Thursday links: Crypto deal-making, futarchy, and predictable randomness

The art of the crypto deal

A Wall Street Journal article on the looming collapse of a Nasdaq-listed crypto payments and digital asset treasury company — AI Financial — highlights the extraordinary economics of the Trump family’s crypto project, World Liberty Financial.

Formerly known as Alt5 Sigma, AI Financial recently “flagged risks about its survival” after reporting a quarterly loss of $271 million, primarily due to the decline of the World Liberty Financial governance token WLFI.

WLFI is AI Financial’s primary asset, acquired in two tranches: First, World Liberty received a controlling stake in AI Financial in return for $750 million worth of WLFI.

The company then raised money from investors to buy an additional $750 million of WLFI, directly from World Liberty.

As reported on President Trump's federal financial disclosure, the second transaction triggered a $540 million payment to the Trump family — because World Liberty's operating agreement entitles another Trump-controlled entity (DT Marks DEFI LLC) to 75% of the proceeds whenever World Liberty sells WLFI.

I believe this is unprecedented in financial history. 

How could it not be? Imagine Apple had an agreement that whenever it sold equity to investors, 75% of the proceeds went to the estate of Steve Jobs. 

Safe to say, it would never sell any equity: A new investor in Apple would have to immediately write down 75% of their investment.

Tokens aren’t equities, of course, so the analogy is not exact (there’s no reason to write-down the value of a governance token because they have no intrinsic value to begin with). 

The effect, however, appears to be the same. 

“The deals saddled AI Financial with an enormous pot of Trump cryptocurrency that since has slumped 70% in value,” The Journal reports.

That 70% decline is suspiciously close to the Trump family’s 75% rake on WLFI. It’s probably a coincidence — I doubt token markets are as efficient as that. 

But it’s probably instructive, too.

The Journal concludes that AI Financial has served “mostly to channel more than half a billion dollars to the Trumps."

Again: AI Financial is a Nasdaq-listed company. 

The Maine use case for futarchy

By dropping out of the race for a Maine Senate seat, Graham Platner has inadvertently illustrated the case for futarchy — the idea that contingent prediction markets can guide decision making.

As the odds of Platner withdrawing shot higher this week, the odds of Democrats winning Susan Collins’ Senate seat rose as well. 

That is useful information to have!

More useful still would’ve been having that same information before he dropped out.

A prediction market contingent on his dropping out (a “decision market,” Alex Tabarrok calls it) would have told us that, without Platner on the ballot, the odds of a Democrat winning Collins’ Senate seat would be about 10 percentage points higher.

Now that Platner has dropped out, 600 delegates of the Democratic party of Maine have another decision to make: Who should replace him as their nominee for the US Senate?

Wouldn’t they like to know who would be most likely to win?

The only way to know with some degree of science is futarchy: prediction market odds on who would win the general election conditional on each possible nominee.

The delegates might ignore that advice and choose someone less likely to win anyway — but they’d at least know the consequences of their decision.

Markets are predictably unpredictable

A new paper argues that, however unpredictable they may seem, financial markets are not a random walk.

Instead, markets create an illusion of randomness — by hiding information, thwarting theory with reality (transaction costs, mostly), and continuously changing in response to the people trying to predict them.

Yet beneath the illusion, Miquel Noguer i Alonso notes that markets remain governed by cause and effect: information arrives, people trade, prices move. That underlying structure shows up in higher-order patterns — volatility clustering, shifting liquidity, “tail thickness” — that reflect a hidden order to prices.

This structure, the paper argues, remains "predictable to the owner of a private signal, a faster pipe, a better model, or a risk tolerance others do not possess."

A second paper offers an example of what a "better model" might look like. Researchers found that markets systematically underreact to certain kinds of corporate news, then trained an AI to exploit that delay. In doing so, the AI inadvertently constructed a signal that foreshadowed bitcoin's 2021 rally:

Amazingly, the AI constructed a portfolio of stocks (the orange line above) that presaged the price of bitcoin — without looking at the market for bitcoin. The signal was based entirely on patterns discovered in ordinary financial news (available to everyone).

I take that as evidence that markets contain many more hidden structures than we've yet imagined — and that AI will be good at finding them.

But Alonso’s work suggests the edge AI finds will prove temporary. As traders exploit new patterns, prices adjust, the patterns weaken, and the search begins again.

The first paper ends on a paradox: Markets are governed by cause and effect, making them theoretically predictable. Yet because every successful prediction changes the market itself, they remain, in practice, unpredictable.

In other words: The game will never end.

— Byron Gilliam