The notes of dpb and David mean, that the number of possible solutions is too large to be useful, if you want to test all of them. In your case testing all 132 over 66 solutions would take many years.
So you have to concentrate on a small subset. There are different ways to find "good" solutions in a huge set, e.g. randomly:
pos = randperm(1:132, 66);
Other methods of global optimizations allows to improve solutions iteratively from different starting points. This does not require to check all possible solutions, but to find a sufficiently good solution in a bearable amount of time.