f2 = @(p)f(p,x);
That relies on x existing in your workspace. If you are going to do that you might as well assume it on the previous line, and use cleaner code.
But your bigger problem is that you are failing to calculate a sum-of-squared errors
f = @(p) p(1) * (0.5 - (1 ./ (1 + exp(p(2)*(x-p(3)))))) + p(4) .* x + p(5);
f2 = @(p) sum((f(p) - y).^2);
p0 = [0.1 0.1 0.1 0.0001 0.1];
p = fminsearch(f2, p0);
