election_title: "Same-ish total, different shape — the sandwich vote"
election_description: |-
  A seven-ballot STAR race built to be read BOTH ways: down the columns (how many
  ballots gave each candidate each score) and across the rows (how much of the 0-5
  range each voter used). LH-only — no BetterVoting election.

scenario_description: |-
  Seven people pick a sandwich. The score totals are close — **Avocado 25, Bacon
  23** — but the two candidates got there in opposite ways, and no single number
  in the Scoring Round shows that:

  - **Avocado is polarizing.** Four voters gave it a 5; two gave it a 1 or a 0.
  - **Bacon is the consensus.** Nobody's favourite (no 5s at all), but nobody's
    enemy either (nothing below a 2). Every score is a 2, 3 or 4.

  That difference lives in the **columns** of the ballot grid, which is what the
  LH engine's `[Score Distribution]` block prints.

  Read the **rows** instead and a different fact appears: three voters used the
  whole 0-5 ballot (a range of 5), while voter 4 scored everyone a flat 4 and used
  none of it (a range of 0). That is what BetterVoting's "Range of Scores" chart
  measures — and it cannot be recovered from the score totals or from the score
  distribution, because it is the other margin of the same grid.

  Neither reading changes the result: Avocado and Bacon are the finalists, and
  Avocado takes the runoff 4-2 with one Equal Support ballot.

num_winners: 1
voting_method: STAR

ballots: |-
  Avocado,Bacon,Cheddar
  5,3,0   # used the whole ballot — range 5
  5,4,0   # used the whole ballot — range 5
  5,3,1   # range 4
  4,4,4   # scored everyone the same — range 0
  0,3,5   # used the whole ballot — range 5
  1,4,5   # range 4
  5,2,3   # range 3

expected_winners:
  - Avocado

# file: same_total_different_shape_c3_b7.yaml