EvidenceBasedProgramming would always seek real evidence that a development practice gives real productivity/quality benefits before recommending it.
Practices with evidence for gains so far include:
Fagin Inspections
Pair Programming
Two Person Offices
Hopefully XP will come up with real evidence for the neneficience of many of its pratcices and combinations of practices. --TomAyerst (with inspiration from ChrisCottee)
(If anyone can come up with more practices and/or citations for the above it would be appreciated)
EvidenceBasedProgramming