Peter Norvig has an essay describing a program to solve sudoku puzzles, even the hardest ones, by combining deterministic logical operations and smart traversal of the possible solutions. The latter is done recursively; here's that function (source):
def search(values):
    "Using depth-first search and propagation, try all possible values."
    if values is False:
        return False ## Failed earlier
    if all( len( values[s]) == 1 for s in squares): 
        return values ## Solved!
    ## Chose the unfilled square s with the fewest possibilities
    _,s = min( (len( values[s]), s) 
                for s in squares 
                if len(values[s]) > 1
            )
    return some( search( assign( values.copy(), s, d)) 
                for d in values[s]
            )
(I've added some spaces, CRs, and tabs for the sake of my eyes; apologies to Dr. Norvig.)
Right below the comment there's a line starting with "_,s".  That seems to be the unpacked tuple (len(values[s]),s) with the minimal value of s.  Is Dr. Norvig using "_" as a variable name just to indicate it's a "don't care" result, or is something else going on?  Are there times when "_" is recommended as a variable name?  In interactive mode, "_" holds the answer of the previous operation; is there a similar function in non-interactive code?
Update
Thanks for the good answers. I guess The Answer goes to Alex Martelli for "value added"; he points out that the "_, vbl_of_interest" idiom is often a side effect of the DSU idiom, which itself has been made largely unnecessary.
 
     
     
     
    