💡 In simple words: Some people say that voting by ranking your choices (called Condorcet voting) is too complicated for real life. But real groups have used it for many years. This page tells the true story with numbers: how many people voted this way in the communities that build Wikipedia and the Debian computer system, what they decided, and how often Swiss citizens say yes when they vote on an idea proposed by other citizens. Every number points to its source, so you can check it yourself.

🎯 In 20 seconds (expert summary): Documented history and results of Condorcet-style voting in large online communities, with every figure pointing to a verifiable source. Wikimedia used the Schulze (Condorcet) method for its Board of Trustees elections of 2008, 2009 and 2011 (3,019 valid ballots in 2008, 2,940 in 2009), then moved to support ratios (2013 to 2017) and to the single transferable vote (2021, 6,873 votes). In the documented record, Condorcet voting at Wikimedia served to elect people; the rare movement-wide votes on subjects used yes or no ballots (licensing update 2009: 17,462 certified votes, adopted; Movement Charter ratification 2024: 2,446 voters, not adopted), and no exhaustive count of Condorcet-based decisions exists, which this page says plainly instead of inventing one. Debian provides the documented example of Condorcet decisions on subjects: its public vote archive lists 74 votes from 1999 to 2026, 29 annual leader elections and 45 votes on subjects, counted with a Schulze-equivalent rule since 2003 and fully recomputable. As a comparison point from direct democracy, Swiss Federal Chancellery data counts 26 popular initiatives accepted out of 235 put to a national vote since the instrument was introduced in 1891, about one in ten. Participation and critiques are covered with numbers first, then with verified research.

Condorcet voting in practice: history and results

Status: documentation page. The facts and figures below were verified online against the cited sources in July 2026; figures naturally evolve as new votes take place. The conclusions drawn for WikiDeal are initial hypotheses. This page does not reinvent research: it summarises and points to sources, and where a statistic simply does not exist, it says so instead of estimating one.

What can be counted, and what cannot

Two different questions hide behind "how much has Condorcet voting been used": votes to elect people and votes to decide on subjects. For the Wikimedia galaxy, there is no central registry of every vote ever held across hundreds of projects and languages. What does exist, and what this page reports, is the documented record: the election pages on Meta-Wiki (with full results since 2004), the SecurePoll voting infrastructure, and, as the richest archive of Condorcet decisions on subjects, the Debian voting archive.

Wikimedia: Condorcet was used to elect people

The Wikimedia movement documents its Board of Trustees elections publicly on Meta-Wiki. In that record, the Schulze (Condorcet) method was used for three Board elections:

2008: 15 candidates, one seat, 3,019 valid ballots, tallied with the Schulze method as stated in the election rules. The results page publishes the full pairwise comparison table and a dump of the ballots, so anyone can recompute the count. The published table even shows a real preference circle among three mid-ranked candidates (Heiskanen preferred to Postlethwaite 841 to 770, Postlethwaite to Saintonge 797 to 769, Saintonge to Heiskanen 745 to 737): the Condorcet paradox happened in a real election, without affecting the outcome, since the winner beat every other candidate head to head.
2009: 18 candidates, three seats, 2,940 valid ballots, with the full pairwise matrix and ballot dump on the results page.
2011: the results page again publishes a full pairwise victory matrix.

The method then changed, which is itself instructive: the 2013 election used support ratios (each voter marks support, neutral or oppose for each candidate), and in 2021 the Board election moved to the single transferable vote (4 seats, 19 candidates, 6,873 votes, Meek counting with a quota). A complete registry of every use of ranked voting across all Wikimedia projects does not exist, and this page does not guess one; the three elections above are the movement-level uses documented on the pages cited.

Wikimedia: subjects were decided mostly without ballots, and never by Condorcet at movement level

How many decisions on subjects were taken by Condorcet ballots in the Wikimedia galaxy, and how many were accepted? To our knowledge, no source documents a single one at movement level. Everyday decisions are taken by consensus and structured discussions, and the rare movement-wide votes on subjects used simple yes or no ballots:

The 2009 licensing update (whether to add the CC BY-SA license alongside the GFDL) is the largest documented movement-wide decision vote: 17,462 certified votes, 75.8 percent in favor, published on the result page; the change was then approved by the Board. Notably, this decision vote drew far more voters than any Board election of its era.
The 2024 Movement Charter ratification is the most recent one: 2,446 individual voters, 73.30 percent support among those who expressed a choice, and 83.78 percent support among participating affiliates, as published on the results page. The Board of Trustees nevertheless voted not to ratify the charter, so it was not adopted.

