[neatze]

> Oh wait…every hedge fund bro is already doing this. And most of them aren’t billionaires. The problem is your model needs to include all the computers playing the market, and it also needs to include the other hedge fund bros themselves. This strategy only dominates if you have more compute than the whole market itself, which you don’t.

Best part of this article, honestly, I did not expected this, of-course one can say we just need more abstractions (how does brain/living things build them?), but this would be ignoring dynamical nature of such problems.

[em500]

That argument is a variation of the Efficient Market Hypothesis (EMH), that at any moment market prices already reflect all information available to the public. Or as the old joke goes, two economists are walking down the street and pass by a hundred dollar bill without picking it up. A little while later one turns to the other and asks "was that a hundred dollar bill on the ground?" To which the other replies "it must be fake, if it was real someone would have picked it up already."

Every graduate level economics and finance student is aware of this joke, and also know EMH with more nuance than "you can't make money with trading". There's tons of research and debate about the various forms of the EMH. But you don't have to believe anything anyone else wrote. There's plenty of data and computing software available, so anyone can try everything for themselves.

[gizmo]

With the Efficient Market Hypothesis it doesn't matter what people think because prices end up converging on reality. Everybody crunches SEC filings and stock/bond prices follow from that. That's relatively easy, and requires very little compute.

Hotz is arguing the opposite. His argument is that the market converges to some kind of consensus price, but that price is a combination of what people think a security is actually worth but also to a large part to what participants think other people mistakenly believe a security is worth.

Crunching SEC filings won't get you anywhere anymore. You also have to reverse engineer what other people believe, otherwise you end up waiting for years (or forever) for the market to come to your point of view. And when everybody acts on what they believe other people are thinking you can spend an infinite amount of compute trying to level each other. Unless one party has the majority of compute then they can outcompute everybody else put together.

[neatze]

This argument (at least the way I am reading), goes deep in uncertainty modeling (Partially Observed Stochastic Games the hardest class to optimize for), stock market here used as analogy more then it is not.

[petercooper]

It reminds me of the rocket fuel equation where you have 500kg to take to space, so you need X kg of fuel. Except now you have 500+X kg to take to space and so you need even more fuel. Repeat. This sort of pattern seems to exist in all sorts of diverse situations.

[threeseed]

It's the sort of point someone high at at party comes up with.

No serious person has ever thought that you can simulate the entire stock market.

And the ability to do this is not limited by compute but rather not having the fundamental algorithms that can accurately predict human behaviour.

[neatze]

Not only simulated, but had pending orders on all NYSE stocks both sides, every day for many years.

[gorbachev]

Yes, every hedge fund bro already does that, but they usually target specific scenarios, they don't model the entire market.

The way to make money this way is to find out a unique scenario, and then create a prediction algorithm for that. I'm not close enough to hedge fund bros to know how feasible that is.

[corsix]

Per https://www.stateof.ai/compute, one of the players in the market has ten thousand GPUs in a private cloud. Out-computing just that one player is hard enough, let alone out-computing the whole market.

[bad_alloc]

And it fundamentally ignores that the market is likely totally incomputable in both the short and long term.

[anonymoushn]

This is more or less how some multibillion dollar firms work. An issue with the strategy is that of course you cannot just do this, your up-front costs to acquire the data and connectivity are tens of millions of dollars.