This is the once-a-week free edition of The Diff, the newsletter about inflections in finance and technology. The free edition goes out to 15,300 subscribers, up 137 week-over-week.

In this issue:

• The Gamer/Arbitrageur to Generalist Pipeline

• Office Politics, Redux

• Vaccine as Perk

• Who Owns User Data?

• Tech Sees Like a State, Test-and-Trace Edition

• Conflicts of Interest

• The Negotiate-With-Hackers Hack

• Small Deals and Stagnation

The Gamer/Arbitrageur to Generalist Pipeline

Many of the most successful investors in the last generation got their start in arbitrage before moving on to other things. “Arbitrage” is a broad term, that ranges from trading the same asset in two different venues to capture price differences (there used to be good money in trading the gap between the price of gold in New York, London, and Hong Kong) to betting on the outcome of a merger that’s been announced but not consummated.

Warren Buffett was doing complicated three-way arbitrage in the 1950s.[1] George Soros' first financial job was arbitraging the price gaps between New York-, London-, and Johannesburg-listed mining stocks. Carl Icahn got his start making a market in options and warrants, and hedging out the risk. Before Elliott Management was one of the largest activist hedge funds, it was a convertible bond arbitrage fund. Robert Rubin’s arbitrage desk at Goldman in the 70s and 80s produced a steady stream of hedge fund stars—Richard Perry, Eric Mindich, Eddie Lampert, Dinakar Singh, and Daniel Och all ran large funds with long winning streaks. John Paulson focused on merger arbitrage before his record-setting killing shorting mortgage-backed securities.

Most of these investors did not focus on arbitrage for the rest of their careers, though. They generalized, and made much of their money outside arbitrage. Why was arbitrage an unusually good training ground for other kinds of investing?

It’s a game, and there’s a metagame.

The game is simple: given a current price, an expected future price, timing, and odds, you can construct a portfolio that maximizes reward for a given level of risk. When there’s news flow, the first arbitrageur to react correctly wins. (Jim Cramer tells a story in his memoir about an oil merger that was rumored to face antitrust action. He was attending Harvard Law at the time, and casually asked his antitrust professor about the deal. The professor said it would go through, and Cramer bought.)

At one level, this arbitrage game is a pure example of advantage gambling. The problem domain is simple. If a company was trading at $30, it gets an acquisition offer at $40, and the stock trades up to $38, you can be reasonably confident that you’re risking $8/share to earn $2.[2]

But how confident are you in that model? Did the stock trade at $30 because that’s what the company was worth to a financial buyer, while a strategic buyer would pay more? Or did it trade at $30 because everyone was betting on the deal before it happened, in which case a failed deal would send the stock lower?

Conversely, once a company has an offer, does this increase the odds of a bidding war? If nothing else, a deal—or a rumored deal—will cause the company’s shareholder base to consist of fewer long-term holders and more arbitrageurs. Perhaps the best case study of this is that when Avon Products got a fake takeover offer, the stock rose even once it was clear the offer was fake: arbs had bought the stock, and that made an actual deal more likely. (Avon was eventually taken over, but years later.)

Arbitrageurs have to use every data point to update their model. If a CEO says something to the media, that’s information; if the same usually-voluble CEO stops talking to the media, that’s informative, too. If a company has received an offer, and the stock trades near the value of the offer, but it starts drifting lower, that’s something that demands an explanation.

The metagame is less about estimating the probability of a known event and more about visualizing the entire space of possible events, and then figuring out the event path from there. That’s a more general skill, but it’s a skill that’s far easier to apply from a foundation of converting every data point into an updated view of the odds.

The game is limited, and the metagame is not, but if you’re not great at the core game, knowledge about the meta is basically useless.

The investors I cited all got their arbitrage done from the 50s through the 80s. The business has changed: today, the arbitrage business is well-understood, and there are fewer inefficiencies. As those efficiencies compress, arbitrageurs' returns on capital become more of a function of their cost of capital, so independent arbitrage operations are less viable. That’s made it less of a way for people on the periphery of finance to work their way in, and more just another strategy.

Which is not to say the dynamic isn’t still there. It’s still common for people to succeed in business and investing after spending time on something that’s a game with a metagame—sometimes, literally gaming. Shopify’s CEO has praised video games and even hired a former competitive gamer as an intern purely because of his gaming experience. He’s also praised Factorio, saying “It’s the one video game that everyone at Shopify can expense.”[3] Magic: The Gathering is fairly popular among quants, and one of the best players of all time, Jon Finkel, is a managing partner at a hedge fund. Poker, of course, makes a strong showing; it’s a core part of the culture at Susquehanna, a trading firm that happens to own roughly $15bn worth of ByteDance (this is not money they manage; it’s the partners' money, which makes them contenders, in dollar terms for the most successful venture investors of all time).

