Cross-sectional mean reversion is a pattern in which assets that have done worse than their peers over a short recent period tend to do better than those peers in the next period, while recent leaders tend to lag. The comparison is across a group of assets at one point in time, which is what cross-sectional means. It differs from ordinary mean reversion, where a single price or spread is compared with its own past average.
A strategy built on the effect ranks a universe of shares by their return over the last day, week or month, buys the weakest and sells short the strongest in equal amounts. In equities this is the short-term reversal effect and it is a core element of statistical arbitrage. Over horizons of several months the sign flips and cross-sectional momentum takes over. The strategy trades often, so transaction costs absorb much of the gross return.