Mercury’s Blog

If you're new here, you may want to get Mercury's Blog by Email or subscribe to my RSS feed. Thanks for visiting!

In my last post on the New York prediction market conference, I discussed the morning speakers. I’d like to finish my thoughts on the day by discussing the afternoon speakers.

Brent Stinski- Media Predict

Brent talked about his prediction market company, Media Predict. They had extensive coverage of their launch in the New York Times and other media outlets. He reported that over 5000 people signed up on the site, though didn’t go into details. Though the initial press focused on Project Publish, the idea where people compete to predict which book will be published by Simon & Schuster by the contest entries, that is not the sole focus of the site. So far, he’s had a promising start.

Ricardo dos Santos- Qualcomm

Ricardo discussed how Qualcomm has used prediction markets internally, in what I would call “Idea Selection.” In new business development, he is always seeking new ideas and new businesses for Qualcomm. They developed an internal process to seek out new ideas, rate the best ideas, and then a prediction market to determine the best of the best ideas. They had hundreds of people participate, and had good results. It was my impression that they are still experimenting with the technology, but have so far had positive responses from their trial.

Christina LaComb- GE Global Research

GE started this project back in 2005, and is using prediction markets for “Idea Selection.” In their case, employees propose ideas which are turned into the contracts to trade. The results of the market order the ideas, and they’ve experimented with what to do with ideas from there. (For example, automatically fund the top idea, or send the top 5 ideas to business managers for further development, etc.)

Steven Sachs- CNBC

Steven couldn’t discuss much of what CNBC is doing, as it’s still in development. It sounds like they are building a public-facing market, and have run internal markets to demonstrate the technology. I look forward to hearing more about it as they get ready to release a product.

———————–

Panel Discussion, moderated by Alex Kirtland

Mat Fogarty- Xpree (formerly EA)
Fortune Elkins- Misys
Brent Stinski- MediaPredict
Ricardo dos Santos- Qualcomm
Steven Sachs- CNBC
Dave Perry- ConsensusPoint

———————–

Robin Hanson-

Robin gave a quick addendum to his earlier talk, where he focused on the cost-value space of a prediction market. He described an evolution from betting markets, which have negative cost (aka profit) though little value to an organisation, to future prediction markets. Fully evolved prediction markets will certainly have a cost to operate, but the output could have tremendous value to a company.

Me-

It was my turn to speak next. I spoke about my research that was published early this year in the Journal of Prediction Markets. Please contact me if you’d like any copies of my slides.

James Surowiecki- author of “The Wisdom of Crowds”

First of all, James is a fantastic speaker. (It was rather humbling to speak just prior to him.) He spoke about prediction markets in a broad sense and made a number of excellent points.

To start off, bureaucracies and hierarchies and bad ways to move information around in an organisation. At the same time, collective intelligence is still quite counter-intuitive. As humans, we have a deep-seated impulse to find the one person with the answer. (This explains the proliferation of prediction market sites that then seek to sell the predictions of the market leaders, which can lead to disastrous results. ~Jed)

What James came back to is that prediction markets need to be measured against other alternatives for information and decisions, and not absolute accuracy. If they provide an organisation a better way of making a decision, why not use them?

His last points were on the topic of leadership and decision-making. Again, humans tend to associate the concept of leadership with the concept of decision-making. Historically, a strong leader doesn’t put too much stake in his follower’s thoughts and opinions; there’s a reason he/she leads. However, John Craven perhaps took the best or at least most unique approach to this problem when it came to the challenge of finding the Scorpion submarine, lost at sea. He developed the methodology which used the intelligence and experience of many professionals to develop the Bayesian map of where the Scorpion could be on the ocean floor. Craven however led the implementation of that methodology through to a very successful conclusion, finding the sub just a few hundred metres away from the expected location, out of thousands of square miles of ocean floor.

Again, James was an excellent speaker, and I highly encourage you to go hear him if you ever get the chance.

———————–

So this is the end of my notes on the New York prediction market conference. I’d like to make a quick reminder that you can subscribe to this blog via e-mail, either by entering your e-mail address in the sidebar to the right or by clicking on this link.

My apologies for taking a week to write this up, but I had some other commitments back in the US, and I only just returned to London.

