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!

I’ve read posts recently by Brad Feld and Sigurd Rinde that discuss horizontal versus vertical markets. This discussion seems particularly relevant to prediction markets.

Brad Feld writes that he loves to get involved with companies that develop horizontal technologies. Specifically,

I personally revel in the tension between engineering, product marketing, and sales. It’s challenging to sell a horizontal product in an emerging market, especially to enterprise buyers. You need a special type of salesperson – one who doesn’t constantly say “I can’t sell that” but instead knows how to ask his early adopter customers “what is your pain around [problem category X]?” and then figures out how to repurpose the horizontal technology to solve this pain.

In a well functioning early stage team, this tension between engineering, product marketing, and sales results in customer driven vertical market segmentation. Rather than having a top down sales effort to “go after the following five vertical markets”, the use cases evolve and the ripe vertical markets become clear.

Sig follows up with:

I can “see” how the “broad horizontal technology” is left spreading slowly over the horizontal plane until it finds cracks of specific needs, the verticals where it can freely flow into. Cheaper, better, efficient, unobtrusive, smart as methods goes.

This is exactly what we’ve seen in the prediction market industry, and it’s an efficient model. Prediction markets are used horizontally in a number of different industries. Pharmaceuticals, manufacturing, R&D, retail, and more.

However, as the industry has started to mature, more vertical markets are developing. Pharmaceuticals in particular have really taken up prediction markets, as their exposure to risk is so large that even marginal improvements can save significant costs. Major banks have the same issues. The Hollywood Stock Exchange has had great success in using their platform for a variety of media-related prediction markets.

I think prediction markets are starting to see a turning point, and this will develop in the next few years. As one company in a vertical market gains a competitive advantage through introducing a prediction market, other companies will follow. The key for early adopters is to be just that: on the leading edge of the technology and implementation.

I’m happy to say that if you’re interested in prediction markets, there are currently two conferences taking place quite shortly that you should know about.

I talked about the London prediction market conference last week, which takes place on October 11-12. There will be a wide variety of participants, including software vendors (NewsFutures, gexID, and nosco), prediction market end users (Microsoft, Financial Times, and Nokia), academics (from Nottingham Trent University and University of Copenhagen) as well as inTrade and myself. I really look forward to seeing everyone here in London!

Dave Perry announced last week that ConsensusPoint is hosting a prediction markets conference in New York City on September 24th! I will be speaking at this event as well, and James Surowecki (author of the Wisdom of Crowds) is the main keynote speaker. The majority of the other speakers are customers of ConsensusPoint that will be sharing their prediction market experiences. (Alex Kirtland of Usable Markets will be the other non-end user speaker.) It should be another exciting and informative event.

If you’re interested in prediction markets on either side of the Atlantic, there are now two great opportunities to learn more about them and hear about the benefits from end users themselves. I sincerely hope you take advantage of at least one of them.

I’m happy to say that there has been a good interest shown in the London Prediction Markets Conference, which will be taking place on October 11-12th. Based on the speaker list alone, this should be an exciting event.

Conferences are really valuable in bringing people in the prediction market industry together. (In that sense, the breaks are the most important part of any event!) You meet the various software vendors (such as NewsFutures and ConsensusPoint), academics doing research in the field, and representatives from companies that have actually used prediction markets in their business (like Google and Microsoft). (Note that the above list are companies that have presented in the past. The specific list for the London conference can be found my clicking the banner below.)

To those who are interested, I’ll be speaking about my experience running two years’ worth of small prediction markets. My focus on the research end has been on how few participants you need in order to have calibrated predictions, and the behaviours observed by participants. The last of my markets (Rowing markets, hosted by Inkling) are finishing in the next week and a half so I will have data from those, as well.

The link above will take you directly to the site where you’ll be able to find information on the speakers, as well as register. I hope to see some of you there!

Update: I edited the second paragraph for clarity.

As we move into what I hope will be a great August weekend for you, I wanted to provide a run-down of some of the more popular articles on Mercury’s Blog, courtesy of Google Analytics.

5 Top “Hands-on” posts to get your Prediction Market project started:

1- Structuring your market
This post talks about the scope of different options you have when assembling the structure of your prediction market.  I discuss continuous double auction (CDA), algorithms (such as MSR or DPM), and other options.

