Reproduce “Adaptive Liquidity in Prediction Markets via Online Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Enrique Nueve, Bao Nguyen, Rafael Frongillo, Bo Waggoner
The authors propose an adaptive prediction-market maker that combines several liquidity settings and shifts their weights as trading conditions change. A hybrid signal balances mean-centered slippage, measuring price impact, against uncentered liability, measuring worst-case inventory exposure. In a stylized simulation with four designed trading regimes, liquidity adapts differently across regimes; theoretical guarantees track switching sequences under stated conditions, not market welfare.
Key insights
- It is possible to dynamically adjust liquidity in prediction markets through cost-function mixtures while preserving fundamental economic guarantees like no-arbitrage and bounded worst-case loss by incorporating specific switching fees.
- The trade-off between execution quality (slippage) and market maker inventory risk (liability) can be effectively managed and optimized using a hybrid structural risk signal as an objective for online learning.
- Online learning algorithms can achieve performance guarantees, specifically switching-regret, for adaptive liquidity mechanisms, allowing the market to effectively track the best sequence of liquidity regimes in hindsight.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.