Algorithmic Competition in Digital Marketplaces: Individual and Market Guarantees

This talk examines how adaptive algorithms influence market outcomes in digital marketplaces. We study two settings where such algorithms are widely deployed. First, we analyse pacing algorithms in repeated ad auctions, which are widely used in practice, and show that they provide individual guarantees while ensuring aggregate welfare that is approximately optimal. Second, we consider time-varying games, which capture the non-stationarities common in online markets. While standard no-regret algorithms are often inadequate in these environments, we show that dynamic no-regret algorithms track equilibria and yield novel efficiency guarantees. Together, these results establish algorithms that deliver both traditional individual guarantees and new market-level guarantees in digital platforms.