05 — The ticket web, and the join onto the existing index

Read the roster first, not here. star-voting-library already carries the canonical abstention ticket map — 07_Concepts/tabulation_engines/BV/abstain_issues_index.md — organised by theme (policy · UI mislabel · export ambiguity · tabulation semantics), with BV test IDs, live elections, and the reproduced library cases. This page does not mirror it.

What this page adds is the thing the roster doesn't have: causal attribution. Which half of the rule produces each ticket, and therefore which tickets close together. That is the join, and it is what turns eleven separate triage decisions into one.

---

🔴 First, the thing that is actually blocking #1407

Every evidence link posted on #1407 is a dead 404. The repo was renamed masiarek/YAML → masiarek/star-voting-library; GitHub's repo-level redirect still works, but the 01_Single_winner/ tree was reorganised into 01_STAR/, so the paths no longer resolve.

Checked 2026-08-02, following redirects:

Link posted on	Path	Status
#1407	01_Single_winner/pet_real_bv_election/small_abstention_c2_b5_lesson.md	404
#1407	01_Single_winner/pet_real_bv_election/small_case_abstention_lesson.md	404
#1407	01_Single_winner/runoff_overturns_leader/Runoff_07_flat_ballot_bv_bug_tf73v9.md	404
#1407	01_Single_winner/pet_real_bv_election/LH_BV_reconciliation_issue.md	404
#884	00_start_here/tabulation_engines/BV/abstain_issues_index.md	404 — moved to 07_Concepts/
#884, #1053, #1090	01_STAR/abstain_bugs/bv11_6xhfp8_full_equal_support.md	200 ✅

@jacksonloper said so on the ticket on 2026-07-15 — "getting 404s from those links" — and #1407 has been silent since. The issue is not stalled on disagreement. It is stalled because its entire evidence base is unreachable.

Fixing four URLs is the highest-value, lowest-risk action available on this whole topic, and it costs one comment.

---

The root

Ticket	State	What it is
#884 — Update abstain behavior for STAR	closed	The policy decision: count as abstain "if they're all equal as well (all threes, all zeros, mix of zeros and nulls)". Decided by Sara, Annie, Ruben & Arend. Adam dissented on the record at the time

Because #884 is closed, each consequence gets triaged as a fresh, isolated bug — and each one looks small enough to defer. That is the mechanism by which this has not moved in a year.

Causal attribution — which half of the rule causes what

The rule is two edits (see 04-options.md). Sorting the tickets by which one causes them is what makes the fix schedulable:

Caused by edit (a) — markAllEqualAsAbstention

These close together, with a two-character change plus the R2 prerequisite. None of them is on the existing roster.

Ticket	State	Symptom
#1053	open, relabelled Discussion	5,5 full-support ballot told it abstained; nTallyVotes: 0 and a winner declared
#1407	open	Flat ballots dropped; BV and the reference engine cannot be reconciled
#1052	open, High Priority	"no ballots have been cast" with 3 ballots present
#1065	open	Same, Bloc STAR, three 5,5,5 ballots
#1384	closed	Same again, 5,5 / 4,4 ballots — closed without the rule changing

Caused by edit (b) — the ?? 0 coercion

These need the data-model conversation and cannot close with (a).

Ticket	State	Symptom
#1090	open	Explicit 0,0 "equal opposition" filed as abstention; CSV can't distinguish 0 from blank
#754	closed	Closed on the frontend half only. The backend still drops [0, null] — see 01-the-rule.md
#894	closed	Plurality: "all three votes are abstentions, yet we claim victory"; wrong voter count
#791, #1160	open	Export ambiguity — on the roster; both are downstream of the same coercion

Caused by the denominator, not the rule

Ticket	State	Symptom
#906	open	Average Supporter Profile doesn't reconcile — nTallyVotes is the denominator
#777	open, Discussion	The umbrella: what should the reporting vocabulary be. Option D in 04-options.md is a concrete answer to it

The prerequisite, misfiled as a sibling

Ticket	State	Why it inverts the priority order
#1035	open, marked "low priority"	An independent zero-denominator bug in the runoff pie. Today's abstention rule hides it by making the widget never render in the worst case. Fixing the abstention rule without fixing this ships NaN% onto live results pages. It is a two-line guard and it gates everything else

Adjacent, frequently conflated, actually separate

Ticket	Why it's separate
#1379	Tie-break protocol — BV deliberately skips head-to-head for 3+-way ties. WAI; the LH divergence is lot-vs-random
#1063	Reproducibility of the random rung. Orthogonal — though flat-score elections reach that rung most often
#1432	Display of roundResults.logs. Would make these cases legible either way
#1420	v2 export. Relevant only in that the nAbstentions / nTallyVotes semantics should be documented in v2 whatever the rule becomes

Reverse index — where each ticket is covered

For when you have an issue number and want the analysis. SVL = star-voting-library, BVQA = this repo.

Issue	Roster	Reproduction (SVL)	Analysis (BVQA)
#884	✅ theme A	01_STAR/abstain_bugs/	01-the-rule.md, 04-options.md
#1053	✅ theme B	abstain_bugs/bv11_6xhfp8_full_equal_support.md	03-reporting-anomalies.md
#1090	✅ theme B	abstain_bugs/bv655_jfrk9t_equal_opposition.md	01-the-rule.md (edit (b))
#894	✅ theme D	abstain_bugs/bv1570_6hv7jf_undecided_plurality.md	this page
#791, #1160, #518, #252, #627, #1421	✅	—	not covered here (label/export cluster)
#1407	—	Flat_scores_ties/ cases 07, 08	all five pages
#1052, #1065, #1384	—	Flat_scores_ties_06, _08	02-blast-radius.md, 03-reporting-anomalies.md
#1035	—	Flat_scores_ties_02, _07	03-reporting-anomalies.md § Anomaly 3 — prerequisite
#906, #777	—	—	04-options.md Option D

Net: the roster carries six tickets this analysis doesn't touch (the label/export cluster); this analysis carries eight the roster doesn't (the counts-and-charts cluster). They are complementary, not overlapping — keep both, and keep the roster canonical for "which tickets exist".

The one-paragraph pitch for reopening #884

884 decided a labelling question — "did this voter express a preference?" — and the implementation answered a counting question, by removing the ballot from the tally entirely. Nine tickets since then are consequences of that gap, four of which have been closed without the underlying rule changing. The label may well be right. The deletion is not: it is what produces "0 tallied votes" pages that still name a winner, and it is why BetterVoting results cannot be reconciled against any other STAR implementation. Reopening #884 to separate the two questions would let eight tickets close — and, verified by running BV's own tabulator, without changing the winner of a single single-winner or Bloc STAR election.