The OP is very interesting and I'd like this discussion to gain momentum.

How would it work in the case of music? How is the "bet" won or lost?

This question hasn't been answered satisfyingly so far IMO.

There is no end date because the only way to exit your bet is to sell your position to someone else. Thus the market price is always the rate at which you win or lose based upon your starting market price.

As I understand prediction markets, there *has* to be an end date. A prediction ends when the event to be predicted takes place. The result should be easily verifiable by anyone after the event. These conditions are not fulfilled by the OP's proposition.

I guess bytemaster's idea will above all lead to speculative bubbles. The nature of such a prediction market reminds me of a pyramid scheme.

Even if it would work out, I fear that the viral nature of this approach rewards mainstream a lot more compared to niches than today's music industry does. The participant has a higher incentive to bet on mainstream than on the long tail (even if the participant is only listening to free jazz himself). The opportunity cost for discovering 'good music' in niches is much higher. I think there should be some kind of reputation system to reward those that pay a constant effort to dig up music that certain groups of people like. Maybe some kind of copy trading could provide those discoverers with a following and help the long tail to pay off.

But still: I doubt the functionality of such a prediction market.