election_title: "Favorite betrayal in STAR (2 of 2) — nine voters demote their favorite and it pays"
bv_test_id: BV2207
bv_election_id: b6xrdr
bv_results_url: https://bettervoting.com/b6xrdr/results
paradoxes: [favorite-betrayal]

scenario_description: |-
  The betrayal — half 2 of the repo's worked STAR favorite-betrayal
  pair. Same 57 voters as half 1
  (bv2206_7mckyg_fbc_honest_tepid_consensus.yaml), except the nine
  Aster-fans now score Aster 4 instead of 5 — strictly below Bluebell on
  their own ballots. Note what they could NOT do: raise Bluebell. She
  was already at 5 — STAR let them equal-top the compromise for free,
  and it wasn't enough, because the problem was Aster's own score total
  keeping Bluebell OUT of the runoff. Demoting him is the only move
  left, and it is a true favorite betrayal (scoring your favorite below
  another candidate). Scores become Clover 72, Bluebell 69, Aster 66 —
  the runoff flips from Aster-vs-Clover to Clover-vs-Bluebell, and the
  Condorcet winner Bluebell takes it 33-18. The nine turned their
  outcome from Clover (their 0) into Bluebell (their 5). The knife-edge
  that makes it a lab specimen, not a strategy: it needs AT LEAST 7 of
  the 9 to coordinate (6 betrayers leave Aster and Bluebell tied at 69
  for the second runoff seat; 5 or fewer changes nothing), plus exact
  knowledge of a 3-point standings gap. This is the ~2% of STAR
  betrayals that pay — the other ~98% backfire (see
  favorite_betrayal_voting_301.md and 06_Other/simulations/).
  Live on BetterVoting (Test ID BV2207): https://bettervoting.com/b6xrdr
  Live results: https://bettervoting.com/b6xrdr/results (BV agrees:
  Bluebell, no tiebreaks).
voting_method: STAR
num_winners: 1
ballots: |-
  Aster,Bluebell,Clover
  9: 4,5,0
  6: 5,0,0
  24: 0,1,0
  18: 0,0,4
expected_winners:
  - Bluebell

# file: bv2207_b6xrdr_fbc_betrayal_pays.yaml