Robot Funds and Bank Regulation

Robot Funds and Bank Regulation

(Bloomberg View) -- Robots.

The robots are winning:

Big hedge funds including Balyasny Asset Management and Tudor Investment Corporation are beefing up their computer-driven approach following seven straight years of investor inflows into the “quant” sector.


“Frankly, we expect to see assets move from human managers to machine managers,” said Tony James, chief operating officer of Blackstone, a big investor in hedge funds, last week.

As a result, more funds outside the traditional quant universe are snapping up computer scientists, overhauling their technology, and tapping new sources of big data and analytics, like Kensho, which automates previously human-intensive research, or research firm 1010data, and credit card-data provider Yodlee.

There are two slightly different things going on here. One is the thing called "quant," in which computers analyze data and make relatively rapid automated trading decisions to respond to signals that they find in the data. The other is just, like, if you are a hedge fund that buys stocks or bonds, and you do lots of fundamental analysis to decide which stocks or bonds to buy, it might turn out that a computer is better at doing that analysis than a human is. Those things shade into each other, but they are at least a little different:

Fundamental equity investing and long-short hedge funds using quantitative analysis as part of their process is “a more likely push than really entering the quant space to trade against people like Quantbot and RenTech. People are too far removed to be able to do it.” 

You can think of the true quant funds as robot investors, doing their own thing, while the others are sort of robot-assisted human investors. The question then is whether investing is more like chess, where human/computer teams are better than either humans or computers alone, or more like driving, where giving humans control over the robots just destroys all the benefits of having the robots in the first place. Maybe human investors' common sense and gut instinct and feel for the market is a helpful check or enhancement on the robots' blind allegiance to their algorithms. Or maybe humans are panicky and irrational and will just mess up the computers' brilliantly logical plans.

In any case, this strikes me as not quite right:

However, a person at one of the funds warned that “any serious person in quant strategies would tell you there’s a limit to the amount of capital these strategies can successfully run.

“There’s a certain irony there,” they said. “Some of the people who’ve been in it for a long time aren’t taking new capital.”

Today’s arms race into quantitative and algorithmically driven strategies will be no different than all other Wall Street arms races of the past: The getting is good at first, until it works. Then it becomes a circular firing squad, as thousands of people and institutions pick each other’s profits off while engaging in similar strategies.

Those concerns focus on the thing called "quant," the smallish existing universe of computer-driven firms that use a particular set of strategies (basically, relatively fast trading of liquid instruments based on arbitrage of statistical patterns). The statistical anomalies that those firms find can only provide so much profit, and the more firms that are focused on exploiting those anomalies, the less profit there is to go around.

But that's not all that's going on here. The addressable universe for robot or robot-assisted investing is all of active investing. (Also, obviously: all of passive investing.) There is no a priori reason that a robot couldn't -- eventually -- be better than most humans at fundamental long-term buy-and-hold stock-picking, or distressed credit analysis, or private equity, or whatever. Statistical arbitrage in liquid instruments is one of the easiest places to apply automated trading and machine learning and so forth, which is why it was the one of the first places those techniques were widely applied. But it's unlikely to be the last.

Elsewhere, "researchers at Carnegie Mellon are working on giving robots introspection, or a sense of self-doubt." 


Here is Mike Konczal arguing against a naive approach to breaking up the banks by bringing back Glass-Steagall, quoting Jeffrey Gordon of Columbia Law School:

“Glass-Steagall divided the world intellectually into two distinct financial arenas, securities markets and banking. Securities markets were subject to the regulatory authority of the SEC, whose major tools were disclosure and enforcement; the banking agencies oversaw banks using a strategy of prudential oversight. Legal specialization aligned this way; so did academic work.”

However, he says, “in the ensuing decades, an increasing share of financial activity took place in the large space of functional overlap between banks and securities markets.” Since those activities were securities markets, “the principal regulatory tool was disclosure, even though the core activity was maturity and liquidity transformation, the sort of activity that we have learned from banking requires prudential oversight for stability.” He concludes that “[t]he consequence of this regulatory mismatch was a massive increase in systemic risk.”

Disclosure regulation really does feel like such an inadequate response to the financial crisis. Not only in the obvious sense that it didn't stop the financial crisis, but even in backwards-looking enforcement, it is just unsatisfying to pretend that the big problem with the mortgage bubble is that mortgage-bond prospectuses didn't properly disclose how bad the subprime underwriting was. The problem was the bad underwriting -- a prudential issue -- much more than the disclosure. It's not even clear that better disclosure would have helped: It was a bubble, and no one reads prospectuses anyway.


