election_title: "Upward monotonicity (Alaska 2022) — AFTER: raise the winner, she loses"

scenario_description: |-
  The BEFORE file (alaska_upward_before) exactly, with ONE change: 7 of the 23
  Palin-only ("bullet") ballots are switched to Peltola>Palin. Nothing else moves.
  This gives the previous winner Peltola MORE first-place support (80 -> 87) — yet
  she now LOSES. Why: the shifted votes pull Palin's first-round count (63 -> 56)
  below Begich's 57, so PALIN is eliminated first instead of Begich; Palin's
  ballots transfer 36 to Begich, and Begich beats Peltola in the final, 93-91.
  Ranking the winner higher made the winner lose — the upward monotonicity paradox
  (Graham-Squire & McCune, arXiv:2301.12075; realrcv.equal.vote/alaska22).
  Model note: at this 200-voter reduction, moving 6 ballots makes an exact 57-57
  Begich/Palin tie (a coarse-model artifact); 7 clears it cleanly for a tie-free,
  engine-independent demonstration. The real election's figure is ~6,000 Palin-only
  ballots, which crosses the threshold comfortably. Companion page:
  upward_monotonicity_alaska.md.

video_script: |-
  Same election as BEFORE — we only moved 7 Palin-only voters up to rank the
  WINNER, Peltola, first. Round 1 now: Peltola 87, Begich 57, Palin 56. Palin is
  out this time, not Begich. Palin's 36 Palin>Begich ballots swing to Begich, and
  Begich wins 93-91. Peltola gained support and lost. That's the paradox: a method
  should never punish a candidate for being ranked higher. STAR and Ranked Robin
  can't do this.

paradoxes: [non-monotonicity]
voting_method: RCV_IRV
num_winners: 1

ballots: |-
  50:Peltola>Begich>Palin
  36:Palin>Begich>Peltola
  29:Begich>Palin>Peltola
  25:Peltola
  16:Palin
  16:Begich>Peltola>Palin
  12:Begich
  7:Peltola>Palin
  5:Peltola>Palin>Begich
  4:Palin>Peltola>Begich

expected_winners:
  - Begich

# file: alaska_upward_after.yaml