Bank Businesses and Robot Traders

Matt Levine is a Bloomberg View columnist. He was an editor of Dealbreaker, an investment banker at Goldman Sachs, a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz and a clerk for the U.S. Court of Appeals for the Third Circuit.
Read More.
a | A


One dumb way to think about the banking industry is that every so often a big scandal comes along and sets the tone for the entire industry, ruling everything that comes after like an astrological sign. Until the next big scandal comes along and the astrological influences change. So now we are under the Sign of the Wells Fargo Fake Accounts, and one effect of that sign is that retail banking is Bad and investment banking is Good. Wells Fargo, which used to distinguish itself from the other big banks by its retail focus and conservatism in investment banking, "disclosed on Friday that new account openings had taken a nose-dive since the scandal over illegal activity at the bank erupted." (I suppose in the past it would have made up for that drop in new accounts by opening fake ones, but that is now frowned upon.) But also on Friday, JPMorgan and Citigroup beat expectations for bond trading revenue, "a challenge to those who think fixed-income profits are mired in a steady, irreversible decline." See people thought that after the last scandal, but the stars are aligned differently now:

For now, at least, Wall Street looks a slightly more comfortable place to be than Main Street. Banks in the US continue to fret over the spillover effects of the fake-account scandal at Wells Fargo, and are urging staff in branches and call centres to focus less on pushing products.

There's a bit more volatility in bond markets, and banks can finally make money in their trading businesses, while retail banking has retrenched into compliance mode. Bank of America continued the trend this morning; "third-quarter profit rose 7.3 percent as expenses fell and revenue from fixed-income trading was better than analysts predicted." Meanwhile Deutsche Bank is going the other way:

Deutsche Bank AG, Germany’s biggest bank, is exploring shrinking its U.S. operations as mounting legal expenses threaten to eat into the firm’s capital, according to two people with knowledge of the matter.

Such an option is being considered as part of the bank’s broader strategy review, which evaluates businesses in the context of regulatory and capital requirements, said the people, who asked not to be identified because the talks are private. The supervisory board of the Frankfurt-based bank discussed the U.S. business at a recent meeting and the topic has come up in talks with U.S. authorities, said one of the people.

Part of the reasoning behind building universal banking businesses before the financial crisis was, like, "this is fun," or "we are good at making money in retail banking, why shouldn't we also be good at making money in bond trading?" But there was also a real, shareholder- and regulator-friendly reason: Different sorts of banking franchises diversify each other. In volatile times, people may not want to open new checking accounts, but they might want to trade bonds. A risky bond-trading business combined with a boring retail-banking business might be safer than that retail business on its own. On the other hand, the more businesses you have, the more opportunity you have to be tainted by scandal. Goldman Sachs, for all its troubles, has probably never opened a fake checking account, but give it time. 

A teal button.

Here is a story about Global Trading Systems, an electronic market maker that acquired a spot on the New York Stock Exchange floor, where, at the end of the day, its floor guy "punches a teal button labeled DONE on a special square keyboard, closing out the stock." Don't you want that keyboard? Doesn't it sound so satisfying to finish a long day of work by pushing a big button labeled DONE? Given the automation in the trading markets, I imagine that the day is not too far off when his keyboard will have only two buttons, OPEN and DONE, and his job will be to try not to touch them accidentally. 

Anyway the reason GTS bothered to put humans on the NYSE floor is that it wants a shot to pitch its non-stock-trading services to public companies:

Using its prowess in equities trading, the firm can cozy up to the corporate giants whose stock it manages at NYSE, providing reams of juicy data on how the companies’ shares trade throughout the day. “That was a big thing that we were attracted to with this purchase,” he says. “It completely changed our ability to promote or show them what we could do.” Ultimately GTS wants to offer those mammoth companies other services, too, including opportunities to trade foreign currencies and fixed-income products such as Treasuries, helping them hedge risk on their balance sheets.

One question you might consider here is: Just how lucrative is high-frequency trading? There is this conspiratorial notion that high-frequency trading firms make billions off the fat of the land, secretly imposing a tax on everyone's trades while hardly employing any humans at all. But if you could do that, you would just ... do that. Right? Buying a NYSE floor specialist operation, building a marketing team, providing tons of transparent trading data to public companies and pestering them to hire you to trade FX, all of that suggests that the core electronic market-making business has become pretty commoditized.

The other thing I enjoyed is that GTS's co-founder and chief executive officer learned the trading business by learning fly-fishing. "Amidst the pandemonium of the floor there’s this natural elegance," says the CEO. "It's an art," says his mentor. "You have to decide what the appropriate fly is."  I guess people have been using vague romantic analogies like that for as long as they've been trading stocks, but they seem somehow more necessary when you're running a high-frequency trading business. If you can't intuitively grasp the details of what the algorithm is doing, your analogies to describe it become airier and more pretentious. The computer is basically tying some flies, you see. To fish for stocks. 

