In what you wrote above $h$ is not the probability amplitude and for most states the equation you wrote:
$$
H|\psi\rangle=h|\psi\rangle
$$
It is only true if $|\psi\rangle$ happens to be an eigenstate of the Hamiltonian.
$H$ can be written in the form
$$
H = \sum_j h_j|j\rangle\langle j|
$$
where the $j$s are just numbers that label each possible outcome of a measurement, they could be $1,2,3,4\dots$ or $\dots -1,-2,-3,0,1,2,3\dots$ or soemthing else depending on how it is helpful to label the outcomes. The $h_j$ are the set of possible values of $H$ and they are not limited to be between 0 and 1, unlike probabilities.
The probability amplitude of the $j$th outcome is $\phi_j = \langle j|\psi\rangle$ and has a square magnitude between 0 and 1. By contrast
$$
H|\psi\rangle=\sum_j h_j \phi_j |j\rangle\langle j|
$$
and $h_j \phi_j$ need not have an amplitude between 0 and 1. $h_j$ could be 1,000,000 and $\phi_j = i/2$ in which case their product doesn't have a square amplitude between 0 and 1.
The expectation value is
$$
\langle H\rangle=\langle\psi|H|\psi\rangle=\sum_j h_j |\phi_j|^2.
$$
Each term in that sum is the product of the probability for that outcome and the value of that outcome.
You might want to read a book about quantum mechanics like Christopher Isham's "Lectures on Quantum Theory: Mathematical and Structural Foundations".