Oceans Yet to Burn

[Oceans Yet to Burn]

neoliberalism-nightly asked:

I think the problem with that math post is that you are distorting the description to match with your ethical beliefs. And I'm pretty sure you are aware, and I may have missed it in the post, but a crucial aspect of the prediction market is to have poor predictors lose their voting power and vice versa. And this generalizes to markets when you use the words associated with scarcity, and you can generalize it to an evolutionary biology view as well.

nostalgebraist answered:

I am aware of this argument – for instance in the paper I linked there is this remark

If long-run market forces lead those with a history of accurate evaluation to become wealthier, then this wealth-weighted average may be a more accurate predictor than an unweighted average.

But I would not say this is a “crucial aspect” of prediction markets, from what I know.  There is a lot of interest, say, in the relatively few currently existing prediction markets, even though these are both rare and “causal” enough that one would not expect to see much of a wealth signal in them due to people going broke or winning big in prediction markets (as opposed to elsewhere).

If this were necessary for the function of prediction markets, or just a crucially important element of the mix, I would expect to see much more of a “we need to wait for the data” attitude among people studying these markets, rather than an intensive use of the data from the currently existing markets to evaluate the prediction market concept itself (as we see in that paper, and in many other places as well).

So we seem to have a real disagreement about how important this mechanism is (or is believed to be).  I’m not saying it’s seen as unimportant because of my values, I really just haven’t seen people treat it as that important.

I also don’t see how this generalizes to non-prediction markets, since there isn’t a notion analogous to “good prediction” for preferences, or if there were it would be something like “being exceedingly average in preferences” and I don’t think there’s a mechanism making people wealthy in proportion to how average they are.

[mitigatedchaos]

It’s probably more important if the market accumulation is broken up by domains, with some secondary importance across domains at a broader level for superforecasters participating.

It may not be that the megapredictor should have almost all the wealth, but rather we might want to allow a capped range or something depending on past performance.  Sounds like a good topic for a study!