The first part of this article examines Patricia O'Grady's recent attempt to identify the method by which Thales might have successfully predicted a solar eclipse. According to O'Grady, some 60% of the potentially visible lunar eclipses were followed 23½ months later by potentially visible solar eclipses. It is shown that this ratio is no more than 23%, and that the method fails to predict after which specific lunar eclipse a solar eclipse will appear. In the second half of the article it is argued that on the basis of his own observations of major solar eclipses, Thales could have concluded that solar eclipses come in clusters of three, the second appearing 17 or 18 months, and the third 35 months, after the first one. In the years after the "predicted" eclipse of 28 May 585 BC, this apparent pattern disappeared again, which would explain why Thales managed to "predict" only one eclipse.