For long lists, that isn't more efficient than shuffling.

Regardless of the length of the list, which is quicker depends upon the ratio of picks.

...for a small number of picks, Ie. If the ratio of picks to list size is less than ~15%, spliceing is quicker than shuffling the whole list:

#! perl -slw
use strict;
use List::Util qw[ shuffle ];
use Benchmark qw[ cmpthese ];

our $N //= 20;
our $S //= 8;
our @nums = 0 .. $N;

cmpthese -1, {
   shuffle => q[ my @s = ( shuffle @nums )[ 0 .. $S-1 ]; ],
   splice  => q[ my @s = map splice( @nums, rand( @nums ), 1 ), 1 .. $
+S; ],
};

__END__
c:\test>868601 -N=10 -S=2
            Rate shuffle  splice
shuffle 894654/s      --    -10%
splice  993686/s     11%      --

c:\test>868601 -N=10 -S=3
            Rate  splice shuffle
splice  769417/s      --     -9%
shuffle 844675/s     10%      --

c:\test>868601 -N=100 -S=15
            Rate shuffle  splice
shuffle 196571/s      --     -5%
splice  207873/s      6%      --

c:\test>868601 -N=100 -S=17
            Rate  splice shuffle
splice  186995/s      --     -3%
shuffle 192359/s      3%      --

c:\test>868601 -N=1000 -S=169
           Rate shuffle  splice
shuffle 19552/s      --     -2%
splice  19968/s      2%      --

c:\test>868601 -N=1000 -S=170
           Rate  splice shuffle
splice  20274/s      --     -1%
shuffle 20569/s      1%      --

c:\test>868601 -N=10000 -S=1998
          Rate shuffle  splice
shuffle 1578/s      --     -6%
splice  1674/s      6%      --

c:\test>868601 -N=10000 -S=1999
          Rate  splice shuffle
splice  1601/s      --     -1%
shuffle 1625/s      2%      --
[download]