Skip to content

BV2105-r2 — the partial ballot, re-counted a year later

Method: Bloc STAR (multi-winner, majoritarian) · 2 seats · Expected winners: Chocolate, Strawberry · full count →

A deliberate re-run of BV2105 (r4dqvd) on exactly the same four ballots, cast again on 2026-08-04 so they are counted by today's tabulator. The question it settles: a year on, does BetterVoting still file a real partial ballot as an abstention? It does. The fresh election returns nTallyVotes 2 / nAbstentions 2, identical to 2025.

What this case is NOT about. The library used to attribute this miscount to bettervoting#1056. That was a mis-citation, corrected 2026-08-04. #1056 is a different defect on the same demo election — a 401 blocking JSON/CSV download and Race Details, introduced by the Editable Ballots work (#979) and correctly closed via #1058. They share only the BV2105 test-document name. The counting defect on this page is a separate bug, filed 2026-08-04 as #1478 on the strength of this election (report archive).

▶ Live on BetterVoting: vote · results ↗ (election w3vvff).

Reference file: bv2105r2_w3vvff_ice_cream_recheck.yaml. Frozen export: bv2105r2_w3vvff_ice_cream_recheck_bv_export.json.

Why a second election was needed

The obvious move is to re-fetch r4dqvd and read its numbers. That doesn't work, and the reason is worth stating because it applies to every closed BV election in this library:

r4dqvd is closed. Re-fetching it today does return nTallyVotes 2 / nAbstentions 2 — but a closed election's stored ElectionResult may simply be the tally computed back in 2025. The re-fetch cannot distinguish "the bug is still live" from "we are reading a year-old result." Only ballots cast through today's tabulator can.

No other case in the library answers it either. The discriminating ballot is one whose non-blank marks are all equal — here a single 1 — because that is exactly what #884's all-equal rule treats as an abstention. Sweeping every frozen export in the repo turns up only one other 2026-minted election with a partial ballot, BV215 (26khr3), and its partial is Ada 5, Bruno 1, blank — two distinct marks, so it is counted either way and settles nothing.

Hence a fresh mint, same ballots.

The election

Bloc STAR, 3 flavors, 2 seats, 4 ballots — one of each kind of ballot:

Vanilla,Chocolate,Strawberry
5,5,5     an all-5s ballot (loves everything)
-,-,-     fully blank — a TRUE abstention
1,-,-     Vanilla=1, the rest blank — a REAL (partial) vote   ← the probe
2,5,4     a full ballot

Winners Chocolate, Strawberry, and they were never in question — Chocolate takes seat 1, seat 2 is a Vanilla/Strawberry runoff tie broken by score (Strawberry 9 > Vanilla 8). The winner path is not what this election measures. The count is.

The result: still broken

Quantity BV2105 (r4dqvd, 2025) BV2105-r2 (w3vvff, today) LH engine
Winners Chocolate, Strawberry Chocolate, Strawberry Chocolate, Strawberry ✓
nTallyVotes 2 2 3 ✗
nAbstentions 2 2 1 ✗
Vanilla score 3 3 total 8, avg 2.7 ✗

The two BetterVoting columns are identical. The 1,-,- ballot is still filed as an abstention alongside the genuinely blank one, and Vanilla's reported figure is still an average over 2 ballots rather than 3 — the single 1 never enters the sum. (BV floors the average for display: Strawberry's 9/2 shows as 4, Vanilla's 7/2 as 3.)

Is this a bug, or the documented policy?

Worth stating fairly, because it cuts against the simple reading. #884 established that a ballot whose marks are all equal counts as an abstention. A ballot bearing a single mark is trivially all-equal, so it falls through the same test — meaning this may be the policy working exactly as written, not a coding slip.

The argument that it is nonetheless wrong: BetterVoting's own tally treats a blank as 0. So Vanilla 1, blank, blank is not a ballot with one mark and two unknowns — it is a ballot reading Vanilla 1, Chocolate 0, Strawberry 0, which strictly prefers Vanilla to both. Under the all-equal rule as applied to the scores it actually counts, that ballot is not all-equal at all, and it plainly expresses a preference. Note the scope: this concerns partial ballots whose non-blank marks are all equal. Partial ballots with two or more distinct marks count correctly.

The LH report (the correct count)

[Divergence from STAR]
  STAR                   = Chocolate
  Choose-One (Plurality) = Vanilla   (differs from STAR)
  RCV-IRV                = Vanilla   (differs from STAR)
  Note: 1 of 4 ballots (25%) had equal non-zero scores, so their ranks were
        decided by candidate priority order. The RCV-IRV result may be an
        artifact of score-to-rank tie-breaking rather than a deep
        difference.
  Note: Ranked Robin (RCV-RR) agrees with STAR, so RCV-IRV is the lone
        outlier — the classic center-squeeze signature.
  Full round-by-round reports (generated for review):
  RCV-IRV rounds: cases_tabulated/bv2105r2_w3vvff_ice_cream_recheck_RCV-IRV_tabulated.txt

--- Bloc STAR Voting Method (2 winners) ---

[Bloc STAR]
 Tabulating 4 ballots to fill 2 seats. Note: 1 of 4 ballots is marked as an abstention.
Vanilla,Chocolate,Strawberry
      5,        5,         5
      -,        -,         -
      1,        -,         -
      2,        5,         4
  ('-' = left blank / abstained; '0' = scored zero — both count as 0 stars.)

[Bloc STAR: Round 1: Scoring Round]
 The two highest-scoring candidates advance to the next round.
   Chocolate     -- 10 -- First place
   Strawberry    --  9 -- Second place
   Vanilla       --  8
 Chocolate and Strawberry advance.

[Bloc STAR: Round 1: Automatic Runoff Round]
 The candidate preferred in the most head-to-head matchups wins.
   Chocolate     -- 1 -- First place
   Strawberry    -- 0
   Equal Support -- 3
 Chocolate wins.
   Runoff math:
     4  ballots cast
   − 3  Equal Support (no preference between the two finalists)
     ─
     1  voters with a preference  (majority = 1)
           Chocolate 1 (100%)  ·  Strawberry 0 (0%)

──────────────────────────────────────────────────

[Bloc STAR: Round 2: Scoring Round]
 The two highest-scoring candidates advance to the next round.
   Strawberry    -- 9 -- First place
   Vanilla       -- 8 -- Second place
 Strawberry and Vanilla advance.

[Bloc STAR: Round 2: Automatic Runoff Round]
 The candidate preferred in the most head-to-head matchups wins.
   Strawberry    -- 1 -- Tied for first place
   Vanilla       -- 1 -- Tied for first place
   Equal Support -- 2
 There's a two-way tie for first.

[Bloc STAR: Round 2: Automatic Runoff Round: First tiebreaker]
 The highest-scoring candidate wins.
   Strawberry    -- 9 -- First place
   Vanilla       -- 8
 Strawberry wins.

[Bloc STAR: Winners — Bloc STAR Voting Method (2 winners)]
 Chocolate
 Strawberry

LH counts 4 ballots, 1 abstention — only the fully-blank row — and gives Vanilla a real 1, for a total of 8.

Does it change a winner?

Not here. The discarded ballot only helped Vanilla, who loses seat 2 either way, so this is a reporting/counting defect rather than a wrong result. But the ballot was cast, and a cast ballot dropped from the tally can flip a closer election — which is the whole reason the case is kept rather than filed as cosmetic.