election_title: "Hamlin & Hua §4.1 — the same marks read pairwise: 60 of 100 voters express no preference"

scenario_description: |-
  The SAME approval marks as hh41_01_approval_as_printed.yaml, read as a score
  ballot (approved = 5, not approved = 0) so the engine prints the pairwise
  matrix and the runoff line. Magnitude is irrelevant — every head-to-head count
  below depends only on the two-class ORDER — but 5/0 is used rather than 1/0
  because the engine's built-in Approval cross-check reads "a score of 3+ is an
  approval" and would misread a 1/0 ballot. Same convention, same reason, as
  ../../black_curtain/cases/Black_Curtain_01b_c3_b5_dichotomous.yaml.

  This answers the obvious follow-up question: if approval misses the majority
  favorite, would bolting a runoff onto the approval ballot fix it? No — and the
  runoff line says exactly why:

    Voters with a preference: 40 of 100 (60 Equal Support).

  The 60 voters who prefer A to B approved BOTH, so in a head-to-head between
  them they say nothing. B wins the runoff 40-0. The deciding preference wasn't
  outvoted, it was never recorded — the information was lost when the ballot was
  marked, not when it was counted. On these ballots B is legitimately the
  Condorcet winner; on the underlying preferences (hh41_02) A is. That gap is
  the compression.

  Claim-check page: ../../../04_Approval/01_Learn/hamlin_hua_2023.md
  Companion set: ../../black_curtain/condorcet_compression.md
  Set overview: ../README.md

voting_method: STAR
num_winners: 1
paradoxes: [majority-failure, condorcet-winner]

ballots: |-
  Count:A,B,C
  60:5,5,0   # approved A and B — and no preference between them
  30:0,5,5   # approved B and C
  10:0,5,5   # approved C and B

expected_winners:
  - B

# file: hh41_03_marks_read_pairwise.yaml