Case analysis: pre-registering a data-snooping test

Before reading Sullivan, Timmermann and White (1999) beyond its abstract, write down: (a) the null hypothesis of the Reality Check for the BLL rule universe and the benchmark; (b) what result in the 1897–1986 sample would indicate that BLL's best rule survives data snooping, and what result would indicate it does not; (c) what result in the post-1986 sub-sample would indicate a real but decaying effect versus a data-mined one; (d) why the size of the rule universe matters for (b) but not for (c). Then read the paper and compare.