Skip to content

A real BetterVoting election, end to end — "What Makes the Best Pet?"

This is a real STAR election run on BetterVoting (BV id pet): 7 candidates, 461 ballots, single winner. It's the worked example behind the screenshots in the runoff percentages lesson and the BetterVoting ↔ the open-source engine page — here you get the whole thing: the election file and the full report from the tabulation engine this repo runs — Larry Hastings' starvote, extended here (the "LH" engine) — side by side, so you can read a real result from raw ballots to winner.

This is Voting 201 — reading a real, full audit report. (A 101 viewer needs only the last three lines: Scoring Round → Automatic Runoff → winner.) For the section-by-section method, see How to read a STAR report.

1. The scoring round — add every star

The engine sums each candidate's stars across all 461 ballots. The two highest advance:

   Dog           -- 1807 -- First place
   Cat           -- 1750 -- Second place
   Bird          --  978
   Rabbit        --  963
   Fish          --  863
   Rat           --  589
   Python        --  449
 Dog and Cat advance.

Dog and Cat are the Finalists. The other five are out — but their stars still counted (they're what made Dog and Cat the top two).

2. The automatic runoff — majority of those with a preference

Now only the two finalists matter. Each ballot's full vote goes to whichever of Dog/Cat it scored higher:

   Dog           -- 190 -- First place
   Cat           -- 173
   Equal Support --  98
 Dog wins.
   Voters with a preference: 363 of 461 (98 Equal Support).
   Dog 190 (52%) vs Cat 173 (48%); majority = 182.

Dog wins. The line self-reconciles: of 461 cast ballots, 363 had a preference between the finalists and 98 are Equal Support (461 − 98 = 363). The same 190 reads two ways — 190 of all 461 isn't the point; 190 of the 363 with a preference is 52%, clearing the 182-vote majority. The 98 Equal Support voters scored Dog and Cat the same, so they sit out this head-to-head (but counted fully in the scoring round). That two-denominator idea is the whole runoff percentages lesson; the engine prints the decisive line as part of its default report. (The saved _tabulated copy expands it into a "Runoff math" funnel.)

3. Did STAR pick the "right" pet? (the Condorcet check)

The full report's pairwise matrix shows Dog beats every other candidate head-to-head — so Dog is also the Condorcet winner, and STAR's runoff result matches it:

[Condorcet Winner]
  Condorcet Winner: Dog — matches the STAR winner

When the score leader, the runoff winner, and the Condorcet winner are the same candidate (as here), the result is about as uncontroversial as it gets. (For the case where they differ, see three notions of "winner" and the Runoff Reversal walkthrough.)

4. Real-ballot details: abstention vs. explicit zero

Real exports are messy. This one has blanks (-) and all-zero ballots, and the report is careful to separate them:

 Tabulating 461 ballots. Note: 1 of 461 ballots is marked as an abstention.

Exactly one ballot is a true abstention — a fully blank ballot (-,-,-,-,-,-,-), no score for anyone. That is not the same as the handful of all-zeros ballots (0,0,0,0,0,0,0), which are cast ballots that simply support no one. Both tabulate as 0 stars, but a blank is "I didn't weigh in" while an explicit 0 is "I rate you zero" — the engine counts only the blank ballot as an abstention. (The score distribution's Abs column tracks per-candidate blanks the same way.)

Where BetterVoting and the engine disagree (and why it doesn't change the winner)

BetterVoting reports this race as 6 abstentions, 455 ballots tallied; the engine counts all 461 and calls only the 1 blank an abstention. BetterVoting's 6 are all flat ballots (every candidate scored equally) — and two of them are real votes: one voter scored all seven candidates 5, another all 4. Those are Equal Support, not abstentions; treating them as abstentions is what makes BetterVoting's score totals run 9 lower per candidate (5 + 4 = 9) and its Equal Support 92 instead of 98 (92 + 6 = 98). The winner, finalists, and runoff margin are identical. This is documented as a reconciliation/correctness issue:

Run it yourself

From the repo root:

uv run python STARVote_LH_tabulation_engine/starvote_larry_hastings.py "01_STAR/04_Real_Elections/pet_real_bv_election/cases/best_pet_c7_b461.yaml"

Re-running rewrites the _tabulated sibling. Because this is a converted BetterVoting export, it uses the nested election: → races: schema (options live under election.options) — exactly the format the convert → validate → tabulate pipeline produces and cross-checks against BetterVoting's own result.