I'm sure I should care about this:

The Financial Industry Regulatory Authority (FINRA) announced today that it has fined Barclays Capital Inc. $1.3 million for systemic Order Audit Trail System (OATS) reporting violations and related supervisory failures.

FINRA rules require firms to transmit to OATS complete and accurate data relating to events in the lifecycle of an order, called Reportable Order Events (ROEs). FINRA found 15 system issues at Barclays Capital that gave rise to OATS reporting violations.

I mean:

“When firms fail to transmit OATS data or transmit inaccurate or incomplete data to OATS, market integrity is compromised because potential violative conduct, including manipulative activity and customer harm, may be obscured,” said Thomas Gira, Executive Vice President and Head of Market Regulation at FINRA. “OATS data is essential to FINRA’s automated equities surveillance program and is therefore critical to investor protection.”

Sure. It is very dull, though. The only fun part that I can see is that "Barclays Capital transmitted more than 3 billion inaccurate or incomplete ROEs to OATS, including omitted special handling codes; inaccurate timestamps, execution quantities and member type codes; and duplicate or erroneous reports." Three billion errors! That $1.3 million fine works out to about 0.04 cents per violation. But Goldman Sachs was fined $1.8 million last year for the same sort of boring violations, except Goldman had 63 billion errors, so it paid about 0.003 cents per violation. Barclays is paying about 15 times as much per violation, which seems pretty steep when you look at it that way. I suppose that's not how you should look at it.

Anyway I guess the main lesson here is that the electronic audit trail that regulators use to monitor the stock market has billions and billions and billions of mistakes!

Hedge funds.

Stamford Harbor Capital, the new firm started by Steven Cohen and led by a longtime deputy, is working with a third-party marketing company that’s meeting with potential clients to gauge interest in investment vehicles that could be started as soon as 2018.

Stamford Harbor is the one that Cohen owns, and that is staffed by employees of his family office, but that Cohen will not supervise and that can therefore raise outside money. (Though it hasn't yet, and says it won't until 2018.) This is distinct from Point72, Cohen's family office, which he does supervise, and which can't raise outside money until 2018, when Cohen's Securities and Exchange Commission ban expires. My view, as I've said more than once, is that it is just fine for Cohen to get back to managing outside money when that ban expires. But my even more strongly held view is: Good lord, why would he want to? If I worked all the time running my hedge fund and made billions of dollars, and then the SEC shut down my fund, I would take that as a sign that it was time to declare victory and retire.

But while Steve Cohen has an apparently insatiable love for running hedge funds, public pension funds' love for investing in hedge funds seems, well, satiable. Sated.

The New Jersey Investment Council voted to cut its target allocation to hedge fund managers by 52 percent, following similar moves by pensions in California and New York.

The New Jersey council on Wednesday unanimously approved a fiscal 2017 plan that calls for reducing its hedge fund exposure to 6 percent from 12.5 percent amid pressure from labor unions to reduce fees to the investment managers.

"You haven’t earned your fees and you haven’t been smarter and we aren’t going to pay for that asset class anymore," said the chairman of the council.

Elsewhere, Bill "Ackman’s Pershing Square Capital Management on Wednesday sold its entire stake of 9.8 million shares in Canadian Pacific Railway," though for some reason Ackman will stay on the board of directors for a while. And here is Tyler Cowen at Bloomberg View on Republican versus Democratic money managers. 

Blockchain blockchain blockchain.

Credit Suisse Research ... I guess tweeted out a research report about the blockchain? That seems right. This tweet is a highlight:

CS Research@csresearch
#Blockchain: Some consider blockchain to be the 'missing piece' of the internet
Twitter: CS Research on Twitter

Here is the full report, which contains some relevant cautions on the blockchain, such as:

Do you actually need blockchain? 'If it ain't broke, don't fix it,' for a blockchain to be relevant you must: (1) require a database, (2) need shared write access, (3) have unknown writers whose interests are not unified, and (4) not trust a third party to maintain the integrity of the data.

Are there many cases like that, in the financial industry? And:

More entry points make a blockchain system more hackable… The hackable 'surface area' of a distributed network increases with each node added.

That point has been proved over and over again on the bitcoin blockchain, where the entry points that have been hacked include Bitfinex, Mt. Gox, Cryptsy, Gatecoin, Bitstamp, BitFloor,, BIPS ...


  • Christopher D. Clack, Vikram A. Bakshi and Lee Braine: "Smart Contract Templates: foundations, design landscape and research directions."
  • Emin Gün Sirer: "How the Bitfinex Heist Could Have Been Avoided."
  • Dmitry Stillermann: "What Doesn't Kill the Blockchain Will Make It Stronger."
  • Izabella Kaminska: "Time to reevaluate blockchain hype."
  • Here's an audio version of Craig Wright's argument that he is bitcoin inventor Satoshi Nakamoto, which seems to consist mostly of him saying a word that I can't repeat here. 


