there are still legitimate forms of objection even to profitable businesses. You might, for instance, think the business is too profitable, and want to disrupt it by cutting out the middleman and letting ultimate buyers and ultimate sellers trade directly with each other, without passing through the hands of market-makers, high-frequency or otherwise. That's fine! That's a rough and partial description of what dark pools are; they're places for fundamental investors to trade without encountering middlemen. And yet exchanges survive, because sometimes people want immediate liquidity and are willing to pay a middleman for it. (Also: Schneiderman seems to be anti-dark pool. )

Or you might worry that the business is too monopolistic, that barriers to entry are too high. That would be an argument for providing equal access to data feeds, for instance, which is a Schneiderman priority. But you should be clear about what that means. The goal -- and the effect -- here is not to reduce high frequency trading. It's to make high frequency trading easier, by reducing the fees that new entrants have to pay to compete with incumbent traders. If exchanges and newswires can't charge HFT firms tens of thousands of dollars for data feeds, HFT will get cheaper, and there will be more of it, not less.

Relatedly, you might worry that this business is profitable due to regulatory rent-seeking rather than actual customer needs. And in fact, common and reasonable arguments about high-frequency trading are that it takes advantage of exchange privileges given to market-makers, or that it games maker/taker fees, or that it otherwise takes advantage of a sophisticated understanding of the rules to earn rents. (And, relatedly, that the desire to earn these rents leads to an escalating arms race: The rents accrue not based on absolute speed, but disproportionately to whoever is fastest.)

You will hear critiques like this from high-frequency traders themselves, about their competitors: "We don't game maker/taker rules, but those guys do." These are mostly arguments for less regulation, rather than more, though to some extent they fit with Schneiderman's desire to crack down on the special privileges that exchanges offer to HFT firms.

Or you might just say, "the people currently doing the business do a bad job of it, I wish someone did a better job." That is a fine thing to say if you are planning to start a trading firm! As a regulator, though, it's somewhat less impressive; a regulator has lots of power to restrict bad businesses but little power to will good businesses into existence. You (probably) can't make trading costs come down by increasing the costs of the guys currently doing the trading.

So, I don't know, there are plenty of sensible positions you can take against high-frequency trading, and Schneiderman even arguably takes some of them. But "high-frequency traders consistently make money trading, and that is too good to be true," is not such a sensible position. If you find yourself unable to understand how the people you regulate make money, that does not in itself mean that you need more regulation. It means you need more understanding.

(Matt Levine writes about Wall Street and the financial world for Bloomberg View.)

  1. Update: Obviously a bid-ask spread is more of a probabilistic fee than a markup on milk. But if you're doing a lot of transactions, as HFT firms are, and making small probabilistic gains/losses on each one with a positive expectation, that converts into pretty consistent income. Here is a calculation concluding that if an HFT firm makes money on 52.5 percent of trades, loses the same amount of money on the other 47.5 percent, and does 10 trades a minute, it will have a losing day once every eight years.

  2. It's unclear, though in the speech he says that they "are even less regulated, have fewer reporting requirements and are far less transparent" than exchanges, which I guess he means to be bad things.

  3. This is what a lot of buy-side firms say, in essence, when they complain about trading costs being too high. Also here is a paragraph from a thoughtful reader e-mail:

    It's also important to understand that privileges which were granted to ensure firms were profitable over time are now used to ensure firms are profitable every day. 10 or 20 years ago, at least in the exchange-listed markets, intermediaries were most certainly not the most active traders in the markets, and most certainly lost money from time to time providing liquidity. Today, intermediaries are the most active traders in the market and brag about never losing money. For all the unanswered phones in the 1987 crash, intermediaries as a group lost a fortune that day (and so Greenspan had to prop them up). In the 2010 flash crash, they made a fortune. In 1987 the feedback loop that accelerated the crash came from investors, and in 2010 it came from intermediaries.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

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

To contact the editor on this story:
Toby Harshaw at