It's been a decade since I've had real analysis so cut me some slack
Assume a season ranking exists for all player and a player has a current rank, r. Also, let us assume a player level, p. Then, the opponent selection pool will consist of the members with following attributes:
r-x<r<r+x, x is the rank radius
p-n<p<p+n, n is the selected level range
If the count of the set formed by the above conditions is less than the desired number, increase x by 1 until that number is achieved.
*#1 ranked player: this player will have the (possibly) maximum r with a bottom heavy set.
*Bottom ranked player: this player will have the (possibly) maximum r with a top heavy set.
*Top performing individuals within each p-n<p<p+n group: these players will have the potential to reach Legend I, level restrictions will need to be applied to brackets in order to gate rewards.
Additionally, a time element could be applied so the same player could not be attacked more than once by the same attacker in h hours.