Sometimes, a Manifold question is isomorphic to a financial asset. For example, the question "What will be the price of the S&P 500 in one year?" is identical to an S&P 500 future dated to one year. On those sorts of questions, is Manifold better-calibrated than financial markets?
This question is non-trivial to answer because I don't know how to extract Manifold-style binary predictions from financial markets (or vice versa). And even if I did, going through a bunch of markets and calculating their calibration would be some amount of work.
I will resolve this question whenever someone (a) comes up with a reasonable idea for how to compare Manifold vs. financial markets and (b) actually does the comparison. Whoever does this will get to make mana by insider-trading on this question.
I earn mana by converting options prices to Manifold odds and trading when they’re out of sync, so I would firmly bet “No,” but I would caveat that there is some complexity to the real financial markets that Manifold gets to skirt. E.g. it typically costs 5%/year (risk free rate) to long a stock. It can cost 100% per year between dividends and short borrow fees to short a stock (e.g. GME in height of frenzy. You’d typically not expect to short for a full year in this case). I would balance that with Manifold having its own cost of capital—loans on the order of 1% per day (3678% per year). In the coming months I’m hoping to set up some bots to get Manifold as close as possible (profitable?) to the the financial markets, at which time the accuracies may converge and we start to see some interesting behavior.