The problem is that you don't have a vector, although you think that you do. Actually y is a scalar value -1, so of course min(-1) will give you that output. This happens because you simply reallocate y on every loop iteration, so at the end of the loop it only has the last value. You could use indexing to fix this.
A much better way to solve this problem is entirely without loops:
>> X = linspace(-pi,pi,10);
>> Y = cos(X);
>> [maxY,idx] = max(Y);
>> maxX = X(idx)
maxX = -0.34907
>> maxY
maxY = 0.93969
Loops are often an inefficient way of solving problems in MATLAB.
