election_title: "Downward monotonicity (San Francisco D7 2020) — BEFORE: Melgar wins"

scenario_description: |-
  The real 2020 San Francisco Board of Supervisors election, District 7 (seven
  candidates), reduced to the final three after four candidates are eliminated and
  their votes transferred — the preference profile from Graham-Squire & McCune,
  arXiv:2301.12075, Table 2. Counted by RCV-IRV: Engardio leads Round 1 with 14119
  first-choices to Melgar's 11652 and Nguyen's 10855. Nguyen is eliminated; his
  ballots transfer and Melgar wins the final 18561-16370. This is the BEFORE half of
  a DOWNWARD-monotonicity pair — Engardio LOSES here. In the AFTER file
  (sf_d7_downward_after), 800 Engardio>Nguyen>Melgar ballots are shifted DOWN to
  Nguyen>Engardio>Melgar — giving the loser Engardio LESS first-place support — and
  Engardio then WINS. Ranking a loser lower makes them win: the downward monotonicity
  paradox. Companion page: downward_monotonicity_sf.md.
  Model note: the Round-1 three-way counts (14119 / 11652 / 10855) and the final
  (Melgar 18561, Engardio 16370) reproduce the paper's published figures exactly;
  Melgar's own second-preference split (not published in the quoted table) is
  reconstructed to complete the three-way, chosen consistent with the paper's stated
  AFTER outcome (Engardio wins once Melgar is eliminated).

video_script: |-
  Three candidates left after earlier eliminations. Round 1: Engardio 14119, Melgar
  11652, Nguyen 10855. Nguyen is out; his votes split and Melgar wins 18561-16370.
  Engardio loses. Now watch the AFTER file: move 800 Engardio voters to rank him
  SECOND instead of first — less support for Engardio — and he wins.

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

ballots: |-
  13319:Engardio>Melgar>Nguyen
  800:Engardio>Nguyen>Melgar
  2500:Melgar>Engardio>Nguyen
  3500:Melgar>Nguyen>Engardio
  5652:Melgar
  6909:Nguyen>Melgar>Engardio
  2251:Nguyen>Engardio>Melgar
  1695:Nguyen

expected_winners:
  - Melgar

# file: sf_d7_downward_before.yaml