Hacker News Reader

Comment by taylorhughes | Hacker News Reader

Hi HN, I helped build Forecast. We're a small team within Facebook's New Product Experimentation group. We think prediction markets have the potential to create structured conversation spaces, to drive constructive discussion about what might happen in the world in a time when that's pretty hard to find. Please give the product a try and let us know if you have feedback!

View on HN

Interesting.

I ran a project in this area for a few years, and saw significant issues. For example, for high-stakes predictions (eg elections) we saw significant gaming because people see it as an opportunity to move the media narrative.

We've seen this in the current US Presidential Elections, with big swings in betting markets for the presidential outcome itself, without the same swings in related outcomes (eg, you see a swing in the price for "the winning president is X" without the price for "X gets above 270 electoral votes changing at all")

Obviously this creates arbitrage opportunities, but the media narrative is the real issue. Have you considered this problem at all?

Edit: Here's an account of manipulation of betting markets during the Obama/Romney election: https://rajivsethi.blogspot.com/2013/09/the-romney-whale.htm...

This is something we're thinking about continuously; it's unclear how central the prediction %s themselves should be. I am hopeful our core community can become more accurate over time, and get better and better at surfacing new information to make the market more efficient. We'll be evaluating this throughout...

That's good, but I suspect continual monitoring of headline markets ("who will win the US Election") is insufficient because FB is big enough to influence the result.

What are you going to do if (when) you detect manipulation? You won't be able to take it offline without the President using it as an example of "liberal bias", and you won't be able to ban users because of the same thing.

I know you won't do this, but I think you should experiment with some lower-stake markets first, and potentially use participation in those as a gate to the high stakes markets.

Gating could be a good idea; I think also leaning less on the aggregate/average and more on the positions from reliable participants in the market would be important once there's any kind of influence or scale to this product.

These are really really important points and the fact that Facebook is launching products without solid and convincing answers to these issues is a big worry.

This is a fascinating product and potentially of great value but the size and visibility of Facebook means that the potential for unintended consequences is BIG.

Hi Taylor, this is one of the cooler (probably coolest) consumer products to come out of facebook in a while. I was following it closely since the announcement a few months back.

My question is: I am sure you guys are familiar with crypto-based prediction markets like augur.net and prediqt.everipedia.org that use smart contracts and real money for predictions. Given that facebook is also building Libra and getting into blockchain, do you guys see that down the line Forecast could be something that moves to facebook's blockchain and/or you guys collaborate with crypto-based products like the above mentioned? That would be truly exciting since it would bring pure market dynamics to the predictions. I'm aware of the regulatory hurdles there, but thought I'd ask anyway. Great job! This is so fun to use.

Thanks! Right now we're focused more on building the community and usability of the product, so the market implementation itself is fairly simple. But we're following the various crypto implementations closely...

Community is always the most important aspect, agreed! Great to hear you guys are watching the crypto prediction markets closely, they are coming up with some novel features. I'm sure they'd all be interested in working with you guys at Facebook too given the users they could get through a partnership. Good luck!

Are the crypto-based markets legal?

Hi Taylor! A few things:

• In Chrome, the website log in system fails silently (and we get "FB.login() called when user is already connected" when retrying).

• The website is blank on Firefox.

• The principle is cool, but not Bayesian enough IMO. Points are expendable (especially when one can get them freely by creating a new account). Better use statistical priors estimating the likelihood that a given source outputs an accurate prediction. Give a bad prior to new accounts, to discourage creating new accounts to "clean the slate" of bad predictions.

• In the current state, this system, if it gets popular, can quickly have diplomatic, political and economic repercussions and consequent abuse that the website seems wholly unprepared for, based on my reading of the guidelines.

> • The website is blank on Firefox.

Do you have Firefox FB containers enabled? I had the same issue, but when I opened the website inside a FB container it works.

I was in a FB container. I tried both while logged off and while logged on Facebook.

The iOS app feels a bit sluggish. For instance scrolling a list will trigger hover state on list items.

What technology is used for the mobile apps?

I predict that it's React Native

bingo

Hi,

Has there been any risk assessment on this feature? If so, any chance this is made public?

Hey Taylor, cool product! But why can't you see the purchase prices of the arguments? As the price moves around over time you can try to look at the date and graph independently, but some indicator showing the movement since purchase on each of everyone's arguments would be a lot more useful! Cheers.

Hey Taylor, and here I was just wondering what NPE was and finding the thought that Facebook farmed out an app to another company amusing. Anyway, your website is trash, but the app is pretty nice and I'm enjoying it so far.

I agree with the other comment that you should use probabilities and not binary choices. Have you looked at metaculus?

The share prices reflect implied probabilities. You’re not required to hold them until expiry.

Is it SOP for the NPE Team to publish to the App Store as "NPE Team", and not Facebook?

Yes, separate developer account (lots of benefits, eg. no shared credentials) but our apps are all transparent about being created by Facebook Inc.

I see nothing on the App Store page that mentions Facebook.

Why the decision to launch for iOS only?

Reply on news.ycombinator.com