election_title: "Minimal tilted cycle — 5 voters, margins 3–1–1 (RCV-IRV)"

scenario_description: |-
  The same five ballots as tilted_cycle_c3_b5_rr.yaml — the minimal tilted
  Condorcet cycle from Brandt, Dong & Peters (arXiv:2411.19857), Fig. 1 —
  counted by instant runoff instead.

  Ben holds just one first choice, so IRV eliminates him first; his ballot
  (Ben > Cara > Ada) transfers to Cara, who wins 3–2 over Ada.

  Worth noting what IRV does NOT see: Ada's 4–1 crushing of Ben is the largest
  majority anywhere on these ballots, and it never enters the count. IRV reads
  first preferences; the cycle lives in the pairwise margins.

  In a cycle there is no Condorcet winner, so IRV cannot be accused of missing
  one — this is a legitimate answer, not a failure. It simply lands on a
  different candidate than Copeland's margins tiebreak (Ada) and than leximin
  (Ada), and agrees with Nanson (Cara). Five voters, four defensible answers:
  that is what a cycle costs. See the folder README.

paradoxes: [condorcet-cycle]

voting_method: RCV-IRV
num_winners: 1

ballots: |-
  2:Ada>Ben>Cara
  1:Ben>Cara>Ada
  2:Cara>Ada>Ben

expected_winners:
  - Cara