On the other hand here's a fund manager in China:

“When you empty your positions, looking at the markets is like sitting in a mountain in spring and contemplating the moon,” he said. “Only then will your investments rise like a dragon above the clouds.”

With patter like that, he could open a high-frequency trading firm. 

Failed operations.

Meanwhile, here is a story about Quantbot Technologies, a computer that crunches the bones of failed human fund managers and digests them into profit opportunities. Wait, what?

“We are seeing a lot of failed operations interview with us,” Mr Botlo said earlier this year.

“If they can develop what they have studied at these fundamental/failed places, they can implement it here. So for us it is really not that bad that everybody wants to dabble in quant, because that obviously means a more alpha potential that we can implement ourselves.” 

That's Michael Botlo, Quantbot's co-founder, talking about how he enjoys feeding his defeated enemies to his computer. They provide it with its most essential nutrient, trading signals: "The platform is robust enough to create scale," he says, "but we don’t have enough signals." And where better to get profitable signals than from people who have ... tried and failed to make money using them?

You can imagine three ways of investing:

  1. A human decides what patterns to look for, and looks for them, and when she finds them, she trades.
  2. A human decides what patterns to look for, and a computer looks for them, and when it finds them, it trades.
  3. A computer decides what patterns to look for, and looks for them, and when it finds them, it trades, and its human owners look on nervously.

Category 1 is fundamental investing, or whatever. ("Technical analysis," too, I guess, if you are staring at charts.) Quantbot seems to be in category 2: Its computer can go find patterns that will predict stock outperformance, but it needs people to tell it what patterns to look for. Number 3 is more of a pure machine-learning paradigm, though even there the machines need human nourishment, in that the humans are the ones giving the machines the data. Each has some form of human input, but the human inputs get more attenuated and harder to see. Every time we discuss this, someone sends me this famous passage from "Hackers," so I am just going to put it here now:

So Sussman began working on a program. Not long after, this odd-looking bald guy came over. Sussman figured the guy was going to boot him out, but instead the man sat down, asking, "Hey, what are you doing?" Sussman talked over his program with the man, Marvin Minsky. At one point in the discussion, Sussman told Minsky that he was using a certain randomizing technique in his program because he didn't want the machine to have any preconceived notions. Minsky said, "Well, it has them, it's just that you don't know what they are." 

Elsewhere in artificial intelligence: "IBM Is Counting on Its Bet on Watson, and Paying Big Money for It." Is it a little weird that Watson, IBM's artificial intelligence program, is getting into the financial-regulatory business but not the, you know, stock-trading business? Lots of firms seem to think that stock trading is a natural venue for artificial intelligence, but IBM is kind of going the other way. Is that just avoiding competition, or does IBM know something that everyone else doesn't?


Last week, Sanford C. Bernstein & Co.'s most metaphysical research team put out a note arguing that, in a world of low interest rates, discounted cash flow models get a bit wonky. This is a modelling point but also an economic one: If interest rates are very low, future cash flows are more valuable compared to present ones, and therefore your valuation of a business depends less on next year's profits and more on your estimates of its prospects many years from now. That's hard, is I guess Bernstein's point. But Aswath Damodaran, the New York University finance professor and Patron Saint of Correct Valuation, is not impressed:

This piece by Bernstein tells me more about how DCF is practiced (or mangled) at Bernstein than it does about DCF itself. As the piece indicates, a Bernstein DCF is a Robo DCF where as the risk free rate changes, nothing else does and not surprisingly the value goes up.

This too is a modelling point but also an economic one. The economic point is that if you expect low interest rates for a long time, you should expect low cash flows for a long time, so your estimates of the business's prospects many years from now should go down. (So the future won't count for as much as you'd otherwise think.) The modelling point might be more important, though: You can't just build a spreadsheet, plop in some mechanically derived inputs, and call it a valuation. If your risk-free rate changes, your growth rate has to change too -- or you need a reason why it wouldn't. You need to think about the value with your whole heart and mind, understand all your inputs deeply, give careful thought to potential future states of the world, and sum up all your analysis in a hand-crafted, deeply engaged valuation model. It is both art and science, and should never be done at 3 a.m. by an exhausted first-year analyst with a political science degree using a buggy template that a second-year analyst in a different industry group sent him.

But! You know. In this fallen world, bankers can't always live up to the teachings of the masters. Sometimes one wants valuation techniques that do not ask much of the hearts and minds of the people using them, which I guess is why price-earnings ratios are popular.

Mean reversion.

