Skip to content

05_Ranked_Robin — Ranked Robin (RCV-RR / Copeland)

Rank the candidates. Compare every pair head-to-head. Whoever beats the most rivals wins.

A Ranked Robin ballot: five candidates — Andre, Blake, Carmen, David, Erin — each with columns 1st through 5th. Instructions at top: rank candidates in order of preference, equal ranks are allowed, candidates left blank are ranked last. This voter marks Andre 1st, Carmen and David both 2nd (an equal rank), Blake 3rd and Erin 4th. Footer: candidates are compared head-to-head in round robin matchups, and the candidate preferred over all others wins.

The ballot (Equal Vote) — note Carmen and David share 2nd place: unlike RCV-IRV, Ranked Robin lets you rank candidates equally.

Ranked Robin is a brand name, not a new method. Peel it apart and only one layer is actually new:

Layer What it really is Since New?
Compare every pair head-to-head; whoever wins the most matchups wins Copeland's method — the count itself first described 1299 (Ramon Llull), rediscovered by Condorcet in the 1780s, formalised by Copeland in 1951 no
Ties broken by summing win margins Copeland//Borda — a standard, long-known construction, proposed as Equal Vote's default by Sass construction long known; 2021 as the default no
The words "Ranked Robin" Equal Vote's branding, coined by Sara Wolk 2021 yes — the name is the only new part

What it does is take the same ranked ballot RCV-IRV uses and count it a completely different way. Instead of eliminating candidates round by round, it runs a round robin: every candidate against every other, like a sports league. Because every ballot is read in every matchup, nothing is ever discarded — and whenever some candidate beats all rivals head-to-head (a Condorcet winner), Ranked Robin elects them.

Which name means what — Condorcet the family, Copeland the count, Ranked Robin and Consensus Choice two campaigns' brands for it — has a page of its own: the naming decoder, which also carries where the name came from, what its coiner says it means, and how Equal Vote's own wording has moved. Which one we ought to lead with, argued both ways → What should we call this method?.

This page is the folder's front door: the method, one worked election, and the index of runnable examples below. The full concept treatment lives next door — Ranked Robin — the method (mechanics, names, family), Why Ranked Robin (the positive case), honest limits (where it struggles), and the full concept index.

How it counts — a worked election

Thirteen voters, four candidates on a left→right line: Ada (left), Ben (center-left), Cara (center-right), Dan (right). Each row is a bloc of identical ballots.

4 : Ada > Ben > Cara > Dan
4 : Dan > Cara > Ben > Ada
3 : Ben > Cara > Ada > Dan
2 : Cara > Ben > Dan > Ada

Count first choices and Ada and Dan lead with 4 each; Ben has only 3. Under Choose-One plurality, one of the two poles wins. Now compare every pair head-to-head instead:

Round-Robin — every pair, head-to-head (For – Against):
   Ben   beats Ada    9 – 4
   Cara  beats Ada    9 – 4
   Ada   beats Dan    7 – 6
   Ben   beats Cara   7 – 6
   Ben   beats Dan    9 – 4
   Cara  beats Dan    9 – 4

Win–loss record — Copeland score = wins + ½·ties (highest score wins; ties broken by total margin, then lot order):
    #  Candidate  W–L–T  Copeland  Margin  Beats
    1  Ben        3–0–0         3     +11  Cara, Ada, Dan
    2  Cara       2–1–0         2      +9  Ada, Dan
    3  Ada        1–2–0         1      -9  Dan
    4  Dan        0–3–0         0     -11  —

Winner — Ranked Robin (RCV-RR): Ben
   beats every opponent head-to-head — the Condorcet winner.

Ben wins 3–0. A majority prefers him to each rival, one on one — so he's the Condorcet winner, and Ranked Robin elects him. The lesson: RR elects the consensus candidate, not the largest faction's favorite. Nobody had to be eliminated, and no ballot went uncounted.

(Honest footnote: RCV-IRV elects Ben here too — Cara is eliminated first and her ballots flow to Ben. This election separates Ranked Robin from plurality, not from IRV. For the case where RR and IRV genuinely part ways, see the Tennessee center squeeze in the examples below.)

Want the whole count — the pairwise matrix, the Smith set, the audit trail? → the full report: ranked_robin_consensus_center.md · run it yourself: .yaml

How it differs from RCV-IRV

Same ranked ballot, opposite counting philosophy — read the whole ballot against everyone, or eliminate until someone has a majority of what's left:

RCV-IRV (Hare) Ranked Robin (RCV-RR)
Ballot Rank, no equal ranks Rank, equal ranks allowed
How it counts Eliminate fewest-first-choices, transfer, repeat Compare every pair; most head-to-head wins
Uses your lower ranks? Only after higher ones are eliminated Always — every ranking counts against every opponent
Elects the Condorcet winner? Not always (can center-squeeze) ✅ Yes, when one exists
Monotonic? ❌ No ✅ Yes
Precinct-summable? ❌ No ✅ Yes (add pairwise matrices)
Exhausted ballots? Possible No — every ballot is read in every pairwise contest

(The canonical, fuller version of this table — plus the cycle question, the naming family, and the sourcing — is on the method page.)

The worked examples

Runnable elections, each isolating one idea. Tabulate any of them yourself.

Where What
The worked intro — the consensus center wins the full report for the election above: Ben beats every rival head-to-head and wins 3–0, though Ada and Dan each hold more first choices (folder)
Condorcet vs. Ranked Robin — worked examples a clean Condorcet winner, a genuine cycle (rock/paper/scissors) and how RR resolves it, and a real 0-wins record
RR vs. IRV vs. plurality — same ballots one ranked ballot set, three winners — the Tennessee center-squeeze (BV-backed, triple-checked: LH / BetterVoting / pref_voting)
Tiebreaks — dead heat → lot the Equal Support column, the ½-Copeland credit, and the full ladder to lot order — and where the LH & BetterVoting tiebreaks diverge
The Copeland score — a draw is worth half a win why "most head-to-head wins" is a shorthand and not the rule: a single ½-credit elects a candidate who ties on raw wins and lost a matchup — and puts Copeland alone against every other Condorcet method, which is the C1/C2 divide made visible
Most matchups won ≠ Condorcet winner the claim-check companion: 18 voters, no drawn matchups at all, and the candidate with strictly the most wins still loses a head-to-head. Every margin is 12–6, so Minimax / Ranked Pairs / Schulze / Split Cycle all tie five ways and only Copeland decides — the mirror image of the case above
Ranked Robin vs Consensus Choice — two winners, one cycle the two rival "consensus" brands on identical ballots: Copeland ties all three, and Equal Vote's margins rule elects Ana while Better Choices for Democracy's "Most Wins, Smallest Loss" elects Celia — the entire practical difference between the two proposals, made visible
Burial — RR's signature wart, worked the sincere/buried pair (BV2208/BV2209): fifteen voters rank the Condorcet winner last, manufacture a cycle, and win the record tie — triple-checked, deterministic on both engines
STAR vs RR — 30 divergence samples an auto-generated dump of 30 elections where STAR and RR elect different winners, spread across candidate field (3/5/7/10), electorate size, and grouped-vs-random structure — each YAML states its own cause (cycle vs dark horse), with RCV-IRV / Approval / Plurality on the same ballots (they scatter — no clean alignment). Empirical companion to the simulation

Same ballot, different count: RCV-IRV (elimination rounds) lives in other methods and inside the comparison sets.

Conversation scripts: the Larry ↔ Adam series (STAR + RCV-IRV) is indexed in Conversation scripts — index.

file: README.md