2- Developing a business case for Prediction Markets
In this post from February I discuss three different broad uses of prediction markets, and how to develop a business case for each.

3- Setting Initial Conditions for Prediction Markets
In this post I wanted to impress upon readers the various considerations for a prediction market’s “Initial Conditions,” the initial stake provided to traders and timescales of contracts.

4- Motivating the ranks
To have a successful prediction market means that you need to get people to participate.  This post talks about how you can make that happen.

5- Forecasting Error Calculator
My apologies if you’ve had trouble accessing this before, but the problem should be fixed.  Follow this link to download an Excel Forecasting Error Calculator, that should help you determine how much your current forecasting errors are costing you, and generate a business case for starting a prediction market.

5 Other top posts - Prediction Market design and Industry commentary

• Notes from the San Diego Prediction Markets Conference
• Prediction Markets for the Stock Market? Huh?
• Thinking about Real Money Trading
• Educating Traders - Including Lessons from Betfair
• Industries Ripe for Prediction Markets

Thank you to all of you for reading, and I look forward to posting more on these and other topics in the coming weeks and months.

—————–

I’d like to make a quick reminder that you can click on this link to receive new Mercury’s Blog posts directly in your e-mail inbox.  Or if you’re into RSS feeds, you can click here to add Mercury’s Blog to your feed reader.

Probability is a dangerous topic in discussing prediction markets.  Many people don’t really understand probability and thus see some of the prediction market “misses” as “failures” instead.  Some of these have been well-publicised, such as the InTrade market for the Senate in 2006, and widely criticised.  But there are two issues that we need to dis-entangle here:

1. The only way to evaluate accuracy of predictions is with a sufficient group or series of predictions.
2. Most prediction market “failures” are one-off events.

Chris Masse at MidasOracle is trying to say that the Karl Rove resignation market at NewsFutures wasn’t “predictive” because it was trading at ~20% when the announcement was made.  This is horse-(manure).  Based on this argument, at what point does a market become “predictive”?  Only when it trades above 50%?  Or perhaps when it trades above 80%, or some other figure?

There is no way to evaluate accuracy of a single binary prediction because it either happens or it doesn’t.  Evaluation requires lots and lots of predictions; only then you can start determining accuracy.  If all the events that were judged to have a 20% chance of occurring actually occur 20% of the time, then the market is calibrated and accurate.  (If five people were all judged to have a 20% chance of resigning and one of them actually did, the judgments would be calibrated.)

This runs into the second issue above: most “failures,” to include the 2006 Senate market, the Pope prediction market, Olympics choice market, and more are one-off events.  They are generally “misses” or “failures” because they a) occur so infrequently that determining accuracy as described above isn’t possible and b) because each time they’re run traders have to learn all over again what signals and information are important.  I believe that all of these markets could be accurate if they were run frequently enough that traders could simply learn from their mistakes, which is a key feedback in a normal market.  When a Pope is elected every 10-30 years, it’s very difficult to re-learn papal politics to trade effectively, and even harder to determine if it’s effective because it won’t occur again for another 10-30 years.  Since that won’t happen, prediction markets are still a good way of aggregating opinions about any of these events happening.

Even election markets fall prey to this phenomenon, though the attention paid to elections means they are traditionally fairly accurate, though I don’t remember seeing any paper on the 2006 elections specifically.  The issue here sometimes becomes one of timescale.  In politics there is a big risk of a candidate doing or saying something stupid pretty much up to the last minute, so traders factor that into their prices, leaving a favourite at 80%, when perhaps they should be at 95%.  That gap gets made up only once they really can’t say anything stupid anymore, such as the day of the election itself.

In summary, recognise that a single prediction is just that: the traders’ aggregated opinion of the likelihood of that event occurring.  Once enough of these judgements are put together, then the accuracy can be determined, and only then.  Remember that an event that only has a 1% chance of happening will still actually take place one in a hundred chances!

——————-

I’d like to also point out David Pennock’s blog post here for more detail.  Calibration, the test described above, is a good test, but not the only test of accuracy.  There are more statistical tests that should be run (again, only with a sufficient number of predictions), but even these are only useful when comparing two prediction methods against each other.