Automated Market Makers (AMMs) are used to provide liquidity for combinatorial prediction markets that would otherwise be too thinly traded. They offer both buy and sell prices for any of the doubly exponential many possible securities that the market can offer. The problem of setting those prices is known to be #P-hard for the original and most well-known AMM, the logarithmic market scoring rule (LMSR) market maker [Chen et al., 2008]. We focus on another natural AMM, the Constant Log Utility Market Maker (CLUM). Unlike LMSR, whose worst-case loss bound grows with the number of outcomes, CLUM has constant worst-case loss, allowing the market to add outcomes on the fly and even operate over countably infinite many outcomes, among other features. Simpler versions of CLUM underpin several Decentralized Finance (DeFi) mechanisms including the Uniswap protocol that handles billions of dollars of cryptocurrency trades daily. We first establish the computational complexity of the problem: we prove that pricing securities is #P-hard for CLUM, via a reduction from the model counting 2-SAT problem. In order to make CLUM more practically viable, we propose an approximation algorithm for pricing securities that works with high probability. This algorithm assumes access to an oracle capable of determining the maximum shares purchased of any one outcome and the total number of outcomes that has that maximum amount purchased. We then show that this oracle can be implemented in polynomial time when restricted to interval securities, which are used in designing financial options.
翻译:自动做市商(AMMs)用于为组合预测市场提供流动性,否则这些市场的交易将过于稀疏。它们为市场可提供的双重指数级数量的可能证券提供买入和卖出价格。对于最初且最著名的AMM——对数市场评分规则(LMSR)做市商[Chen et al., 2008],设定这些价格的问题已知是#P-难的。我们关注另一种自然的AMM,即恒定对数效用做市商(CLUM)。与LMSR的最坏情况损失界限随结果数量增长不同,CLUM具有恒定的最坏情况损失,这使得市场能够动态添加结果,甚至可以在可数无限多个结果上运行,此外还有其他特性。CLUM的简化版本支撑着多个去中心化金融(DeFi)机制,包括每日处理数十亿美元加密货币交易的Uniswap协议。我们首先确定了该问题的计算复杂性:通过从模型计数2-SAT问题的归约,我们证明了为CLUM的证券定价是#P-难的。为了使CLUM更具实际可行性,我们提出了一种高概率有效的证券定价近似算法。该算法假设能够访问一个预言机,该预言机能够确定任何单一结果的最大购买份额以及达到该最大购买量的结果总数。然后我们证明,当限制在区间证券(用于设计金融期权)时,该预言机可以在多项式时间内实现。