As a Technique for Creating Equivalent Groups, When Is Matching

As a technique for creating equivalent groups, when is matching preferred over random assignment?

A) when a large number of subjects are available and can be used
B) when some extraneous variable is known to correlate with the dependent variable
C) whenever a within-subjects design is preferred over a between-subjects design
D) whenever a potential confound exists, but you aren't aware of its presence