If a documented track record of Condorcet voting on subjects matters for WikiDeal, the reference is not Wikimedia but Debian.

Debian: the documented archive of Condorcet decisions

The Debian voting archive publishes every project-wide vote since 1999, each with its ballot, tally and outcome. As of July 2026 the archive lists 74 votes between 1999 and 2026:

29 annual Debian Project Leader elections (electing a person);
45 votes on subjects: 34 general resolutions plus the original constitution, constitutional amendments, logo choices and a few superseded proposals.

Since a 2003 constitutional amendment, all these votes are counted with cloneproof Schwartz sequential dropping, a Condorcet method equivalent to the Schulze method. "How many were accepted" has a nuance here: many Debian ballots choose among several options rather than saying yes or no, and every ballot includes a built-in default option (such as "further discussion") that a proposal must beat. The winning option of every vote is published on its page, with tallies that anyone can recompute; well-known examples include the 2019 general resolution on init systems and the 2022 general resolution on non-free firmware. Averaged over the archive, Debian holds roughly one leader election plus one or two subject votes per year: even the community with the best-documented Condorcet practice votes rarely.

Participation, resistance and critiques

The numbers first, all from the pages cited above: 3,019 ballots (2008 election), 2,940 (2009), 6,873 (2021), 17,462 (2009 licensing decision), 2,446 (2024 charter ratification). Participation in movement-wide votes has thus ranged from a few thousand to under twenty thousand, in communities whose contributor base is much larger; the 2013 election report also documents operational problems with the voting infrastructure, a reminder that election tooling needs maintenance.

Resistance to voting is documented inside the communities themselves. Meta-Wiki hosts the essay Polls are evil ("don't vote on everything, and if you can help it, don't vote on anything"), and the English Wikipedia community insists that polling is not a substitute for discussion. The 2024 charter episode documents a different kind of friction: a community vote in favor, followed by an institutional decision not to ratify, both published side by side on the same results page.

Verified scholarship analysing these processes in depth (full references with links on the research page):

Reagle (2010), Good Faith Collaboration, a book grown out of a doctoral dissertation, with the full text online, describes Wikipedia's consensus culture and its deliberate wariness of voting.
Konieczny (2009) and Shaw and Hill (2014) evaluate whether power concentrates in wiki communities, with contrasting findings.
Leskovec, Huttenlocher and Kleinberg (2010) analyse thousands of Wikipedia adminship votes and show how social dynamics shape outcomes.
O'Mahony and Ferraro (2007) follow how Debian built and revised its electoral governance over years.
Schulze (2011), the founding paper of the Schulze method, documents its adoption by free software communities.

One honest limitation: no doctoral thesis or peer-reviewed article specifically evaluating Condorcet voting inside the Wikimedia movement was found when this page was written; the works above are the closest verified scholarship, and pointers to better sources are welcome.

A comparison point: Swiss popular initiatives

The Voting portal main page describes the Swiss popular initiative as an inspiration for expressing the degree of stakes. The Swiss record puts numbers on what such an instrument yields. According to the data published by the Swiss Federal Chancellery (pages on the chronology of popular initiatives and accepted popular initiatives, available in German, French and Italian, built on the Chancellery's open dataset), 235 popular initiatives have been put to a national vote since the instrument was introduced in 1891, and 26 have been accepted: an acceptance rate of about 11 percent, roughly one initiative in ten (figures as of July 2026).

The reading proposed for WikiDeal (initial hypothesis): a low acceptance rate is not a failure of the instrument. It reflects a deliberately high bar for changing fundamental rules, while still letting any group of citizens force a public debate and a binding vote. A wiki community using initiative-style decisions should likewise expect most proposals not to pass, and should value the debate they trigger as a result in itself.

Sources

All sources verified online in July 2026:

Wikimedia Foundation elections, master list and per-election pages: overview, 2008 rules, 2008 results, 2009 results, 2011 results, 2013 results, 2021 results.
Wikimedia movement-wide decision votes: licensing update 2009, Movement Charter ratification 2024.
Voting infrastructure: SecurePoll extension.
Community essays: Polls are evil, Polling is not a substitute for discussion.
Debian: voting archive, 2003 constitutional amendment, init systems general resolution, non-free firmware general resolution.
Switzerland: Federal Chancellery pages on popular initiatives, their chronology and the accepted ones; explanations in English on ch.ch.
Reagle, Joseph (2010). Good Faith Collaboration: The Culture of Wikipedia. MIT Press. Full text online. Other academic references: see the research page.