Skip to content

Glossary — Ranked Robin & the Condorcet family

Method-specific terms for Ranked Robin (RCV-RR) and the broader Condorcet family. Shared, cross-method vocabulary lives in the main glossary.

  • Ranked Robin (RCV-RR) — a Condorcet tabulation of a ranked ballot: compare every pair head-to-head, elect the most head-to-head wins (Copeland-style; ties broken by total margin). No center squeeze — do not lump it with IRV. The name is the newest thing about it: coined 2021, for an algorithm from 1951, of an idea from 1299. → Ranked Robin (RCV-RR / Copeland); every other name it travels under: the naming decoder
  • Round-robin voting (aka paired comparison, pairwise, tournament voting) — plain-descriptive names for the whole family, not for this method. Everyone plays everyone; the names differ, the format doesn't. Use one of them — or "Condorcet" — whenever you mean the family, because "Ranked Robin" is one branded member of it. Watch the abbreviation: in this repo RR* always means Ranked Robin, never "round robin." → the naming decoder
  • Llull method (aka Llull voting) — the same algorithm as Copeland, credited to Ramon Llull, who described it in Ars Electionis (1299), rather than to Copeland, who arrived at it independently in 1951. Rare, but not fringe — Wikipedia's article opens "The Copeland or Llull method." Note that Llull's own scoring differed (see Copeland, below). → Ramon Llull
  • Instant Round-Robin Voting (IRRV) — a separate rebrand of the same family, and ambiguous enough to pin down before you rely on it: electowiki defines one specific method (ranked ballots, approval ballots as the tiebreak), Wikipedia redirects the term to the family, and the Condorcet Canada Initiative flies it as a banner; it is also used loosely for Schulze. The same shrink-or-swell problem as "Ranked Robin," one letter over — which is the argument for saying Copeland when you need to be exact. (Community-wiki and advocacy sources; cited for usage, not as referees.)
  • Ranked (ordinal) ballot — the ballot Ranked Robin reads: order over degree of support, shared with RCV-IRV and STV (the term is defined once in the main glossary). Unlike IRV's strict ballot, Ranked Robin naturally allows equal ranks (weak ranks) — see Strict vs. Weak Ranks — Not All Ranked Ballots Are the Same and scored vs ranked Scores vs. Ranks — Don't Confuse Ranks and Ratings.
  • Condorcet method — any ranked method that always elects the candidate who beats every other head-to-head (the Condorcet winner) when one exists. A family: Ranked Robin, Ranked Pairs, Schulze, Minimax, Copeland.
  • Condorcet cycle — a rock-paper-scissors situation (A beats B beats C beats A) with no Condorcet winner; each Condorcet method resolves it differently. → Cycle Resolution — why Minimax, Ranked Pairs, and Schulze exist
  • Copeland — Ranked Robin's tabulation. Two conventions are both in use: wins − losses, and wins + ½·ties (what the engine prints, and what BetterVoting and pref_voting score). They are affine transforms of each other, so they always give the same ranking — but neither equals the raw win count, which ignores draws and can disagree with both. Llull's original was a third convention, 1/1/0 — a draw scored a full point for both, i.e. wins + ties. That one is not an affine transform of the other two (a candidate's tie count varies), so it can genuinely rank differently: in a four-candidate field, two wins and a loss ties one win and two draws under 1/½/0 (both score 2), while Llull's scoring puts the drawer ahead 3–2. Check which convention a source means before comparing tables. → worked case a draw is worth half a win
  • Ranked Pairs (Tideman) — lock in the strongest pairwise victories first, skipping any that would create a cycle.
  • Schulze (beatpath) — decides via the strongest "beatpaths" between candidates.
  • Minimax (Simpson–Kramer) — elect the candidate whose worst pairwise loss is smallest.
  • Borda — a positional ranked method (points by rank); ranked but not Condorcet-compliant. Contrast with Ranked Robin, which is often mistaken for it: Borda scores rank numbers; Ranked Robin uses only each ballot's order to decide pairwise winners — see a blank is ranked last.
  • Blank / unranked (ranked last) — a candidate you leave off a ranked ballot is placed below every candidate you did rank (not a specific tier like "5th" or "6th" — the numeric label is irrelevant to the count). Multiple blanks are tied with one another (Equal Support). → a blank is ranked last
  • Bucklin (Grand Junction) — a ranked, median-style method; ranked but not Condorcet. (Spelled Bucklin, not "Buckling".)

The shared criterion terms — Condorcet winner / loser / efficiency / compliance, monotonicity, summability — are defined in the main glossary (they apply across methods).

file: glossary_ranked_robin.md