First of all, congratulations to Dave Perry and ConsensusPoint for assembling so many interesting speakers and attendees into a beautiful venue (the Tribeca Grand Hotel in New York.) With people like Robin Hanson and James Surowiecki anchoring the morning and afternoon, it was bound to be a great event.

Here are some of my notes from the day:

Robin Hanson-

Robin gave a fairly standard introduction to prediction markets lecture that some may have seen at other events or downloaded from his website. It was a good overview of the topic.

The question and answer period was the most interesting part with Robin. He was asked about manipulation, and provided some fairly convincing answers that manipulation shouldn’t be a worry (at least with the correct incentives.) Robin described the situation in terms of sheep and wolves. Sheep aren’t that knowledgeable; they are trading for any number of reasons, and are the “noise” in the marketplace. Wolves take advantage of that, and consequently they look for markets with lots of sheep. With better information, the wolves will easily have plenty to “eat.” The net result is that those noisy markets are accurate markets.

Another concept he talked about was creating a “fudge” account. Let’s say you want to weight one set of traders more than another, or simply want to “move” the forecast in one direction or another; create a “fudge” account to conduct those transactions. If after the account has been running for a while and it’s positive, you’ll know you’ve done a good job fudging. But if the fudge account is negative, you don’t know more than the market so just stop fudging and leave the market to itself. It’s a great idea, and fairly easy to implement.

Dave Perry-

Dave talked about ConsensusPoint and prediction markets in general. One of the more exciting clients they are dealing with is BestBuy, which is rolling out prediction markets to every single employee, all the way down to the hourly workers in the retail stores. BestBuy is making a strong push for prediction markets, and it will be exciting to hear about the results. (I’m very bullish for them.) That they developed an internal DVD to help promote the markets showed the dedication they have to the project.

Dave also showed off a demo site with their prediction market software. The newest development is that they have now implemented three different trading interfaces: Simple, Standard and Advanced. The Standard is their traditional interface, where you choose to buy or sell a certain number of shares or a dollar value at which to purchase shares. The Advanced interface shows both the buy and sell order books so you have more information when making a trade. The Simple interface just asks questions about how a person views the contract; ie, do they “believe it is likely” to occur. They are then led to select how many shares or dollars to spend. While it doesn’t quite match the ideal world I’ve discussed before, it’s a great step towards accommodating people with different backgrounds, who may or may not understand “trading” on a financial market.

When discussing BestBuy, they had a couple of instances where the prediction markets were particularly valuable. One was where a forecasting team was trading against their official forecast, demonstrating the power of anonymity in the markets! The second case was where a potential problem showed up by a dropping price in the market, even though the executive dashboard for the project was “green” across the board; the market showed the problem a couple of weeks before the official reports made their way to the executive suite.

Finally, Dave stated two great reasons why prediction markets are used. One, is that it pays people for their knowledge of their company and their industry. Two, it also pays people to “shut up.” Markets are self-selecting and very equal; a person can be a wallflower in a conference room and not be heard, but they will be heard on a prediction market. The influence of the stereotypical loud, brash, know-it-all is greatly minimised.

Fortune Elkins- Misys

Fortune is very passionate both about prediction markets and her company, Misys. In her company, as in most others, “uncertainty costs money.” Fortune has been successfully testing prediction markets at Misys, and has found that they have been working quite well. She’s also had some interesting results, where the person that made massive winnings by (correctly) betting against the crowd was an outsider to the project in question, but someone with extensive knowledge of the company itself. It shows that anyone can be successful in a market, so long as they use the knowledge, experience and resources they have within the company.

Karim Tarwahi- MyCurrency

Karim is a former derivatives trader that has founded MyCurrency. One of their first products is a website (homepredict.com) that attempts to forecast housing prices by individual house and by zip code. This functionality has the ConsensusPoint software at its core, though through a totally custom-designed interface. One of the things they want to do is provide a forum where real estate agents can prove that they know what they’re talking about, and then leverage that reputation. Again, while not strictly a prediction market, MyCurrency is using forecasting as a key component of social networking and reputation building.

—————-

So those are my notes and thoughts on the morning portion of the conference. I hope you find them helpful and informative. Please contact me if you have any questions, and I’ll be posting the notes from the afternoon portion of the day later this week.