election_title: "Ossipoff's 303 — the first-round LEADER is eliminated"

scenario_description: |-
  A constructed 5-candidate, 303-voter profile published on rangevoting.org
  (Warren D. Smith's site, section 12, credited to Mike Ossipoff). Reproduced
  here exactly as given, because it is a sharper example than the one it is
  usually filed under.

  The textbook center squeeze eliminates a Condorcet winner who holds FEW first
  choices — that is the whole mechanism, and it lets a defender answer "well,
  nobody actually wanted them first." This profile removes that answer.

  C holds 100 of 303 first choices — the LARGEST first-choice bloc in the
  field, ahead of D (53), B (51), A (50) and E (49). C is also the Condorcet
  winner, and not narrowly: C beats A 202-101, B 202-101, D 201-102 and
  E 201-102 — roughly two to one against every rival. RCV-IRV eliminates C in
  round 3 and elects D.

  The elimination order is what does it. E goes first (49), and every one of
  those ballots reads E>D, so D climbs to 102. A goes next (50), and those
  ballots read A>B, so B climbs to 101. C, who led the entire count on 100, is
  now the lowest of the three left and is cut — at which point C's 100 ballots
  read C>D and hand D the election 202-101.

  Read this fairly. It is CONSTRUCTED, not a real election, so it proves that
  RCV-IRV CAN do this, not that it usually does — Condorcet failures showed up
  in 2 of 182 US RCV elections studied. Its one-dimensional layout (A-B-C-D-E
  left to right, C in the middle) is the standard spatial model rather than a
  transcript of any electorate. And the source page is score-voting advocacy
  with a polemical tone; the PROFILE is arithmetic and reproduces exactly, which
  is why it is here and the rhetoric is not.

  Bare letters are deliberate: this is an abstract published illustration, and
  A-E are the source's own labels.

  BV-CONFIRMED. Race 1 of live election qycpbx (Test ID BV2281) is these 303
  ballots counted by BetterVoting's own IRV tabulator, and it elects D, matching
  this file exactly; race 2 is the same ballots under Ranked Robin and elects C.
  tieBreakType 'none' in both. Frozen export: bv2281_qycpbx_bv_export.json.

  Live on BetterVoting: https://bettervoting.com/qycpbx
  Live results: https://bettervoting.com/qycpbx/results
bv_test_id: BV2281
bv_election_id: qycpbx
bv_results_url: https://bettervoting.com/qycpbx/results
voting_method: RCV_IRV
num_winners: 1

# One-dimensional electorate, A-B-C-D-E left to right; C is the center.
# Source: https://rangevoting.org/rangeVirv.html section 12, credited to
# Mike Ossipoff. Claim-checked in this folder's README.md.
ballots: |-
  50:A>B>C>D>E
  51:B>A>C>D>E
  100:C>D>B>E>A
  53:D>E>C>B>A
  49:E>D>C>B>A

expected_winners:
  - D

# file: bv2281_qycpbx_ossipoff_irv.yaml