I recently got to spend ten minutes explaining prediction markets to a (nice, smart) software billionaire. He had already been exposed to the basic idea, but from me he came to understand the larger potential for markets on decision consequences. He said they could be useful inside for-profit firms, like hedge funds. I suggested that software firms could also benefit from better estimates on user satisfaction, rates of bugs, and making deadlines. He quickly countered that software visionaries, in charge of implementing an unusual vision, shouldn’t be held to the conventional wisdom of a crowd.
To my knowledge, the Microsoft prediction market was not controlled by management, nor was it widely gamed. Neither did it fail to work. It did, however, fail to catch on, and that is important.
> and whose reasoning would be well guarded information, and with investors that can buy a move, publish something influential, and then sell a move.
Yep. What's the problem? They reveal their information as they buy up shares of that move. That's the whole point!
gwern: "Who could be paid by investors to trade full-time moves" ... and whose reasoning would be well guarded information, and with investors that can buy a move, publish something influential, and then sell a move.
> With four world chess stars to suggest the moves for the world team,
Who could be paid by investors to trade full-time moves in a prediction market, rather than volunteering a little free time; and such a prediction market could draw on every poor world chess star.
> explanations of the moves, a bulletin board,
You don't think prediction markets have bulletin boards or explanations?
> plurality voting,
Yes, that's the point: whatever result they got from plurality voting - you know, the most horrible form of voting in the world? - should be considered a lower bound for the quality of moves which might be selected by a more intelligent voting system. Like prediction markets.
> and no attempts to speculate on moves in any way.
Indeed. A lower bound.
Specifically, a lower bound which gave one of the greatest chess players of all time the hardest-fought game of his life. Your challenge is not to show that voting masses can go toe to toe with a world champ, because that's already been done for you; your challenge is to show that prediction markets would turn in *worse* results than a brain-dead plurality voting result from a massive group of amateurs (and a few experts working for free).
gwern: With four world chess stars to suggest the moves for the world team, explanations of the moves, a bulletin board, plurality voting, and no attempts to speculate on moves in any way.
Doesn't look like market to me. Look at the overhead of operation, too: one man, vs a lot of people wasting a lot of time, and they can't even make a tie (which is an enormous fail as I would really expect that they'd play very very conservatively).
I was wondering if you'd comment on it... Try reading the WP article - "although Kasparov won, he admitted that he had never expended as much effort on any other game in his life" - and reflect on how much better prediction markets would do compared to dumb plurality voting.
So maybe prediction markets needn't do as bad as you assume
What the poster wrote was, "I doubt that play by market price games would have beaten Kasparov."
And Kasparov won the contest.
You're inventing assumptions to reply to, as usual, gwern(0).
Moral: LW custom notwithstanding, straw man arguments aren't the epitome of rationality.
In one of those beautiful incidents which turn out to largely refute a commenter's random claims, this has already been done almost as described, with Kasparov himself: http://en.wikipedia.org/wik...
So maybe prediction markets needn't do as bad as you assume. (Both Kasparov and the voters had access to computers and used them extensively, leveling that particular playing field.)
I like this point. In a world without computers, general prediction markets would have done worse than Kasparov at playing a chess game. Certainly, play by voting games weren't first tier and I doubt that play by market price games would have beaten Kasparov either even if you could have implemented a prediction market that played in real time at 2 hours for 40 moves.
" Alas this seems another example of the usual excuse making;"
Why do you think people make excuses? Why don't you address *that* next time?
MSFT had just the market described here. It failed because it was controlled by the mgmt, with no losses allowed, so was widely gamed. Most orgs have a rent-seeking management agency layer that would be undermined by prediction mkts. In cutting age innovation orgs where principals still are in control, environment is too fluid to structure a market
Like dmytryl said, but for the record, I disagree that near-mode is the more rational mode of thought. Both modes have rational strengths and weaknesses, and rationality is obtained only through their combination. But it's true that those who call themselves "rationalists" around here mean near-mode when they say "rational." Theirs is the shortfall of Monomoniacalist thinking ( http://tinyurl.com/6pt9eq5 )
"...shouldn’t be held to the conventional wisdom of a crowd."
There are perhaps two ways of interpreting this.
1. The notion that "people don't know what they want until you put it in front of them". Radical ideas could be aborted too early if value determined by judgement (including expert judgement) is a poor match for value in use.
2. Personal reputations could be badly damaged if a big decision is made that contradicts the 'advice' of a prediction market, and fails, and much more so than if the same decision had been made sans a relevant prediction market. What effect could prediction markets have on daring strategies or risky plans? Could it result in greater homogeneity of products? What if a manned mission to Mars was decided against based on the outcome of a prediction market?
@37175155a229f3794da539908bc129c1:disqus "what fields lend themselves more to prediction markets and what fields don't?"
Perhaps government activities more than private enterprise? A bigger issue might be how prediction markets affect the status and decision making of owners, managers, boards, investors and politicians. The software billionaire in the post might have an overly romantic view of his visionary capacity, but it's a big call to write off his position (and experience) as making excuses. Costs are not always what we want them to be. On the other hand, prediction markets will probably teach us quite a bit about the psychology of powerful decision makers.
The ones where you can expect people to actually be able to answer. Try opening "prediction market" on solutions to hard math homework, with some scoring points in place of money, compare this to hiring 1 expert that actually can solve this stuff. Or "prediction market" on best chess moves, vs hiring Kasparov.
But this leads to an interesting an question: what fields lend themselves more to prediction markets and what fields don't? And why?
Because he was not making excuses?