I periodically read stories about investment managers who had big losses, and then rebounded, and are now doing great, and I guess they are inspiring examples of the triumph of the human spirit over adversity? Here is one:

“He’s an artist,” said Ken Shibusawa, chairman of Commons Asset Management Inc. and the great-great grandson of the founder of Tokyo’s stock exchange. “Even if others tried to do what he does, they wouldn’t be able to.”

That's about Hideto Fujino, who did poorly enough in 2009 that he sold his asset management business for 3,240 yen ($31), but who is now back on top, "beating 97 percent of peers" and managing 200 billion yet of assets. The story discusses his impressive background, his strong track record, and his thoughtful investment process, but I can never read things like this without imagining that they describe literal coin-flipping contests. "He flipped six tails in a row, and was down to his last $31, but then he flipped heads eight times and now he is beating 97 percent of peers. What an artist!" How could you tell the difference?

People are worried about unicorns.

Here are some more details from Partner Fund Management's lawsuit against Theranos, the Blood Unicorn (Elasmotherium haimatos), none of which should be that surprising. The short version is that Theranos told PFM that it had stuff that worked, and it turned out that the stuff didn't work, and had never worked. Theranos denies this, or at least, it disagrees with the lawsuit:

In a statement issued Monday, Theranos responded: “The suit is without merit, the assertions are baseless, and the plaintiff is engaging in revisionist history. Most of the company statements the plaintiff has cited in its suit were made after the time the plaintiff invested, and could not possibly have been the original basis for investment. This wholesale reliance on post-investment statements, therefore, negates the claim that the plaintiff was misled.”

That is a pretty technical defense: Perhaps Theranos's products never worked, and perhaps it said that they did, but if it only said that after PFM invested, then no harm, no foul.

People are worried about bond market liquidity.

Hey here's another Treasury trading venue, this one for people worried about high-frequency traders:

David Rutter, the former head of the biggest electronic venue for Treasuries, says his startup will launch a new trading platform called LiquidityEdge Select this week. According to Rutter, a big draw is that it will enable clients to shut off bids and offers from firms they suspect are using hair-trigger algorithms to trade against their orders. He’s enlisted Cantor Fitzgerald to backstop the transactions and signed up about 90 clients, including most of the Treasury market’s 23 primary dealers and several high-speed trading firms.

It is fun to watch the Treasury market repeat the painful electronification process that the equity markets went through. All of this -- high-frequency traders pushing banks out of the market making business, suspicion that those HFTs offer "phantom liquidity," the growth of new venues promising to protect real investors from predatory HFTs -- has pretty much happened before. But the Treasury market lacks a Regulation NMS -- the rule that links all the U.S. stock exchanges together -- which has its good and bad points. On the one hand, if you build a walled garden for Treasury trading, it is fairly safe: You can't be forced out into the wide scary world of a national market system. On the other hand, if you build a lot of walled gardens for Treasury trading, who will you trade Treasuries with?

To be sure, a proliferation of trading platforms could potentially harm liquidity more than help it.

New venues may poach clients from the incumbents -- BrokerTec, Rutter’s former employer, and eSpeed -- but that may just lead to shallower liquidity across more venues and result in a Treasury market that’s more fractured than it is now.

Me recently.

I wrote about the Libyan Investment Authority's lawsuit against Goldman Sachs over some derivatives trades, which Goldman won last week.

Things happen.

"New York, even more than Frankfurt or Paris, is emerging as a top candidate to lure banking talent if London’s finance industry is damaged by Britain’s divorce from the European Union." UK looks at paying billions into EU budget after Brexit. Saudi Arabia’s bond sale: what to watch for. Puerto Rico’s Governor Warns of Fiscal ‘Death Spiral.’ Flush and Dominant, Australia’s Banks Come Under Pressure. CFTC Poised to Modify Source-Code Proposal. UK buyout and hedge funds make progress in gender diversity. Leading New York Law Firms Lag in Including Women and Minorities. AmEx’s Plan to Dethrone Chase Sapphire. UK prosecutors question Barclays executives for Libor probe. Challenging the Myth of Financial Intermediation. Beware the Code FreezeWikiLeaks Release Reveals Hillary Clinton’s Sympathy for Wall Street. "When he was mayor, Giuliani dressed up as a woman for a videotaped comedy sketch in which Trump — playing himself — groped him." Donald Trump thinks SNL is rigging the election and should be canceled. Bank teller action figure. Jessica Pressler goes shopping with Ryan Lochte. Man mixing LSD and cough syrup saves dog from imaginary fire. 

If you'd like to get Money Stuff in handy e-mail form, right in your inbox, please subscribe at this link. Thanks! 

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author of this story:
Matt Levine at

To contact the editor responsible for this story:
James Greiff at