Trade-based Asset Model using Dynamic Junction Tree for Combinatorial Prediction Markets

Prediction markets have demonstrated their value for aggregating collective expertise [Arrow et al. 2008]. Combinatorial prediction markets allow forecasts not only on base events, but also on conditional and/or Boolean combinations of events [Hanson 2007]. We describe a trade-based combinatorial prediction market asset management system, called Dynamic Asset Cluster (DAC), that improves both time and space efficiency over the method of Sun et al. [2012], which maintains parallel junction trees for assets and probabilities. The basic data structure is the asset block, which compactly represents a set of trades made by a user. A user’s asset model consists of a set of asset blocks representing the user’s entire trade history. A junction tree is created dynamically from the asset blocks to compute a user’s minimum and expected assets.