election_title: Exercise 15a, second look — the same Yes/No ballots read pairwise

scenario_description: |-
  The SAME 35 ballots as ex15_approval_yes_no.yaml, with each approval written
  as a 5 and each non-approval as a 0, so the engine prints the pairwise matrix
  and names the Condorcet winner. The magnitudes carry no information here —
  every head-to-head count depends only on the two-class order (approved above
  not-approved), which is what makes this a DICHOTOMOUS PROFILE.

  The payoff: the head-to-head order reproduces the approval order exactly.
  Blair beats Cosmo 13-7 (15 voters approved both, so they express no
  preference) and beats Ada 20-7; Cosmo beats Ada 15-8. Blair > Cosmo > Ada,
  the same ranking as the approval totals 28 > 22 > 15.

  That is guaranteed, not luck. On a dichotomous profile "more voters strictly
  prefer x to y" reduces to "more voters approve x than y", so a Condorcet
  winner always exists and approval voting agrees with every Condorcet method.
  The caveat lives one folder over: real approval ballots are compressed from
  richer opinions, and the Condorcet winner of the uncompressed opinions can be
  someone else — see method_comparisons/black_curtain/condorcet_compression.md.

  Exercise page: ../ex15_read_the_ballot.md
  Live results: https://bettervoting.com/d4v2dh/results (BV2258, race 2 of 2 —
  the same 35 ballots as race 1, cast as 5/0 scores; BV's STAR count agrees:
  Blair. The frozen export sits beside the sibling yaml,
  ex15_approval_yes_no_bv_export.json — one export covers both races.)

bv_test_id: BV2258
bv_election_id: d4v2dh
bv_results_url: https://bettervoting.com/d4v2dh/results

voting_method: STAR
num_winners: 1

ballots: |-
  Count:Ada,Blair,Cosmo
  15:0,5,5
  8:5,5,0
  7:5,0,5
  5:0,5,0

expected_winners:
  - Blair

# file: ex15_approval_pairwise.yaml