ABSTRACT We develop an equilibrium model of learning by rational traders to reconcile several empirical regularities: Cross sectionally, most individual speculators lose money; large speculators outperform small speculators; past performance positively affects subsequent trade intensity; most new traders lose money and cease speculation; and performance shows persistence. Learning from trading generates substantial endogenous liquidity, reducing bid–ask spreads and the impact of exogenous liquidity shocks on asset prices, but amplifying the effects of real shocks. Introducing slightly overconfident traders increases bid–ask spreads, hurting all traders. Finally, behavioral theories cannot reconcile all of these empirical regularities.