I said yesterday that it's great that Instagram is copying Snapchat's most popular feature, because that means I don't have to learn how to use Snapchat. But that was a lie, because it turns out that now Instagram is Snapchat. Look, I am old, and Snapchat's interface has always felt to me like defusing a bomb. Now that has infected Instagram -- not just the Stories bit, but all of it. There are all these unexplained gadgets, and I'm afraid that if I touch the wrong part of the screen, my pictures will disappear, or be covered in lasers, or be nude. I want the carefully curated latte art back.

Anyway, here is an article about how Instagram is meta-innovating in the innovation space, by just copying Snapchat instead of innovating. (Also: "He cited the ability to understand the daily life of North Koreans or refugee workers through Stories and scenes they’d never put on Instagram.") And: "Advertisers want to be associated with the trendiest, newest thing," the poor dears, "and this year, that’s Snapchat."

People are worried about unicorns.

Some people worry about a unicorn bubble, but others wonder: If there is a unicorn bubble, how can I make money from it? Erin Griffin:

So let’s say, hypothetically, you’ve come to the same conclusion – we’re in a tech bubble — and you’re in the position to put your money where your mouth is. Can you actually short the tech bubble? The short (ha) answer is “no” because the companies and their investors are privately held. But that hasn’t stopped some investors from coming up with creative methods.

She mentions approaches like shorting public companies with exposure to startups, or buying shares in public incumbents that might benefit from the failure of hyped unicorn disruptors. 

That's fine I suppose. But it seems to me that the way to profit from a bubble is by selling into it, and that people sometimes focus too narrowly on short-selling into it. If you think that there's a bubble in private tech companies, and that venture capitalists are too eager to invest in any old random company at ludicrous valuations, then the most direct way to profit from that is:

  1. Found some dumb startup.
  2. Put together some dumb pitch deck.
  3. Go around to some dumb VCs and get them to fling money at you.
  4. Pay yourself a big salary. 
  5. Sell some of your shares on the secondary market at a high valuation.
  6. Quietly fold your startup when the bubble bursts.

It is sort of the "Producers" strategy, except it might be legal. (Not legal advice!) If you try it, and it works, do let me know.

Elsewhere: Not so fast with the vampirism there, Silicon Valley! (Earlier.) And: "How Silicon Valley helps spread the same sterile aesthetic across the world." ("Minimalist furniture. Craft beer and avocado toast. Reclaimed wood. Industrial lighting. Cortados. Fast internet.") And: "In Defense Of Bubbles." And people should be more worried about Rule 701:

Growth companies and startups compete for talent with larger companies based on the potential wealth creation these employees hope to generate from stock options. As startups stay private longer and attain higher valuations, the aggregate dollar amount of stock options they grant has increasingly been tripping over a $5,000,000 threshold baked into SEC Rule 701. That threshold obligates startups to provide robust disclosure (including detailed financials) on a recurring basis. While that may seem manageable, even if it’s undesirable, the real problems arise when startups don’t realize that they’ve already triggered a Rule 701 problem without having made the required disclosure.

People are worried about non-GAAP accounting.

Here's "One More Reason for Investors to Worry About ‘Earnings Before Bad Stuff’":

Companies that report significantly stronger earnings by using tailored figures like “adjusted net income” or “adjusted operating income” are more likely to encounter some kinds of accounting problems than those that stick to standard measures, according to research by consulting firm Audit Analytics.

I guess that makes sense. Aggressiveness about accounting is aggressiveness about accounting, and companies that try to make their earnings look better by reporting non-GAAP earnings are more likely to make their earnings look better by lightly cooking the GAAP numbers too. You could imagine an opposite dynamic: Perhaps every company wants to make its earnings look better, and the ones who don't go the open route of reporting adjusted earnings are more likely to go the secret route of cooking the GAAP numbers. But apparently not.

On the other hand, here's AllianceBernstein's Joseph G. Paul:

In our view, the spread between GAAP and adjusted earnings is a symptom of the dearth of investment opportunities in a world awash in capital.

Why do we say that? Let’s look more closely at what’s behind most of the growth in adjustments. The lion’s share is due to larger and more frequent write-offs of goodwill and asset values. So it’s not that companies are becoming more cavalier about accounting standards; it’s that they are getting more aggressive about writing off impaired assets.

People are worried about bond market liquidity.