Henning Makholm's post is something like your second attempt. What follows here is something like your first attempt, a recursive definition analogous to the way $+$ and $\times$ are defined. They key point is that you must define the comparison of larger numbers in terms of the comparison of smaller ones.
$$\begin{align}
0\le y & \text{is true for every $y$} \\
S(x) \le 0 & \text{is false for every $x$} \\
S(x) \le S(y) & \text{is true if and only if $x\le y$}
\end{align}$$
So for example to find whether $S(S(S(0))) \le S(S(0))$ you ask whether $S(S(0))\le S(0)$ and then whether $S(0)\le 0$. This is false by the second clause above, so $S(S(S(0))) \not\le S(S(0))$. The recursion must terminate in every case, because either the left-hand or the right-hand argument must eventually become 0. Note that every possible comparison falls under exactly one of these three cases: Either the left-hand argument is zero, or else the left-hand argument is a successor and the right-hand argument is zero, or both arguments are successors.
The first and second clauses here are like your $bigger(S(0), 0)$ idea, but you need to specify how $0$ compares with every other number in order to get the required base cases.
Once you have this you can define the other relations in terms of it. For example, $x<y$ can be defined to mean $(x\le y) \land \lnot( x= y)$.
This is of course equivalent to the other definition: one has $x\le y$ if and only if $\exists z. x+z=y$.