We can model each voter's preferences by assigning a real number, called a score, on the interval $[0, 1]$ to each candidate. The goal of each voter is to elect a candidate with the maximal score according to their own preferences.
Voters are allowed to coordinate and the voting system doesn't need to be deterministic. My intuition tells me approval voting where ties are broken uniformly at random but I'm unsure.
Edit: For clarity, all voters vote as tactically as possible. Voters will only compromise if doing so will ensure/increase the probability of a better outcome. Omniscience also applies to knowing how each voter fills out their ballot.