In fact, the term “metagame” is mostly used in those fields—real-time strategy games require a combination of trainable skills, reaction time, and the ability to identify which playstyles will produce an edge conditional on competing with someone who also has inhuman reaction time. Collectible card games require continuous inferences about what the other player can do next, but also how they plan to win. Since each player is selecting cards from a much larger collection, it’s a field that rewards adversarial R&D—building a strategy specifically to beat the currently dominant strategy.

In all of these games—as in arbitrage—success is ultimately bounded by external rules that are consciously set by third parties. The games are competitive, and the gap between winners and near-winners is minimal. There’s such a thing as a persistent skill advantage, but not a persistent meta advantage, because the metagame is on display every time someone plays.

Which means all these people have succeeded through a three step process:

1. Master the game

2. Master the metagame

3. Master the meta-metagame of applying the skills necessary for #1 and #2 to something with less defined rules but unlimited upside.

[1] The actual story is not relevant to the point, but pretty interesting. A chocolate company Rockwood & Co., had excessive cocoa inventory. They didn’t want to sell the cocoa and pay a dividend, which would have been taxable, so they set up a scheme where they’d exchange cocoa warehouse certificates for stock. Buy a share for $34, exchange it for warehouse receipts, sell the receipts for $36, repeat. In this deal, Buffett says the main transaction cost was subway tokens. The deal was a classic nerd snipe, too: if a company can buy back its stock for less than the value of its easily-liquidated inventory, the right move is not to do the arbitrage but to own the underlying stock. And more generally: if you are offered a deal and feel very smart for accepting, consider how smart your counterparty might be feeling.

[2] Why doesn’t it trade at $40? The boring answer is the time value of money, and the interesting answer is a combination of investors' asymmetric skill and their incentives. If you owned the stock at $30, you were probably betting on the underlying business, which is a different skill set from estimating the odds that an acquirer will successfully complete a deal. And even if you are confident in that, it’s very hard to tell your boss or your outside investors that you owned a stock, the company received an offer, you gambled on the offer going through, and you lost. The gap between where a stock trades once the offer is made and what the value of that offer is represents a fee that less specialized stock-pickers pay to arbitrageurs for the service of handicapping the deal’s odds.

[3] Not that this is an ideal career plan. I have more than one friend whose career trajectory went vertical when they stopped spending most of their free time getting very, very good at video games.

A quick note to readers: overwhelming majority of new readers find The Diff when it’s shared by fans. If you enjoyed this post, please pass it along to someone who seems to be getting a little tired of the latest metagame they’ve mastered.

Share

A Word From Our Sponsors

Here’s a dirty secret: part of equity research consists of being one of the world’s best-paid data-entry professionals. It’s a pain—and a rite of passage—to build a financial model by painstakingly transcribing information from 10-Qs, 10-Ks, presentations, and transcripts. Or, at least, it was: Daloopa uses machine learning and human validation to automatically parse financial statements and other disclosures, creating a continuously-updated, detailed, and accurate model.

If you’ve ever fired up Excel at 8pm and realized you’ll be doing ctrl-c alt-tab alt-e-es-v until well past midnight, you owe it to yourself to check this out.

Elsewhere

I have a new piece in Medium talking about airlines and loyalty program economics. Large US airlines' market values right now are below the value of their loyalty programs, but loyalty economics make it hard to separate the two.

Office Politics, Redux

Earlier this week, Coinbase’s Brian Armstrong published a blog post arguing that the company should focus on its mission, to the exclusion of other social issues. Shortly thereafter, Coinbase offered severance to anyone who disagreed with the stance. (Covered in The Diff here and here.

This has, naturally, led to a great deal of positive and negative feedback. On the negative side, many of the critiques argue that since every company has a political impact, and since many people are deeply passionate about their politics, it’s both hopeless and exclusionary to keep that out of the office. This is a valid point of view; while the original essay explicitly mentions the fact that Coinbase’s aims interact with the political sphere, it’s always possible to extend that further.

On the other hand, one critic, former Twitter CEO Dick Costolo, took the opportunity to fantasize about making a snuff film of people who agreed with the memo being executed. Obviously, this should not be read as a realistic threat of violence—as the history of social media shows, ideas are easy and execution is hard—but it highlights how fraught these discussions can get. If discussions about discussions about politics can prompt public figures to such violent musings, actual conversations about politics are bound to be even more intense. Someone can easily agree with the general point Costolo made and also think “I would not want to deal with the fallout if somebody said that on the office Slack.”

