An academic analysis like what you linked to has to assume everybody have settled on the same definitions and valuations.

See my other comment describing what a perfect answer looks like.

I solve huge NP-hard problems every day, I've never seen any suggestion there is a useful way to solve marking pricing.

You have not seen any suggestion to solve this problem, does not mean it isn't NP-hard.

>> we just might not have to computing power to calculate it.

There is a perfect answer, and there will almost never be enough computing power to calculate it. It is always approximated.

So you'll have more "perfect" answers than people, and no way to select any one of them.

It's not a valid argument against the fact that everyone does have their marginal benefit and marginal costs for every good and service that can translate to a binary decision of whether they would purchase a good or service at a given price.

Just because you can't see it doesn't mean it's not there. And I'm not saying it's impossible to find out individuals marginal benefits and costs, because in markets you very well can. Stock traders, when evaluating whether a stock would go up or down, would often buy a bunch of stock, see how the market takes it, then sell a bunch, and by measuring how fast the market eats his orders, he can estimate the aggregate marginal benefit and cost of holding the stock, and then speculate accordingly.

I think this fact is more like an axiom.

With NP-hard problems like Sudoku, that's easy. Slightly simplifying, the definition of NP-hard is "easy to check the answer, hard to find it".

EDIT: Added rabbit ears for simple.

You don't need to solve an NP hard problem to beat human intervention. You just have to be faster and more accurate.

I do think that computers could outperform humans on the supply side, calculating the costs of production relative to competitors better than humans. The demand side is trickier but again computers would do a better job calculating available cash and then ranking marginal goods' likelihood of being purchased based on existing prices.

As long as prices and production processes are public knowledge computers will do better than people at modeling the economy. Humans only gain an advantage through insider knowledge and trade secrets.

Is it? I mean, I see how it's reflexive (and so chaotic) but NP-hard is the halting problem: given a program and it's input, decide if it halts. E.g. what is the input to your market function? It's like asking if the world halts.

The thing about all complexity analysis is that it ignores input/output.

Just that its NP-hard doesn't mean it's as hard as the halting problem, it could be easily decidable in exponential time.

Complexity analysis is all about the size of the input, it doesn't ignore it.