election_title: "Equal ranks — the same 74 voters use the whole scale, and Delia wins"
scenario_description: |-
  The twin of equal_rank_cohesive_consecutive.yaml: the SAME 74 voters, the SAME preference order over Alice, Bilal, Cato and Delia — Figure 11 of Théo Delemazure & Dominik Peters, "Generalizing Instant Runoff Voting to Allow Indifferences" (EC'24, arXiv:2404.11407) — with exactly one thing changed. Every voter now uses the full width of the 0-5 scale instead of consecutive numbers: a top class is still 5, but everything the voter is not enthusiastic about drops toward 0.
  Check it: each row below induces the same indifference classes as the matching row in the companion file. Row 1 is still Alice=Bilal=Cato > Delia. Row 5 is still Delia > Alice=Bilal=Cato. Nobody reordered anybody. Only the gaps moved.
  STAR elects DELIA. The companion file, same order, elected Alice. And 38 of these 74 voters rank Alice in their top class, so by the paper's respect-for-cohesive-majorities axiom the winner had to be Alice, Bilal or Cato — Delia is precisely the candidate that cohesive majority excluded. Approval-IRV, which the paper proves satisfies the axiom, elects inside the permitted set on both files. Split-IRV elects Delia on both.
  So this file is the witness for a claim this library should state about its own preferred method rather than wait to be told: STAR does not satisfy respect for cohesive majorities. The axiom is defined on weak orders, a score ballot induces one, and here is a perfectly ordinary 0-5 profile that breaks it. That is not a defect peculiar to STAR — the paper proves no Condorcet method that reads only pairwise margins satisfies the axiom either (Proposition 3.4), so Ranked Robin, Schulze, Ranked Pairs and Minimax are all outside it. Approval-IRV's claim is genuinely unusual, and worth reporting straight.
  One reading note, so the report does not surprise anyone: Bilal and Cato post identical score totals here, so STAR's second finalist slot is settled by lot. It does not touch the result — Delia beats Bilal 30-28 and Cato 30-28 on the same ballots, so the lot picks the loser of the runoff either way.
  For the Approval-IRV and Split-IRV counts, and the axiom check itself, run tools_adam/pref_voting_tabulation_engine/approval_irv_report.py.
paradoxes: [condorcet-cycle, majority-failure]
bv_test_id: BV2297
bv_election_id: j9wvv4
bv_results_url: https://bettervoting.com/j9wvv4/results
voting_method: STAR
num_winners: 1
ballots: |-
  Count:Alice,Bilal,Cato,Delia
  18:5,5,5,0    # Alice, Bilal and Cato equal-first
  10:5,5,0,3    # Alice and Bilal equal-first
  10:5,0,5,3    # Alice and Cato equal-first
  16:0,5,5,5    # Bilal, Cato and Delia equal-first
  20:0,0,0,5    # Delia alone on top
expected_winners: [Delia]

# file: equal_rank_cohesive_wide_gaps.yaml