Hilbert Matrix, Gaussian Elimination with varying pivot strategies, and computation error.

Question

I'm doing a project for my Numerical Analysis class about computational error related to Gaussian elimination, gaussian elimination with partial pivoting, and gaussian elimination with scaled partial pivoting. One exercise is to use each method on a Hilbert Matrix or size n=10 and n=20. In theory, the amount of error should decrease as I go through the methods: GE -> GE+Partial Pivoting -> Ge+Scaled Partial Pivoting, but for the Hilbert matrix the opposite seems to be true. The least error occurs for just normal GE. Have I done something wrong or is this how Hilbert matrices behave?