One potentially healthy solution is for more companies to be explicitly political in their aims. The CEO letter in Palantir’s S-1 is an effort in this direction. Palantir doesn’t have to worry as much about political arguments spiraling out of control, because they’ve tried to select a workforce that a) all agrees that a small subset of issues are the most important, and b) is on one side of those issues. (There’s still room for debate, but a debate over “how” is less distracting than a debate over “what.”) This is more or less what Coinbase’s critics have read into the Coinbase post, so there’s little additional cost in actually saying it.

Vaccine as Perk

China’s vaccines are close to formal approval, and are currently being distributed under emergency-use terms. As a result, they’ve become a way to trade favors. Per the New Yorker, which quotes a biotech investor in China:

“Some of these friends used to work at Sinopharm, and they’ve seen people they trust at the company vaccinate themselves,” he said, explaining that such individuals would have early access to clinical results. The investor didn’t find this inappropriate, because participation was voluntary. He pointed out that Gu Fangzhou, the scientist who developed China’s first live polio vaccine, in 1960, had administered it to his infant son before mass trials were carried out. In the United States, Jonas Salk had done the same thing with his own polio vaccine. At the University of Pittsburgh, Salk’s wife and three sons were voluntarily injected in 1953, two years before the vaccine was declared to be safe and effective.

A thirty-four-year-old in Beijing told me that he was offered the C.N.B.G. vaccine because his company engaged in business with Sinopharm. This seems to be a uniquely Chinese addition to the Pittsburgh model—from what I can tell, there’s nothing on the historical record about Jonas Salk building guanxi dose by dose.

This is the retail form of vaccine diplomacy, which will be an important feature of the world in the coming months.

Who Owns User Data?

Facebook has sued two companies that scrape Facebook’s contents through a browser extension. The browser extension market usually involves users getting services and giving away data for free, and this is right at the fuzzy boundary between the data that belongs to the user and data that’s in Facebook’s custody. Everything the extensions are tracking is data that Facebook shows logged-in users, although the extensions generally track other users' information as well.

This is a case where the critique that Facebook owns users' data is actually quite useful: if the users own the data, Facebook doesn’t have any grounds to object, but if users don’t, then the extensions are hoovering up information that belongs to Facebook, not to them.

Tech Sees Like a State, Test-and-Trace Edition

Amazon plans to run 50,000 Covid tests a day for its 1.37m front-line employees, and has identified 19,816 positive cases. Amazon estimates that this is 42% less than the age- and geography-adjusted average for the US. This testing pace is extremely aggressive relative to the US. Amazon expects to test 3.6% of its employees daily, compared to 0.25% for the US as a whole. And because Amazon runs a planned economy—with some very careful planners—it can enact the sort of fine-grained responses that national governments have struggled to.

Conflicts of Interest

BaFin, Germany’s anti-fraud regulator, has banned employees from trading stocks, after the discovery that they had been actively trading Wirecard while it was being investigated for fraud, but before it collapsed. It is, of course, much more newsworthy that they were doing the trading in the first place, and might point to an important difference in the assumptions behind each legal system. US securities regulation tends to assume that, given the opportunity, traders will exploit every possible information asymmetry, and that securities laws need to protect against all the forms of exploitation that cause net harm. In a market where they’re less rapacious, it might take a while before anyone realizes that those trades could happen, and it takes a scandal to ban them.

The Negotiate-With-Hackers Hack

Earlier this week, I linked to a story on why cyber-risk insurers tend to pay ransoms rather than try to recover data. The Treasury Department has since warned that anyone who pays such a ransom needs to ensure that they’re not violating US sanctions. Since it is not especially likely that a North Korean hacker syndicate will go through a standard KYC/AML procedure, this amounts to a ban on paying ransoms for ransomware. Assuming it’s enforced, that’s very exciting news for everyone in the data-recovery business, and a nightmare for insurers.

Small Deals and Stagnation

A common theme, in The Diff and elsewhere, is the observation that public equities markets love any company that can credibly promise a stream of low-volatility cash flow, even if its growth is not spectacular. The latest instantiation of this: big banks are increasingly focused on smaller M&A deals. Large deals have better economics—when they happen. But returns are lumpy, and may not arrive at all. Smaller deals are less risky on a per-deal basis, and pursuing a lot of them adds diversification. When large banks look for smaller deals, they’re admitting that a more predictable but less profitable advisory business is ultimately what their shareholders want.