Skip to content

Condorcet methods — a reading list, with the nomenclature fixed first

The Condorcet literature is genuinely hard to enter, and the reason is almost never the math — it's the names. "Condorcet method," "pairwise method," "round-robin voting," "tournament solution," "Copeland," "Ranked Robin," "Consensus Choice" all show up in the first ten minutes of reading, at four different levels of generality, from four camps with four agendas. This page is the way in: one taxonomy that makes the names stop sliding, then the sources worth your time, each with its lean marked.

Level: reference · deep dive (a 201/301 aid). Start with the naming decoder if you just need the brands untangled; come here for the literature behind it.

First, the one idea that fixes the nomenclature

Before any book: "Condorcet method" is not a method. It's a property — elect the candidate who beats every other head-to-head, whenever such a candidate exists — and dozens of different algorithms have it. The literature's own organizing scheme for those algorithms comes from Peter Fishburn (1977), and it sorts them by how much of the pairwise data the rule actually needs:

Fishburn class What the rule reads Members you'll meet Where in this repo
C1 only who beat whom (the tournament — a matrix of wins/losses, margins discarded) Copeland, Smith set, Schwartz set, Uncovered set, Top Cycle Ranked Robin (a Copeland variant) · Smith set
C2 the margins too (how badly each pair went) Minimax, Ranked Pairs, Schulze (beatpath), Kemeny, Split Cycle cycle resolution
C3 more than the pairwise matrix — the rule must go back to the ballots and iterate Dodgson, Young the ranked-ballot zoo

Once that ladder is in your head, most of the confusion resolves itself:

  • "Condorcet" / "pairwise" / "round-robin voting" all name the family (the property), at different levels of formality and friendliness. "Round-robin voting" is a recent Wikipedia reframing of what the field has always called Condorcet methods — which is exactly why those two articles read as redundant.
  • "Tournament solution" is the academic name for the C1 tier specifically — the theory of what to do with nothing but a win/loss graph. If a paper says "tournament solution," it is not talking about the whole family.
  • Copeland is one C1 rule (most head-to-head wins, ½ per tie). Ranked Robin is a brand for Copeland plus a defined cycle tiebreak. Consensus Choice is a further brand. None of these is a new tier of the taxonomy — they're leaves.
  • The famous disagreements inside the family are all about cycles. When a Condorcet winner exists, every rule above elects the same person. C1 vs. C2 vs. C3 only matters when the pairwise results loop — which is why so much of the literature is about a case that is empirically rare. Keep that proportion in mind while reading; see cycle resolution and the repo's divergence ledger.

The reading trick: whenever a source names a method, ask which tier? A claim proved for C1 rules ("depends only on the tournament") does not transfer to Minimax, and a Minimax result does not transfer to Copeland. Half of all cross-talk in online voting arguments is a tier mismatch.

Free, online, start here

Source Why The lean
Voting Methods — Eric Pacuit, Stanford Encyclopedia of Philosophy The best free survey in existence, and scrupulously consistent about terms. Read this before any wiki: it defines the family, the criteria, and the individual rules in one vocabulary, and it's maintained. Neutral / academic. Sells no method.
pref_voting — Wesley Holliday & Eric Pacuit The nomenclature as runnable code. Every Condorcet rule under its standard name, definition in the docstring, and the modules are literally organized by Fishburn's tiers (c1_methods, margin_based_methods, iterative_methods). The library is already declared in this repo — see cross-checking with pref_voting. Neutral / academic.
Handbook of Computational Social Choice (the complete book, free, hosted by co-editor Ariel Procaccia — CUP 2016, ISBN 9781107060432), ch. 3 "Tournament Solutions" — Brandt, Brill & Harrenstein The modern restatement of the C1 tier, rigorous and self-contained. Where "tournament solution" is properly defined. Read it here first: tournament solutions is the repo's map of the chapter, with two of its figures turned back into runnable ballots. Neutral / academic. Dense.
Wikipedia: Condorcet method Fine for the neutral family term and notability. But note the round-robin voting umbrella rename is contested and has seen editorial churn — part of why the family looks unsettled from outside. NPOV pressure, but unstable on this topic.
electowiki The only place many branded variants (Ranked Robin, Consensus Choice) are defined at all. Cite for definitions. Reform-advocacy-adjacent. Weak for verdicts — never cite it for a criterion claim.

