Earthquake prediction aims to specify the time, location, and magnitude of a future quake, distinguishing itself from probabilistic long-term forecasting and real-time early warning systems. While scientists in the 1970s were highly optimistic about finding a reliable method, by the 1990s continued failures led many to deem it impossible, and no large earthquake has ever been demonstrably and successfully predicted. The most famous claimed success, the 1975 Haicheng earthquake, was later found to have had no valid short-term prediction, and extensive searches for reliable precursors have proven fruitless. Evaluation of predictions requires statistical testing against random chance, but earthquake clustering complicates this—a naive method based on clustering can "predict" 5% of quakes, highlighting the difficulty of beating the null hypothesis. The stakes are high, as failures have led to political purges (e.g., after the 1976 Tangshan earthquake) and legal convictions (e.g., the 2009 L'Aquila scientists for giving undue assurance). Consequently, while a minority holds that prediction might be possible with enough resources, most scientists are pessimistic or believe it is inherently impossible.