We invite you to join a new experiment we cooked up at Yahoo Labs called Predictalot, a game that takes NCAA tournament pick 'em to entirely new extremes.

Here’s how it works. You can invest virtual points on almost any prediction you can think of about the men’s college basketball tournament, like “Duke will advance further than Texas,” or “Every final four team’s name will start with the letter U.” For example, you can invest one point predicting that there will be exactly 2 upsets in the Elite Eight; and, according to the current Predictalot odds, you;ll win ten points if you're right.

You can keep your predictions and see how they turn out, or you can sell them at any time, even while you're following the action on the court. How does that work? Think of your predictions like stocks that can be priced, sold and traded (for virtual points, of course) just like you’d buy and sell stocks, options and futures on the financial exchanges of the world. Predictalot offers hundreds of millions of such predictions, constantly computing prices for them all.

Throughout the tournament, the prices of all the predictions in the system will go up and down based on what’s actually happening in the tournament and on what everyone else thinks those predictions are worth. Based on the value of the predictions you invest in you’ll accumulate points and climb your way to the top of the Predictalot leader board.

In technical terms, Predictalot is what is called a combinatorial prediction market. Some of us at Yahoo Labs, along with academics like George Mason University professor Robin Hanson, have been studying and writing about them for quite some time, and dreaming about trying them out in the real world. We built the first version of Predictalot during an internal Yahoo Hack Day. And this year we used the Yahoo Application Platform to finally build a public version of the game. Like any Yahoo App, you can install Predictalot to play right on the Yahoo homepage. (Note that we're not special in that regard: anyone can develop a Yahoo App that's available to millions of Yahoo users; there's good sample code, it supports YUI and OpenSocial, and it's easy to get started. Seriously, try it.)

After some fits and starts, a few late nights, and eventually a few all nighters, we’re proud and excited to go live with Predictalot, version 1.0. I can’t rave enough about the talent and dedication of the research engineers who gave the game a professional look and feel and some real production speed. They turned a very pie-in-the-sky idea into a reality. We have a lot feature ideas in mind for future versions, but the core is in place and we hope you enjoy the game.

With 9.2 quintillion outcomes, or nearly as many as the number of individual insects on Earth, Predictalot is to our knowledge the largest prediction market ever built. We hope that with your participation it will prove a great test to the limits of what the wisdom of crowds can produce. We'd also like to pave the way for the serious use of combinatorial prediction market technology in a host of other arenas down the road.

Smarter markets will be extremely valuable in a variety of applications, especially online, where markets and exchanges decide prices and allocations for all kinds of goods and services, including advertising inventory. And more and more, the matching of buyers with sellers – which is the core function of any market – is requiring sophisticated algorithms, machine learning and optimization technology. Predictalot attempts to help us better understand these dynamics while also giving us the great satisfaction of picking Duke to go all the way.

(Duke alum) David Pennock and the Predictalot Team:
Mani Abrol, Janet George, Tom Gulik, Mridul Muralidharan, Sudar Muthu, Navneet Nair, Abe Othman, Daniel Reeves, Pras Sarkar

P.S. If you'd like to learn more about the technology hidden in the engine and the scientific factors at play, read our post about Predictalot on the Yahoo Research website. And don’t forget to also check out all of Yahoo's fun tourney experiences.