Abstract: The article briefly introduces the basic law and method of statistic test of earthquake prediction efficiency and two statistic evaluation methods, i.e. Z grading of prediction efficiency value and V grading of correlative evaluation. Taking 1996, and 1986 to 2003 consultation ideas from countrywide earthquake trend and 1987 to 1998 monthly consultation ideas from Seismological Bureau of Xinjiang Uygur Autonomous Region as examples, we make statistic evaluation of mid- and short term prediction efficiency. The conclusion shows that having-earthquake prediction of present mid-term (annual) earthquake prediction is valid on the whole, which passes test, i.e. Z₁ equals to 0.2~0.3, and efficient to a certain degree. Non-earthquake prediction passes test on the whole, and Z₀ equals to 0.15 or so, which is less than Z₁. Capacity and level of short-term (monthly) earthquake prediction are very poor, and is up and down around valid prediction and blind prediction.

Key words: Earthquake prediction, Statistic test, Efficiency evaluation, Blind prediction