The papers that define the words

  • Peter C. Fishburn, "Condorcet Social Choice Functions," SIAM Journal on Applied Mathematics 33(3), 1977, pp. 469–489. The source of the C1/C2/C3 classification above, and of "Condorcet social choice function" as a precise term. If you read one primary source for vocabulary, read this one. The lean: neutral; pure taxonomy.
  • H. Peyton Young, "Condorcet's Theory of Voting," American Political Science Review 82(4), 1988, pp. 1231–1244. The historical correction, argued properly: Condorcet the man was after a maximum-likelihood estimate of the "correct" ranking, which is closer to Kemeny than to what now carries his name. Genuinely clarifying about why the modern label is a bit of a misnomer. The lean: academic; sympathetic to the maximum-likelihood reading.
  • T. Nicolaus Tideman, "Independence of Clones as a Criterion for Voting Rules," Social Choice and Welfare 4(3), 1987, pp. 185–206. Coins the term, defines the criterion, introduces Ranked Pairs. Pairs with this repo's clone independence in Ranked Robin. The lean: pro-Condorcet (Tideman's own rule).
  • Markus Schulze, "A New Monotonic, Clone-Independent, Reversal Symmetric, and Condorcet-Consistent Single-Winner Election Method," Social Choice and Welfare 36(2), 2011, pp. 267–303. Long, but the definitions section is a de facto glossary for the C2 tier. The lean: the author is the method's inventor and advocate.
  • Wesley H. Holliday & Eric Pacuit, "Split Cycle: A New Condorcet-Consistent Voting Method Independent of Clones and Immune to Spoilers" (arXiv:2004.02350; published in Public Choice 197, 2023). A modern paper whose background sections carefully re-lay the entire Condorcet landscape before adding to it — often the clearest available statement of what the older rules actually do. The lean: academic, proposing their own rule; unusually explicit about trade-offs. → claim-checked in this repo, with the Schulze spoiler tabulated: Split Cycle.
  • Samuel Merrill III, "A Comparison of Efficiency of Multicandidate Electoral Systems," American Journal of Political Science 28(1), 1984, pp. 23–48* — expanded into Making Multicandidate Elections More Democratic (Princeton, 1988). The origin of the *Condorcet-efficiency numbers everyone still quotes: Plurality, Hare/IRV, Borda, Approval and Condorcet rules simulated under impartial-culture and spatial models. Read it before citing any Condorcet-efficiency figure — including this repo's. This library's own measured run independently reproduces the shape of Merrill's impartial-culture results, which is some evidence both are doing the arithmetic right. The lean: broadly sympathetic to Condorcet and Approval methods, and predates STAR entirely — which makes it a useful outside check rather than a friendly one.
  • James Green-Armytage, T. Nicolaus Tideman & Rafael Cosman, "Statistical Evaluation of Voting Rules," Social Choice and Welfare 46, 2016, pp. 183–212. 54 rules tested for manipulability and utilitarian efficiency across five data-generating processes. The empirical counterweight to axiom-only argument — and honest enough to report that Hare (IRV) and Condorcet-Hare come out most strategy-resistant, a result that does not flatter every camp. The lean: Condorcet-sympathetic authors reporting an inconvenient finding — which is why it's worth citing.

Books

Every book below is on the repo's annotated shelf or belongs beside it; leans are stated in the house style.

  • Duncan Black, The Theory of Committees and Elections (1958) — the book that brought Condorcet back into modern analysis, and where the median voter theorem is proved: single-peaked preferences along one dimension guarantee a Condorcet winner at the median. Start here if you want to know why the pairwise criterion looks so natural to this camp — and notice that the guarantee needs one dimension, which is where the argument gets interesting. The lean: Condorcet-sympathetic, foundational. → shelf entry.
  • T. Nicolaus Tideman, Collective Decisions and Voting: The Potential for Public Choice (2006) — the major Condorcet-sympathetic treatise, by the inventor of Ranked Pairs. Works through what we should want from a collective decision before asking which rule delivers it, and defines its criteria as it goes. The single best book if you want the family argued at full strength. The lean: pro-Condorcet, criterion-driven; a valuable opposing-camp read for a STAR-centered library. → shelf entry.
  • Hannu Nurmi, Comparing Voting Systems (1987)* and Voting Paradoxes and How to Deal with Them (1999) — Nurmi's whole method is definitional hygiene: name the rule, name the criterion, show which fails. The right books for a reader whose complaint is "the words keep sliding." The lean:* survey-neutral; Nurmi catalogs rather than advocates. → shelf entry.
  • Iain McLean & Arnold B. Urken (eds.), Classics of Social Choice (1995) — translations of Condorcet, Borda, and Ramon Llull (whose ~1299 round-robin predates Condorcet by five centuries). Reading where the names came from is clarifying in a way no summary is. The lean: historical, neutral.
  • Dan S. Felsenthal & Moshé Machover (eds.), Electoral Systems: Paradoxes, Assumptions, and Procedures (2012) — Felsenthal's opening chapter is a tight taxonomy of single-winner rules and the pathologies each admits, with the terminology pinned down. Excellent as a lookup table. The lean: academic; the editors are critical of several popular rules and say so.
  • Donald G. Saari, Decisions and Elections: Explaining the Unexpected (2001) — read for why methods disagree (his geometric account of aggregation is superb), but know that Saari argues against the Condorcet criterion and for Borda. That makes him a useful adversary here rather than a guide. The lean: pro-Borda, openly. → shelf entry.

Advocacy — read it, knowing where it stands

The Condorcet camp argues well, and the strongest version deserves a serious reading. Ours are marked, too.

Where each idea lands in this repo

You just read about… Run it here
Condorcet winners, pairwise matrices pairwise counting · the math behind Condorcet
Copeland / tournament solutions (C1) Ranked Robin · Smith set
Cycles, and what the tiers do about them cycle resolution · the Condorcet-winner paradox · runnable: the two profiles
Split Cycle, and "immunity to spoilers" Split Cycle, claim-checked — the paper's central claim, run
Clone independence clone independence in Ranked Robin
Why the brands differ the naming decoder · Ranked Robin vs. Consensus Choice
Whether a rated method should chase the Condorcet winner at all three notions of winner · scores vs. ranks
Cross-checking a real count three ways cross-checking with pref_voting