BV2132 — Pet poll: four methods, one electorate, three winners¶
Our own answer to Equal.Vote's meta_pets ("What Makes the Best Pet?"), rebuilt small and tuned so the methods actually disagree. The same 22 voters rank the same three pets (Dog, Cat, Fish), tallied four ways in one BetterVoting election. The result splits three ways — and the candidate a majority actually prefers wins only the two "good" methods.
▶ Live on BetterVoting: vote · results ↗ (election ykjjhy, Test ID BV2132).
Frozen export: bv2132_ykjjhy_pet_poll_four_methods_bv_export.json.
The electorate¶
Twenty-two voters in three blocs, each with a clear preference order:
9 voters: Dog > Cat > Fish (Dog partisans)
7 voters: Fish > Cat > Dog (Fish partisans)
6 voters: Cat > Fish > Dog (Cat first)
Cat is the Condorcet winner — beats Dog 13–9 and Fish 15–7 head-to-head — but has the fewest first choices (6). That single fact is what makes the methods diverge.
Four methods, three winners¶
| Race | Reads | Winner | Our engine | BetterVoting | Read · run |
|---|---|---|---|---|---|
| Plurality (choose one) | first choices | Dog | Dog | Dog | page · …_plurality.yaml |
| RCV-IRV | ranked, elimination | Fish | Fish | Fish | page · …_irv.yaml |
| Approval | approve any | Cat | Cat | Cat | page · …_approval.yaml |
| STAR | score 0–5 | Cat | Cat | Cat | page · …_star.yaml |
Every race's winner is cross-checked: our LH engine (starvote_larry_hastings.py) and BetterVoting's tabulator agree on all four.
Why they split¶
- Plurality → Dog. Dog has the most first choices (9), so choose-one hands it the win — even though 13 of 22 voters rank Dog last. Classic first-past-the-post: a polarizing plurality beats a broadly-liked majority. Cat, the consensus, is a spoiler'd also-ran with 6 first choices.
- RCV-IRV → Fish. IRV looks past first choices, but it eliminates the fewest-first-choice candidate each round — that's Cat (6). Cat's ballots transfer to Fish, which then beats Dog 13–9. The Condorcet winner is squeezed out before the final round — a textbook center squeeze.
- Approval → Cat and STAR → Cat. Both let voters express support beyond one name, so Cat's broad acceptability shows up: Cat is approved by 22 of 22 non-Dog-partisans, and in STAR advances on score (78) then wins the runoff 15–7. These two methods recover the Condorcet winner.
So the candidate a majority actually prefers head-to-head (Cat) wins Approval and STAR and loses Plurality and IRV — the whole argument for cardinal methods in one 22-voter poll.
Per-race engine reports¶
Each race's full LH tabulation is in its _tabulated mirror:
- Plurality →
…_plurality_tabulated.txt(Dog 9, Fish 7, Cat 6) - RCV-IRV →
…_irv_tabulated.txt(Cat eliminated R1 → Fish 13, Dog 9) - Approval →
…_approval_tabulated.txt(Cat 22, Fish 13, Dog 9) - STAR →
…_star_tabulated.txt(Cat 78 & Fish 62 advance; Cat wins runoff 15–7)
Provenance¶
Created via create_bv_test_election.py as a single multi-race BV election (ykjjhy) with made-up ballots — the first multi-race case built through the script. BetterVoting's Results for all four races match our engines exactly.
The pets set¶
- BV2132 — Pet poll (three winners) — this one (Cat wins Approval and STAR).
- BV2133 — Pet poll II (four winners) — a 4th pet splits Approval from STAR too, so all four methods disagree.
- Pets Governance (five methods, LH-only) — multi-winner: majoritarian sweep vs proportional representation.
See also¶
- Folder overview: pet_poll_four_methods — README
- The single-method center-squeeze counterpart: BV2131 — Tennessee (Ranked Robin)
- The real inspiration: Equal.Vote's
meta_pets("What Makes the Best Pet?"), whose four methods barely diverge — this case tunes the ballots so they do. - Condorcet efficiency — topic hub · Glossary