Overcoming Bias: Was Intrade being manipulated in the last month?

The link is good apropos, but the question that interests me is a bit more general. It seems to me that the apparent failure of Intrade to function as a predictor/knowledge aggregator before the US presidential elections was an important natural experiment. What do you think about the explanations and implications? I don't think it's too hard to discuss the knowledge aggregation issues without being bogged down in the specificities of US politics.

New Comment
18 comments, sorted by Click to highlight new comments since:

N.B.: OB posts are automatically linked on the right sidebar.

So you'd prefer to have the discussion about it on OB? I think there are many good reasons to have it here.

EDIT: Replying to a private message, I wrote a bit more about my reasoning: "As I wrote in the post, the link was just apropos to start a discussion about the issue, here, on LW. I contemplated whether to post it as a [LINK] or as a more generic "open-ended question to the audience". I decided for [LINK] because that more clearly reflects that there's no original content in my post. Should I change it to a more generic title? Do you disagree with posting even that?"

Anyone else think LW is over-moderated?

Edit: The cost of excess moderation is an invisible cost, in the form of people not posting interesting/useful stuff because they're too intimidated. There's no good place for most Overcoming Bias-style blog posts here anymore, for instance, and I see that as a byproduct of excessive moderation (and the fact that there's no good middle ground between Discussion/Main).

Anyone else think LW is over-moderated?

Not me. I think the moderation here is working as it should.

There's no good place for most Overcoming Bias-style blog posts here anymore

I think that's a good thing, for reasons that are confirmed every time I look at the New On OB sidebar.

  1. Yes, the cost of excess moderation is inconspicuous.

  2. This post is perfectly appropriate for the Discussion, I think. More generally, it's fine to discuss things here, even if there are other places on the Web to discuss them. This is because discussion here can be of a higher quality, and is more directly relevant to LWers' interests.

  3. I am not sure whether LW is over-moderated. How would one tell, apart from personally feeling intimidated?

I am not sure whether LW is over-moderated. How would one tell, apart from personally feeling intimidated?

I'm using the word "intimidation" to refer to any kind of aversion to posting more stuff associated with getting downvoted or criticized. Obviously, a huge reason that downvoting does anything is because people don't like getting voted down, and when they do get voted down, they work hard to avoid behaviors resembling the one that caused them to get voted down. They will also tend to avoid behaviors that they saw other people get voted down or criticized for.

Empirically lots of people report feeling intimidated.

I don't think there's a reluctance to spend time writing longer-form stuff: many LWers write blogs. On Less Wrong, though, there is a bit of a culture of "how dare you share that with us", "how dare you put that in Main and not Discussion", "is this really appropriate for Less Wrong", etc. (as seen in this thread and elsewhere). And unlike reddit, where your submission is silently downvoted in to obscurity, on LW it's pretty public that you screwed up since item scores impact visibility less. This leads to a gradually reducing trickle of submissions to Discussion and Main, and then people say things like "Less Wrong is gradually running out of things to talk about" (more like we're all becoming conformists who don't introduce new topics because they "might not be appropriate for Less Wrong").

Potential fix: have a single subreddit (with a less-trafficked "new" section) so that open thread posts/discussion posts/main posts are all part of the same system and the userbase collaboratively decides how much visibility your post gets (which makes screwing up less public, and solves the "discussion vs main" problem for short, speculative essays like you see on Overcoming Bias). In the extreme, go upvotes-only (kinda like Hacker News).

I really think that LW has overcorrected for the eternal september problem--if anything, seems like the quality of Main posts has gone up drastically since the site's creation (with a corresponding decrease in their frequency). Maybe it will be the next generation of online forums that finally get things right :)

I'm using the word "intimidation" to refer to any kind of aversion to posting more stuff associated with getting downvoted or criticized.

You say that like it's a bad thing.

I don't mean to cast it inherently bad or good, I was just trying to define the term. Of course a certain amount is good.

You are right: Lots of people feel intimidated. I was overly skeptical, and missed the evidence staring me in the face. And it's true that lots of people apparently feel confused about the Discussion/Main line of demarcation.

Yep.

Fine as is IMO.

It seems to me that the apparent failure of Intrade to function as a predictor/knowledge aggregator before the US presidential elections was an important natural experiment.

BTW, if you want hard numbers on this - Brier scores and RMSEs, I've calculated them for a number of predictors including Intrade. In general, Intrade was a mediocre predictor, outperformed by most predictors. In the case of the Senate races it came close to Nate Silver, but I have reason to believe that Silver scored poorly on the Senate races since Wang & Holbrook apparently called all Senate races correctly instead of missing 2 like Intrade and Silver.

[-]zaph20

I think this explanation from Wiblin* is most likely: "Nonetheless, I think this is more likely tha[t] a broad pool of Intrade participants [were] being enthusiastic about Romney against all the evidence, and [were] unaware that they could get better odds elsewhere." Is there enough evidence to investigate whether something more sinister is at work? I certainly don't know the details on Intrade and other similar markets, but perhaps there should be more stringent transparency rules to prevent potential manipulations.

Implications: What has always struck me as difficult for prediction markets is the fact that they aren't pricing an underlying "thing"; they're pricing uncertainty itself. Even in a futures or options market, there is an underlying right to purchase at a specific price, even if that right is useless due to the current price of the commodity or stock being lower. While the options or futures contract has a potential value on a certain date, there isn't anything of value being bought and sold on a prediction market. It's just bets based on information that everyone will know on a specific date. So to me, all the pluses you get from other markets that pertain to large groups grappling over how to price good don't apply to prediction markets, because there aren't any underlying assets to price. To me, markets do a good job of rationally arriving at prices because of the constant negotiations and comparisons going on for the underlying assets. People make predictions on where market prices will go, but to me that isn't the same as these being prediction markets. The implication of all this is that for me it's not surprising to hear that there was such a large arbitrage potential between the prediction markets. I'm not such a believer in the efficient market hypothesis to believe that there isn't arbitrage potential in asset markets, but I would predict that these are fewer and smaller than there would be in prediction markets. I don't see this as an unsolvable problem, but to me it shows what prediction markets are good for, which is to keep people honest in their premises and expectations. Take a global warming discussion with a AGW proponent on one side and an AGW skeptic on another. Sans a prediction market, either person could make as dire or as rosy a claim as they like; once they start putting real money on the line, both will likely become more interested in accuracy. If all prediction markets did was to routinely get people to adhere to rational discussion (and thus adhere to Aumann's agreement theorem), arenas such as public policy would improve immensely. So, it doesn't matter that the Intrade bets didn't reflect the best of odds; that just means one of the rational actors hadn't fully adjusted yet; once their account was debited, once would assume that they were in agreement now with the party on the other end of the bet.

*Corrected

As an InTrade bettor, I think a lot of the mispricing is just driven by wrong beliefs. In the primary season, I could usually get good prices to short Ron Paul and I thought it was likely to be because American InTrade bettors (internet savvy, ok with questionably legal things involving money, nerdy) overrepresent Paulites. No manipulation, just sampling problems.

I assumed that demographics of InTrade made it pretty likely they'd overprice marijuana legalization, too, and I had to pull my money out of the Colorado market with a quickness once I'd heard the referendum was actually getting editorial page endorsements. :)

I think this explanation from Hanson is most likely

That post was authored by Robert Wiblin.

[-]zaph10

Thanks, I corrected that.