> every single one of billions of trades, someone is always wrong... every person I have spoken to in finance clams up immediately when I mention anything remot

> every single one of billions of trades, someone is always wrong... every person I have spoken to in finance clams up immediately when I mention anything remotely quantitative. It's not just because I'm being boring, they know they are playing a game of poker that is never not-on. No serious financial professional would expect their strategy or any strategy in existence to be 100% correct. A strategy that is correct 51% of the time is more than good enough. You diversify and hedge, and build up an uncorrelated portfolio. Suppose you've honed in on alpha strategy. You probably don't even want to optimize and fit in further, with past data. Instead, you find a beta strategy that is uncorrelated and run those two together. With the amount of capital firms the likes of Citadel are working with, it's impossible to trade in a single product, and so scaling up, in the case of an MM, is to trade everything all at once. This is also repeated in the article. "[CitSec CEO] diversifies the firm’s earnings so it can better stomach an occasional loss in any part of its business" > current strategies are not information problems that need new science, they are work problems that piece together solved ones, and whose efficiencies come from competency in related domains Applied math is not about pushing mathematical theory, necessarily. Then there is a difference between theory and practice, and there is a difference between practice and sizing trades with this scale of AUM. Academics can talk and write papers about LLMs or CRDTs, but they aren't going to be the ones who will be affective at bringing it to market.