This document discusses the opportunities and challenges of enterprise forecasting systems and prediction markets. It provides examples of prediction markets that outperformed official forecasts in companies like HP, EA, and a major pharmaceutical company. However, it also outlines problems in getting prediction markets off the ground, like ensuring sufficient participation and actionable results, and sustaining long-term interest due to attrition and a lack of community. The document proposes that Xpree is working to address these challenges through their forecasting system, such as allowing for accurate results with moderate participation and an easy to use interface that encourages spontaneous participation.
Recently joined Xpree, but have been in prediction markets for 10+ yearsCaltech (Bossaerts, Fine, Ledyard liquidity paper) It’s not just the population and the assets, it’s the mechanism that matters. Seminal work in that space.BRAIN as response to 1998 Plott & Chen work.Designed, architected, and implemented BRAIN from 2002 to 2008A lot of people ask my why I left HP. I’ve been in this space for a long time, but until recently hadn’t seen anyone that really understood the problems that make enterprise forecasting fundamentally different than public markets. The first attempts in this space had an ‘if we build it they will come’ mentality. While there has been a lot of good press, the traction was lacking. I joined Xpree because of their single-minded focus on delivering the promise of PMs in a way that fits the corporate need.Today, like everyone else, I’ll talk about some successes. But, what I personally think is more interesting is where PMs have failed to deliver. We’ll go through a few of the major stumbling points, and then discuss what Xpree has cooking to meet this challenge.
5 minutes on these. I can talk for days on this. No notes needed.
Population: There is always a tension on the size of the population in corporate PMs. On the one hand, we want lots of voices in the chorus. On the other, there are a number of factors. (1) We don’t want to tie up lots of people unnecessarily; (2) many decisions are made by relatively small groups; (3) as PMs move to a more strategic role in the company, the sensitivity of the questions prohibits wide inclusion.Results: Point estimates are not useful for many types of strategic planning. Many clients ask, “What would I do differently if I knew that? Will I know far enough in advance to react?”Mechanisms: while many clients appreciate the analogies to the stock market, many also fear the complexities of financial instruments. On the one hand we have stock-market like mechanisms that require what many people in this room might consider basic understanding of markets (shorting, covered calls, etc), which intiidate quickly. On the other, we have extremely straightforward, unidimensional, slider mechanisms. But, they make the message space muddy. For example, there is no way to communicate that you feel very strongly that a number is a little low vs. weakly that it’s way too low.Some systems, like BRAIN, get much of their power from calibration that requires synchronous training.
Assume we meet every challenge on the previous slide and the clients engage in a PM implementation.The kickoff is usually big. But, after traders place a few trades and poke around a bit, the novelty wears off and the attrition is massive. Why?We feel that this is largely due to the fact that the systems focus on feedback to the end client. They fail to realize that for a PM to thrive, the network effects must be strong, and if we don’t create an engaging user experience, the client’s needs won’t be served. PMs systems, while they have often included comments, don’t create a sense of community, any kind of team-based spirit, or sense of importance of voice to the user.Because many markets have long horizons and also because often markets are all launched in bursts, there is sparse activity. Even if there is enough betting in the aggregate, the lack of quick change discourages others and the market whithers.Users, like the clients, find mechanisms and payoffs confusing. More traditional stock market mechanisms have too many levers for the average user. Simpler, MSR based systems, are usually either based on mid-point or start-price trading. These hide much of the mechanism, but also make the valuations more mysterious. Our experience shows that if you can’t explain how to play the game in 1-2 minutes, you’ve lost the user.In light of the issues with using real cash, both wrt the CFTC and to employing AMMs, prizing has become the best practice in PMs. Those that just reward the best players magnify risk attitudes (go big or go home), and those that use lotteries are often perceived as unfair by those in the lead. Both of these systems leave little of interest for the relatively uninformed liquidity trader, who soon stops playing. Bye, bye liquidity.
Xpree has spent a great deal of the last year thinking about these challenges. It is our mission to bring the promise of PMs to companies, and build it in to the corporate culture with a well-supported, social ecosystem of prediction, delivering measurable ROI for the client.Our new mechanism at once is simple and feels like a watercooler bet with a friend, but also provides an extrordinarily rich message space for the user. Unlike traditional market mechanisms where convergence requires many iterations, a buy-and-hold strategy in the Xpree X2 system conveys a great deal of information all the way until closer.The simple UI allows players to visualize the implications of their bets, and to view their bet in a simple, declarative statement. The payoff is stable until close, although interim sales are allowed and priced efficiently.Etc…..
Here is a screenshot from X2. On this page, a user sees the bets he is currently holding. As you can see, the player is able to forecast over a range of possible outcomes. He places a bet and has a guaranteed payoff if he’s right. He is able to view the current crowd’s perceived probability that he is right, and therefore the current market value of his bet.